Croatia
Selected Issues and Statistical Appendix
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar
PubMedClose
Publication Date:
12 May 1997
eISBN:
9781451817232
ISBN:
9781451817232
Language:
English
Keywords:
ISCR; CR; foreign currency; broad money; national bank; opportunity cost; privatized bank; U.S. dollar; price level; depreciation rate; deutsche mark
Download PDF (6.5 MB)

This Selected Issues paper and Statistical Appendix deals with the data interpretation problems in the Croatian balance of payments and discusses in detail the overestimation of the size of the external current account deficit in the official statistics. It examines the behavior of the monetary aggregates in Croatia using econometric methods. The structure and performance of the banking sector is also analyzed, including the effects of competition and ownership structure on profitability and interest rate spreads.

Abstract

This Selected Issues paper and Statistical Appendix deals with the data interpretation problems in the Croatian balance of payments and discusses in detail the overestimation of the size of the external current account deficit in the official statistics. It examines the behavior of the monetary aggregates in Croatia using econometric methods. The structure and performance of the banking sector is also analyzed, including the effects of competition and ownership structure on profitability and interest rate spreads.

I. THE BALANCE OF PAYMENTS IN CROATIA: RECENT DEVELOPMENTS, DATA INTERPRETATION PROBLEMS, AND SUSTAINABILITY ANALYSIS2

A. Introduction and Overview

1. This chapter aims to shed some light on balance of payments (BOP) data problems and trends with a view to helping better interpret and analyze external sector developments and to evaluate against this background external sustainability.

2. Important conclusions from the first part of the chapter are that: (i) the size of the current account deficit—which surged from near balance in 1994 to over 10 percent of GDP in 1995 before declining to about 7 percent in 1996—has likely been overestimated in the official statistics by significant margins; and (ii) errors and omissions, which jumped to 8 percent of GDP in 1995, are likely to have peaked in that year.

3. Two main reasons are offered for the overestimation of the current account deficit. The first is that part of errors and omissions have likely reflected sustainable current earnings. It is estimated that this factor contributed at least 1½ percentage points of GDP to the overstatement of the current account deficit in 1995. The second reason is that most of the foreign exchange inflows of unknown origin have been recorded as capital inflows when more of these inflows should in fact have been registered as current receipts. It is estimated that this could have resulted in biasing upward the 1995 current account deficit by about ½ percentage points of GDP, although the extent of overestimation could be as high as 3 percentage points of GDP.3 This overestimation is likely to continue until data from a comprehensive package of surveys become available to replace those on unidentified foreign exchange inflows in the compilation of the BOP.

4. The main identified components of errors and omissions are expected to decline rapidly due to a mixture of exogenous developments (such as increased use of the banking system for remittances and a diminution of unregistered trade credit) and statistical improvements (especially regarding the measurement of foreign direct investment (FDI). The decline in errors and omissions as shown in the most recent available data (through September 1996) is in line with these expectations. Declining errors and omissions on account of improved measurement of current earnings will tend to reduce the corresponding overestimate of the current account deficit.

5. The key conclusion of the second part of the chapter is that reasonably plausible medium-term projections and various alternative scenarios point to a sustainable balance of payments in Croatia, notwithstanding the high current account deficit and negative export growth registered in the official statistics in 1996. In addition to the overestimation of the current account deficit, this conclusion mainly reflects favorable prospects in the tourism sector, which should more than offset any expansion in the trade deficit.4 It is also important to note that export performance in 1996 was, for several special reasons, significantly better than the data on face value would indicate. The various alternative scenarios suggest that the BOP could remain manageable in the presence of adverse shocks, including resumption of limited regional hostilities that would severely impact tourism and private capital inflows.

6. The rest of the chapter is organized as follows. The first part discusses the main data interpretation problems related to errors and omissions and the arbitrary allocation of foreign exchange inflows. The second part focuses on interpreting important recent developments and prospects in key BOP components and presents alternative external sector scenarios.

B. How to Interpret the BOP Data5

7. The extraordinary size of net errors and omissions makes the Croatian BOP unusually hard to interpret. From a relatively modest US$0.1 billion in 1994, errors and omissions increased sharply in 1995 to reach US$1.3 billion, equivalent to about 20 percent of imports or 8 percent of GDP. This increase coincided with a sharp deterioration in the officially-recorded current account from near balance in 1994 to a deficit of over 10 percent of GDP in 1995, and therefore served to “finance” a large part of the recorded current account deficit in the latter year. Significant problems also arise in interpreting the BOP because of the somewhat arbitrary approach to assigning the large flows of foreign exchange cash and checks sold and deposited by residents to various categories of the current and capital accounts for purposes of estimating line items. Gross inflows into these foreign currency accounts amounted to about US$4 billion in 1995, with net inflows totaling about US$0.8 billion.

Errors and omissions

8. An important data source for the BOP is the International Transactions Reporting System (ITRS). Transactions involving international payments are recorded comprehensively by all banks and compiled in the ITRS. If the ITRS were the only data source used, errors and omissions would by definition not arise because the balance sheets of banks must always be in balance. However, the balance of payments relies on other sources of information for a number of important categories, notably for trade, transportation, and short-term credit.

9. As shown in Table 1, the discrepancy in net import data between (i) payments data from the ITRS and (ii) the customs data compiled by the Central Bureau of Statistics (CBS) and used for the BOP, coincides markedly with the direction of movement in errors and omissions in the period 1994–1996. Note that the difference reported by the two sources in export data has gradually narrowed, but ballooned in import data by about US$1.2 billion in 1995. The substantially higher imports from customs data than from payments data point to unrecorded import financing as one of the main reasons for errors and omissions.6 Such sources of finance that are inadequately captured with existing statistical systems could include unregistered trade credit, unregistered foreign direct investment, and underreported workers’ remittances (including income earned abroad that is not repatriated but instead used to pay for imports by drawing down balances held abroad). Accordingly, four main potential sources of errors and omissions have been identified and are discussed below.

Table 1.

Croatia: Difference Between ITRS and Customs Data on Trade in Comparison with Total Errors and Omissions, 1994–1996

(In millions of U.S. Dollars)

article image
Sources: Customs Office and National Bank of Croatia
• Unrecorded trade financing in the form of delayed payments for imports and trade credit with a maturity of under 3 months

10. In view of security risks, trading partners often required advanced payments for exports to Croatia in its early years (1992–94). When the situation normalized, exporters gradually allowed the usual practice of requiring payment within a short period (mainly 30 days) after receipt of the product. To provide an indication of the significance of these developments, one month of delayed payments for imports amounts to US$625 million of trade credit. A rough estimate based on parameters derived from surveys is that unrecorded trade credit increased by US$300 million in 1995 and a further US$100 million in 1996, and thus contributed these amounts to errors and omissions.7 8

• Underestimation of FDI

11. An improved ITRS was introduced on July 1, 1996, which has provided more accurate data and has indicated that FDI has been underestimated. For July, August, and September 1996 (the first three months for which improved data are available), FDI is reported to have been US$138 million, compared with US$111 million for the first six months of 1996 and US$81 million for 1995. While the high inflows in the first half of 1996 are in line with expectations (reflecting the flotation of shares in Pliva and Zagrebacka Banka on the London Stock Exchange),9 the surge under the improved ITRS suggests, along with anecdotal evidence, that there have been large numbers of small-scale FDI transactions which were previously misreported or underreported.

• Payments made from foreign accounts, using hoarded cash, and from unreported foreign earnings

12. There is ample evidence that Croatians hold large savings abroad. A 1994 survey of Austrian banks put these savings at about US$2 billion in Austria alone. Hoarding had reportedly become widespread during the period of hyperinflation. Foreign earnings that do not flow though the banking system also appear very sizable. Noteworthy in this respect is the surge in workers’ remittances in 1996 by an estimated US$230 million (or 1½ percentage points of 1995 GDP) which cannot be explained by an increase in the number of foreign workers, and therefore points to increased use of the banking system for formerly nonreported cash remittances. These pools of money may constitute important sources of funds to finance imports for which payments are not registered.10 In so far as such imports are measured by customs officials, this could explain a substantial part of the residual errors and omissions. This explanation is in line with the surge in errors and omissions in 1995, which saw an increased demand for consumer durables and investment goods following an improvement in the economic and security outlook and enhanced confidence that the government and the banks would ensure that frozen foreign exchange savings accounts held in Croatia would be protected.11

• Advance payments for exports

13. It is likely that payments for exports have been advanced as confidence in Croatia’s ability to deliver contracted exports has increased in line with the improved security situation. This could explain why payments for exports have grown more rapidly than actual exports. Especially in the shipbuilding sector, advance payments play an important role and contribute to positive errors and omissions when orders result in advance payments that outpace actual exports. Thus, for example, customs records in the first half of 1996 show exports of US$101 million. With payments in the first half more in line with annual exports of US$400 million—consistent with the customs data showing exports of ships of about US$200 million in the second half of the year—this factor may have contributed about US$100 million to errors and omissions in the first half of 1996.

14. In sum, these four sources of errors and omissions all point to underestimation of capital inflows. In addition, evidence suggestive of unreported foreign earnings used to finance recorded imports points to an underestimation of current inflows, perhaps on the order of 1½ percent of GDP in 1995 because of unrecorded workers’ remittances alone. Importantly, it is not expected that errors and omissions will continue at the pace registered in 1995 and the first half of 1996. According to survey results, the float from delayed import payments and the stock of credit with an under three-month maturity appears to have reached a more normal size, implying that these sources of finance can now be expected to stay broadly in line with imports and therefore contribute only about US$50 million annually to future errors and omissions. The earlier improvements and the results from an FDI survey to be introduced in the first quarter of 1997 should also help to significantly reduce errors and omissions from underreported FDI. In addition, the payments for imports from foreign saving accounts will decline to the extent that this source of funds is being depleted and that the demand to restore stocks of durable consumer and investment goods has been satisfied. Finally, workers’ remittances through the banking system seem to be replacing previously underreported remittances in cash. Data for the third quarter of 1996 and estimates for the remainder of the year already suggest that errors and omissions will continue to decline from the very high level recorded in 1995.

15. If errors and omissions nevertheless persist at a high level and in line with a continued large difference between registered payments for and actual imports, the source of unrecorded payments is likely to be a more continuous one, such as import financing from workers’ remittances and small scale exports that go unrecorded. Additional survey work in these areas would need to be undertaken.

Foreign Currency Inflows

16. In addition to the issue of errors and omissions, the other key factor that complicates the interpretation of the Croatian BOP concerns the arbitrary allocation in the present methodology of large foreign currency inflows and outflows in the form of cash and checks sold to/bought from and deposited at/taken out of Croatian banks (Table 2). The primary source of inflows consists of deposits in the foreign exchange accounts of residents, followed by purchases of cash and checks from residents, remittances from abroad to resident accounts, and purchases of cash and checks from nonresidents. The gross inflows, at a sizable US$2.8 billion in the first half of 1996, exceeded export earnings over this period. Foreign currency outflows were also large, with cash and check withdrawals the main item, followed by purchases of foreign currency and travelers’ checks, for a total of US$1.9 billion over the same period. Nevertheless, the net flows were quite significant: annualized about 11 percent of GDP.

Table 2.

Main Foreign Currency Flows, 1995–96

article image
Source: National Bank of Croatia

17. As the origin of and underlying reason for most of these inflows is not well known, they are somewhat arbitrarily allocated under the methodology currently used to construct the BOP: about a quarter of the cash, check, and deposit inflows from residents are allocated to current services (travel credits or tourist earnings), and the rest to capital inflows. Remittances from abroad to residents’ accounts and cash and checks purchased from nonresidents are fully treated as current earnings. Circumstantial information and some survey evidence supports the notion that the sizable net inflows reflect in particular the five sources discussed below.

• Nonresident spending in foreign currency

18. Nonresidents (notably embassies, aid agencies, UNPROFOR, UNHCR and IFOR and their stationed personnel) undertake part of their expenditure in foreign currency cash, which is likely to show up in foreign currency deposits or foreign currency sales by residents. Expenditure by the U.N. alone is estimated to have been US$180 million in 1995, with amounts declining sharply in 1996. Preliminary data on changes in the stocks of nonresident accounts for 1996 indicate annual payments of about $600 million, many of which were probably for goods and services but were reflected in the capital account in the official statistics.12

• Tourist earnings

19. Based on overnight stays and survey data on tourist expenditure—which unfortunately is confined to the main tourist season and the traditional tourist resorts—overall expenditure is estimated to have been about US$700 million in 1994, falling to US$444 million in 1995 and rising again to about US$859 million in 1996.

• Unidentified export and transportation earnings

20. With the open borders, notably to Bosnia and Herzegovina, it is likely that Croatian citizens earn substantial sums in unrecorded trade with this area. The main incentive to underreport export earnings is the significant income taxes in Croatia. The current BOP methodology also underestimates the earnings potential of the transport sector. At present, about 10,000 transport companies operate in Croatia, of which only about 100 are captured in the present transportation survey. Very crude estimates indicate that this sector generates US$200 million more in service earnings than reported.

• Workers’ remittances in cash

21. Croatia has a very large expatriate community, and workers often bring part of their earnings home in the form of cash during holidays, or more frequently for daily or weekly commuters. Some of the seasonality (notably during the Easter, Christmas, and summer seasons) in sales of cash by nonresidents and deposits of foreign exchange by residents is due to this factor. Workers’ remittances through the banking system have increased rapidly from US$243 million in the first half 1995 to US$343 million in the first half of 1996. This trend is expected to continue, with the result that there should be a further reduction in (unrecorded) remittances in cash.

• Repatriation of foreign savings

22. Finally, as discussed above, there is ample evidence that suggests that Croatians hold large amounts of savings abroad, including evidence from a survey of Austrian banks. Part of these savings may be repatriated because of the slightly more favorable terms offered in Croatia. Total foreign currency deposits in Croatia (including declining frozen foreign exchange deposits) increased at a very rapid pace, by about US$0.7 billion in 1995 and US$0.8 billion in 1996, with banks placing about two-thirds of the increment with foreign banks.13 14

23. The current BOP methodology records foreign currency inflows (summarized in Table 2) by treating 75 percent of the foreign exchange cash and checks purchased from residents (ITRS No. 701) and 75 percent of deposits into foreign currency accounts of residents (ITRS no. 797) as a capital account item (i.e., as “other capital inflows”). The remaining 25 percent of these receipts are treated as travel credits or tourism inflows. Foreign exchange remittances to resident accounts (ITRS no. 781) are fully treated as current transfers. Similarly, on the debit side, 75 percent of the foreign exchange sales (ITRS no. 700) and 80 percent of the foreign exchange withdrawals (ITRS no. 897) are treated as capital account items (that is, “other capital outflows”). The remaining 25 percent and 20 percent respectively, are treated as travel debits or tourism outflows.

24. There would appear to be two main problems with this methodology. First, the net inflows recorded as tourism receipts clearly reflect inflows other than those from tourist earnings. This issue came rather to the forefront in 1995, which was a poor year for tourism with a virtual halving in overnight stays by foreigners, although official statistics show a mere 11 percent decline in net tourism receipts. Secondly, the fixed percentages have a significant impact on the size of the officially-reported current account deficit, but make it very arbitrary. As Table 3 indicates, if all cash and check remittances (ITRS No. 701 and 797 on the credit side and 700 and 897 on the debit side) were regarded as current account items, the deficit would be about 3 percentage points of GDP smaller in 1995, and nearly 4 percent of GDP (annualized) smaller in the first half of 1996. With the trend in these cash and check items showing a gradual increase, they would appear to reflect mostly sustainable current earnings, rather than a finite repatriation of foreign-held savings in which case a hump or bell-shaped pattern would be expected.15

Table 3.

Impact of Alternative Allocations of Foreign Exchange Flows on the External Current Account, 1995, 1996 H I

article image

The allocation under the existing BOP methodology is 20–25 percent. The allocation is to the category tourism (credits and debits).

25. Estimates have been made (see attachment) to derive the trend part of the flows as a way of estimating the relative size of current earnings in the overall foreign exchange inflows. The results indicate that a larger part of inflows should be classified as current flows than in the methodology now used and imply that the current account deficit in 1995 is overestimated by about 1½ percent of GDP on this account.

26. A fundamental solution for these statistical problems would be to replace the data on the foreign exchange flows by estimates based on survey information on the five main items identified above, with the information on overall foreign exchange flows functioning as a check on overall accuracy. As a first step, and in line with recommendations by the Fund’s Statistics Department, a more extensive tourism survey and extended transportation surveys will be undertaken, starting in early 1997. These will be used to provide more accurate estimates for tourism and transport income. Simultaneously, and making use of ITRS and banking sector information, more accurate estimates of non-resident expenditure by non-tourists need to be made, and improved criteria for non-resident accounts need to be developed. However, before more comprehensive information becomes available, it will be difficult to abandon foreign exchange flows as the core source of data.

C. BOP Sustainability

27. Medium-term projections for the BOP point to external sustainability. A first key reason is the solid prospect for growing tourist income since income from this source at present is still only a fraction of past income. A second key reason is that performance and prospects of the export sector are better than they appear on the basis of U.S. dollar-denominated customs data for 1996. Finally, as noted above, the current account deficit is likely to have been overestimated by a substantial margin. This is reflected in a relatively modest debt build-up. And while the debt-to-GDP and debt service ratios are projected to increase again during the reconstruction phase, they do so after a recent decline and they are expected to remain well within reasonable bounds in the baseline outlook. Thus, external sustainability when evaluated in the context of the overall BOP, and not the trade sector alone, appears good. Following a brief discussion of the prospects for the main balance of payments categories (trade, tourism, current transfers, and capital flows), various scenarios indicate that Croatia’s BOP is manageable in the presence of adverse shocks.

Merchandise trade

28. Export performance in 1996 was disappointing, with exports as measured by customs data declining by 2.6 percent in U.S. dollar terms. Several special factors contributed to this performance. First, the valuation effects of exchange rate movements, which tended to reduce export growth substantially when measured in the U.S. currency. The deutsche mark is a more representative unit of account for Croatian trade, and exports measured on that basis actually increased by 2.4 percent. Second, export shipments through customs have been substantially weaker than payments received for exports, with export payments rising by 8 percent in the first half of 1996 over the same period in 1995, in contrast to an 8 percent decline for customs-cleared exports. About one third of this difference reflects advance payments for ships, whose export in 1997 will therefore rise considerably. The remainder could reflect the substantive decline in reexports of processed goods observed in the data.16 As there is no indication in the price data that profit margins have declined, re-exports may very well pick up in 1997. The relatively poor export performance in 1996 also reflected the slow growth in Croatia’s export markets. According to data from the WEO, Croatia’s export market expanded by only 1.5 percent in terms of U.S. dollars in 1996.17

29. It is anticipated that export performance will pick up over time, supported by the positive effects of structural reforms on competitiveness. In addition, the growth in Croatia’s export market is expected to pick up significantly in 1997, reaching 6.6 percent based on WEO data. Exports should be raised further in 1997 by growing re-exports as well as the orders-indicated rise in exports of ships.18

30. Despite buoyant economic growth, imports increased only moderately in 1996, although growth was higher when imports are measured in deutsche mark instead of U.S. dollars. Imports of investment goods rose by 15 percent, while imports of intermediate and consumer goods combined inched ahead at about 2 percent. A key explanation behind the moderate import growth appears to be that the surge in imports experienced in 1995 (when imports jumped by 44 percent) satisfied at least part of the demand for replenishing stocks of retail and durable goods. This explanation is consistent with the observed decline in 1996 in short-term trade financing associated with stock building, and the decline in errors and omissions in the first nine months of the year which may have been associated with lower imports for consumer durables financed by drawing down savings held abroad.

31. Over the medium term, imports are assumed to resume growth in line with nominal GDP, with an elasticity modestly above unity. Meanwhile, export growth, which is assumed to be two percentage points below import growth in 1997, picks up only modestly thereafter, thus allowing for a widening trade deficit in the medium-term baseline projections.

Tourism

32. After a dismal and disappointing performance for tourism in 1995 following the shelling of Zagreb and the coast, and “Operation Storm” in late summer, tourism nearly doubled in 1996 to 16.5 million overnight stays by foreign tourists. This number is still a far cry from the pre-war peak of 60 million overnight stays and, with its renowned coast, Croatia’s potential tourism growth is surely substantial. The baseline projection is for Croatia’s tourism to grow at a rate which declines from about 30 percent in 1996 to 15 percent in 1997 and afterward, reaching about 80 percent of its former size over a period of 5 years. The upside potential mainly comes from the possibility of attracting more high-paying tourists; the downside risk stems from too-slow restructuring of partly state-owned tourist companies, and regional security difficulties.

Private and government transfers

33. At 1½ percent of GDP in 1995, Croatia received a relatively large inflow of Government transfers for a country with its per capita income. The main reason is cosmetic: in the official BOP classification, about sixty percent of the Government transfers in 1995 were actually pension payments received by Croatian residents who had worked abroad and which should thus be classified as private transfers. With the Croatian work force in other countries substantial and relatively old, such pensions are likely to continue to grow. In line with the trend, growth in real terms is projected at 5 percent. This contrasts sharply with the outlook for genuine government transfers, to aid the reconstruction of Croatia and to deal with the refugee problems. These transfers have diminished rapidly in 1996, and are expected to dry up after the reintegration of Eastern Slavonia has been completed, now scheduled for 1997.

34. The bulk of overall transfers consists of workers’ remittances, which have been increasing rapidly and may continue to do so as a result of an improved banking network and confidence in the Croatia banking system. With unemployment high, with ongoing restructuring and the lay-offs it implies, and wages in several neighboring countries a multiple of those in Croatia, there is added reason to believe that the number of Croatians working abroad and their remittances will continue growing. Conservatively, the overall real growth is projected at a rate of 2 percent, giving very little weight to the unknown impact of a possible further increase in the use of the banking system to remit earnings.

Prospects for capital flows

35. Until recently, Croatia’s access to international capital markets was very limited. Up to 1995, disbursements of medium- and long-term loans net of scheduled amortization were negative. Access then improved rapidly, but initially only with both the government and commercial banks taking loans from foreign banks at relatively short maturities (up to 2 years). The assessment of an investment grade rating for Croatia by three rating agencies (Moody’s, Standard & Poor, and IBCA) in January 1997 paved the way for borrowing at more desirable maturities. Indeed, in February 1997, Croatia was able to tap the eurobond market with an oversubscribed issue of US$300 million in 5-year bonds, and a favorable spread of 80 basis points over comparable U.S. Treasury securities.

36. With access thus established, the overall size of foreign borrowing from private markets over the medium term should be primarily driven by the demand for foreign loans by the Government and the banking sector, provided the overall borrowing stays within reasonable limits, the macroeconomic and political situation remains stable, and the regional situation peaceful. Thus, following the successful Eurodollar bond issue, recourse to the international capital markets is set to increase markedly in 1997, with the budget deficit financed entirely through foreign loans (including those from the IBRD). This is projected to continue for a number of years, when reconstruction demands are still high. However, it would be desirable if foreign borrowing tapered off over the medium term in order to avoid a significant increase in the external debt burden.

37. Regarding other sources of foreign inflows, the prospects for FDI are also good. For example, discussions for major investments in oil and gas exploration are well underway, while plans for large scale investment in infrastructure (notably roads) have been developed. In addition there is further scope for portfolio investment in several large enterprises—however, at this juncture, data on portfolio investment are not separately recorded.

38. With the pipeline of approved loans, especially from the EBRD and IBRD increasing, the scope for multilateral lending, including through fast disbursing structural adjustment credits, is set to increase further. The increase may taper off, however, once the need for adjustment credits diminishes. Suppliers’ credits, guaranteed and non-guaranteed, also increased rapidly in 1995 and especially 1996 and have become an important source of financing. This latter source is not likely to fall as long as imports are growing and regional security risks remain within acceptable limits.

Sustainability analysis

39. In line with the prospects highlighted above, the baseline medium-term outlook is presented in Table 4. This outlook assumes an unchanged method of compiling the BOP, and projects a marked decline in the current account deficit over the medium term to about 3 percent of GDP by the year 2001. While the baseline allows for a further deterioration in the trade balance in absolute terms, this is more than offset by a growing service sector, notably on account of tourism. Errors and omissions are conservatively projected to fall to zero in 1997 (and stay there), and this drop is assumed to have only a minimal impact on either current earnings or imports. At the same time, net non-loan inflows are expected to moderate.19 Reflecting these trends and those in other capital flows discussed above, financing is expected to shift from net non-loan claims to FDI and net medium- and long-term loan disbursements. The emphasis is initially on loans and gradually shifts to FDI.

Table 4.

Baseline Medium-Term Balance of Payments, 1995–2001

(In millions of U.S. Dollars)

article image
Sources: Croatian authorities and staff estimates.

Does not include debt that was excluded from London Club agreement, and claims on international reserves of the former Socialist Federal Republic of Yugoslavia.

Excludes short-term trade related debt.

40. In the baseline scenario, the ratio of total external debt to GDP peaks at about 30 percent, a figure well within a sustainable range. The debt service ratio is projected to increase to about 12 percent in 2001, as the average maturity is seen to decline with rescheduled Paris and London Club obligations being refinanced with loans of a shorter maturity.20 With exports of goods and non-factor services amounting to just over 40 percent of GDP, a stable debt service ratio of 15 percent would be achieved if the average maturity of loans were 5 years.

41. It is clear that there are substantial upside and downside risks in Croatia’s medium-term balance of payments outlook. At the same time, there are a number of factors that cushion the impact of these risks. The downside risks relate to the speed in the response of the export and tourism sector to structural reforms and regional security. The upside risks include the considerable potential for tourism and uncertainties regarding the assumed drop to zero in errors and omissions since they may turn out to continue to be positive. The foreign exchange flows in the form of cash and deposits by residents pose both an upside and downside risk, which the baseline tries to evenly balance. An important cushion is the reserve position banks maintain abroad.

42. Below, two alternative scenarios are developed to illustrate the downside risk (Table 5). The first scenario features a delayed pick-up of export growth to 4 percent in 1997, climbing only gradually to the projected growth rate of 8 percent in 2001. This is offset in the scenario by a modest reduction in imports of on average 1½ percentage points during the coming 4 years. This reduction in imports is less than half the envisaged build-up in reserves over these years and the situation would be quite manageable through appropriate demand management. To ensure the eventual pick up of export growth there would be need for a significant strengthening of enterprise restructuring and labor market adjustment.

Table 5.

Alternative Medium-Term Balance of Payments Scenarios, 1996–2001

(In millions of U.S. Dollars)

article image
Sources: Croatian authorities and staff estimates.

Does not include debt that was excluded from London Club agreement, and claims on international reserves of the former Socialist Federal Republic of Yugoslavia.

Excludes short-term trade related debt.

43. A second alternative scenario assumes a resumption of hostilities in the region, with the negative impact falling primarily on tourism and capital inflows. With tourism revenues falling back to the low levels experienced in 1995, gross inflows of loans drying up, net non-loan inflows unchanged, foreign direct investment falling back to about US$250 million, short-term trade financing declining, a substantial adjustment in the economy would be required. The budget deficit would need to be scaled back drastically in the absence of foreign financing. Imports would have to fall by about 18 percent before resuming growth to keep reserves unchanged from their 1996 level. However, imports would still remain about 22 percent above the levels observed in 1994. The required drastic adjustment in imports would come about mainly because of reduced income induced by lower tourism income and higher taxes (and possibly exchange rate adjustment). While the import compression under this scenario is manageable, moderation in foreign borrowing and prudence regarding the maturity structure of the foreign borrowing—notably the avoidance of short maturities—remains advisable as long as a scenario of regional insecurities can not be ruled out.

44. A third scenario illustrates the upside risk that errors and omissions do not fall to zero by assuming this item will continue at about US$300 million annually and reflect in large part unrecorded current earnings. Higher errors and omissions on this latter account would imply income levels that are higher than envisaged in the baseline outlook, and thus also imply somewhat higher-than-projected imports. Assuming a high average propensity to import with roughly half of the initial impact leaking out, this would lead to an additional reserve build-up of about US$150 million annually. Alternatively, the program of import liberalization could be accelerated.

ATTACHMENT: Estimation of the Permanent and Transitory Parts of Net Foreign Exchange Inflows

45. For the estimation below, a series of “non- tourism inflows” has been constructed which consists of monthly foreign exchange inflows in the form of resident cash and deposits, and non-resident sales of foreign currency minus an estimate of tourism inflows. This estimate of tourism inflows is constructed by multiplying the number of overnight stays of foreign tourists with an estimate of their expenditure based on survey information from 1994, and adjusted for the movement in the price level. It is assumed that expenditure by the first 100,000 stays—a minimum registered in virtually every month and outside the standard tourist season—is double the expenditure indicated by the survey. The motivation for this adjustment is that the survey evidence relates mainly to normal tourists, whereas the year round number reflects high expenditure businessmen, and U.N. and other officials.

46. The inflows show a relatively steady trend except for the period from late May to August of 1995, when the shelling of Zagreb and the coast occurred and “Operation Storm” was initiated. This is confirmed by regressing the non-tourism inflows on a combination of linear and non-linear trends, and a dummy for “Operation Storm” (Table 6). The trend component is, on average, US$69 million higher in the first half of 1996 compared with the whole of 1995. Similar specifications explain about 70 to 80 percent of the variation in the inflows. Such a percentage of the non-tourism inflows could be regarded as reflecting flows of a permanent nature, notably current earnings.

Table 6.

Regression Results for Non-Tourist Inflows

article image
Source: staff estimates

47. On this basis, it can be calculated that about 50 percent of cash and check purchased from and deposited by residents could represent current earnings that are over and above the share of inflows identified in the existing methodology for compiling current earnings in the BOP. Thus, about 75 percent of such flows should be registered as current earnings. As indicated in Table 3, this would entail an overestimation of the current deficit of about 1½ percent of GDP in 1995, and 2 percent in 1996 on this account.

II. An Analysis or the Behavior of Monetary Aggregates in Croatia21

A. Introduction

48. The monetary aggregates in Croatia have grown enormously in the period since end-1993 (Table 1). Nevertheless, inflation has stayed below 4 percent and the exchange rate has been fairly stable, trading between 3.5 and 3.8 kuna per deutsche mark since then. An additional point of interest is the high share of foreign currency in Croatian broad money, with foreign currency inflows so far showing only slight signs of abating. At first glance this high share of foreign currency might suggest that some form of “currency substitution” is taking place in Croatia. In fact, foreign currency appears to be held mainly as an asset (albeit an asset whose value has been depreciating over most of the period considered here), in an environment where there are few other options available for portfolio diversification.

Table 1.

Croatia: Growth rates of monetary aggregates

article image
Sources: National Bank of Croatia

Reserve money consists of the reserves of the National Bank of Croatia and currency. Narrow money (M1) consists of kuna-denominated demand deposits and currency. Kuna-denominated broad money (M2) consists of M1 plus kuna-denominated time and savings deposits (quasi money). Broad money (M4) consists of M2 plus foreign currency denominated quasi money. Required reserves against foreign currency deposits are held abroad.

Data for 1996 are for the first 11 months of the year.

49. In an effort to understand better the behavior of various monetary aggregates in Croatia, and to investigate their usefulness as intermediate targets for program purposes, this paper estimates various money demand functions. The main conclusion is that the monetary aggregates have not grown in a way that is predicted by standard theories of money demand with respect to the effects of changes in economic activity (the scale variable) or the opportunity cost of holding money (the interest rate). Demand equations with these variables do not appear to be satisfactory because of two main findings. First, although estimates of the elasticity of demand for real money balances with respect to activity are roughly consistent with evidence for other industrial countries, in general the results with respect to the interest rate elasticity are of the wrong sign—that is, the estimated demand for money is positively correlated with the opportunity cost of holding it. Second, deviations of estimated money demand from actual observations are highly serially correlated. This suggests that some other, perhaps unobservable, factor is driving movement in money demand.

50. A possible explanation, consistent with both findings, is that a substantial and sustained remonetization has been, and appears to still be, taking place in Croatia in the wake of decades of high and hyper-inflationary conditions. In such a situation, money demand could rise without a drop in the interest rate or a rise in the level of economic activity, resulting in spurious estimates of the output and interest rate elasticities. With poorly estimated and understood money demand, the usefulness of the corresponding monetary aggregates as intermediate targets is severely diminished.

51. The extremely short time period for which consistent data series are available unfortunately limits the econometric procedures which can reasonably be applied and necessitates the use of parsimonious time series models with fairly strong restrictions on the lag structure.22 Thus, this paper does not pursue the error-correction approach, which has been a popular model in the empirical money demand literature since it identifies both the long-run and short-run dynamics of money demand. On a priori grounds, it would be reasonable to expect that a long-run trend, as opposed to a cycle, could not be isolated with only two and a half years of data. In addition, application of this model requires that a stationary error term be obtained, but this was only the case in the empirical work undertaken when the parameters of the long-run money demand function were theoretically implausible.

52. In the second part to this paper, the sensitivity of the shares of various components in broad money (M4) to interest rate differentials is investigated directly and the results indicate that some fraction of their variation can be explained by changes in the relevant interest rate differentials. The findings do, however, also throw into question the validity of including foreign currency-denominated deposits in a measure of broad money. Including such deposits would be reasonable if they were a close substitute for similar deposits denominated in kuna. If this were the case, then the share of kuna deposits in M4 would be highly responsive to the interest rate differential between kuna- and foreign currency-denominated deposits. In fact this is not the case, suggesting that foreign currency deposits may be held primarily for asset diversification.

B. Money Demand: Theoretical Motivation and Empirical Estimates

53. The demand for real money balances is often modeled as a function of: (i) some measure of economic activity or a scale variable; (ii) the nominal interest rate or some measure of the opportunity cost of holding money; and (iii) exogenous shocks to money demand. When money balances bear no interest, the opportunity cost of holding money is given by the nominal interest rate on an alternative asset, say bonds.23

54. In Croatia, there are several factors which complicate the interest rate part of the story. First, almost all deposits are interest bearing, so only the opportunity cost of holding currency can be correctly modeled as a function of just nominal interest rates on other assets and activity. The opportunity cost of holding demand deposits is the differential between the interest rate on a less liquid asset, say savings deposits, and the interest rate on checking accounts. The opportunity cost of holding time or savings deposits, similarly, is the differential between the interest rate on those deposits and an even less liquid asset. In addition, a large fraction of the broad money aggregate is held in foreign currency, bearing different interest rates than kuna holdings. To compare these interest rates, those on foreign currency deposits must be converted into ex ante kuna interest rates. This involves computing expected exchange rate changes against a basket which reflects the currency composition of the foreign exchange deposits.

55. An even more fundamental question to be interpreted in the context of the empirical results is whether the foreign currency component of what is called broad money in Croatia should be regarded as money at all, given that the characteristics of this component may be more like those of a financial asset. Even though foreign currency deposits satisfy the store of value role of money, the transactions role is far less certain. However, foreign currency deposits are only marginally less liquid than similar kuna-denominated deposits and they can be easily converted to liquid kuna, if desired.

56. Broad money (M4) is divided into four components for empirical analysis: (i) currency (C); (ii) demand deposits (DD); (iii) kuna-denominated savings and time deposits or kuna quasi-money (S, D); and, (iv) foreign exchange time and savings deposits, or foreign currency-denominated quasi-money, (S, D). The empirical analysis itself is divided in two parts: (a) an estimation of the money demand functions for three of the monetary aggregates; and (b) an estimation of the how the components of broad money demand are individually affected by interest rate differentials.

Estimating the demand for M4

57. Starting with the broadest monetary aggregate (M4), the basic model is:

M 4 / P = L ( Y , i L , i M 4 , ε ) ,

where M4 = C + DD + S, D + S, D P is the price level, Y is a scale variable, ε is an exogenous shock to money demand, iL is the bank lending rate, and im4 is a weighted average of the interest rates on the components of M4, calculated as

i M 4 = θ C i C + θ D D i D + θ S O D i k + ( 1 - θ C - θ D D - θ S O D ) i F = θ D D i D + θ S O D i k + ( 1 - θ C - θ D D ) i F

58. The θ’ weights, computed as the respective shares of each component of broad money in the total; ic, iD, iK, and iF are the interest rates on currency (trivially 0), demand deposits, kuna-denominated time and savings deposits, and foreign currency-denominated time and savings deposits respectively.

59. The opportunity cost of holding M4 instead of some alternative asset is measured using the return differential against the interest rate on borrowed funds (iL, the bank lending rate) as the return on the alternate asset. This is justified in two ways: (i) the fact that the market for bonds and other financial instruments is so small and still sufficiently undeveloped that the interest rates on these instruments probably do not represent a good benchmark; and (ii) if firms (or households) which borrow fonds fulfil the marginal condition that the return to capital (or durables) is equated to the borrowing rate, then the bank lending rate represents the return to capital (durables) which is an alternate asset to holding broad money. However, rather than using the return differential itself as an explanatory variable, the individual components are allowed to enter the estimated equations separately. This is because the lending rate is only a proxy and imposing the restriction of equal coefficients on each interest rate generally gave a poorer fit, as well as less plausible parameter estimates.

60. The expected exchange rate depreciation that is needed to construct a proxy for the expected return on foreign currency deposits is calculated using a purely statistical, backward looking, time series model of the exchange rate. Such a model should do better than the theoretical restriction implied by uncovered interest parity, since, in the case of Croatia, this restriction implies that systematic and persistent forecast errors have been made over long periods of time: the differential between Croatian and foreign interest rates has consistently been positive, yet the kuna has rarely depreciated. The model that was eventually used measured the value of the kuna against a basket of dollars and deutsche mark assumed to be purchased at the average exchange rates of 1995, where dollars represent 30 percent of the basket at the time of purchase.24 The forecasting model projected the current depreciation rate of the kuna on the lagged depreciation rate and one lag of the inflation rate. Further lags of these variables and lags of various nominal interest rates were found to have no additional forecasting power. Rather than factor the estimate of the expected depreciation directly into the interest rate on foreign currency deposits (iF), it too was allowed to enter separately in the regression that follows. Specifications where the estimated expected depreciation was factored directly into the interest rate generally gave worse results overall.

61. In estimation, it is assumed that the money demand function is in log-linear form (so that coefficients on quantities represent elasticities and the coefficients on interest rates represent semi-elasticities), and that various lags of the left- and right-hand side variables may be important in explaining the current demand for money. Specifically, ignoring constants, and possible linear trends, it is assumed that

ln ( M 4 t + P t ) = a ( L ) ln ( M 4 t - 1 / P t - 1 ) + b ( L ) ln ( Y t ) + c ( L ) i L t + d ( L ) i M 4 t + e ( L ) d e p + ε t

where a(L), b(L), c(L), and e(L) are polynomials in the lag operator, and dep is the estimated expected depreciation.

62. An important empirical question is whether the money demand functions, so estimated, are stable. In this context, stability is not taken to mean parameter constancy but the existence of a stationary error term in the money demand function. The implication of such stability would be that as long as some linear combination of all the variables in the money demand equation were found to be stationary—even if not stationary individually—then a statistically reliable forecast of real balances could be obtained, based on independent forecasts of the right hand side variables.

63. In the event that the error terms from the estimated equation appear to be stationary and thus a stable money demand function can be estimated, a second question is whether the estimates are theoretically plausible. Plausibility affects the extent to which forecasts from the money demand equation can be considered reliable and useful.

M4 Demand—Empirical Results

64. Estimation of the demand function for M4 did not yield plausible results. Several different models were estimated, but the essence of the results was robust to these different specifications, and generally cast doubt on the existence of a stable and predictable demand function for M4. The basic problems with all the estimates can be illustrated by considering the simplest OLS specification that was used: the dependent variable was the log-level of real balances (M4/P) while the right-hand side variables were a scale variable (measured by an average of the indices of industrial production and real retail transactions), the relevant interest rates for M4 deposits and bank credit, and the expected depreciation of the exchange rate.

65. The results were as follows:

article image

significant at the 5% level

significant at the 1% level

66. The elasticity with respect to activity (1.41) was very large and of the expected sign but it exceeded estimates typically found in other money demand studies. On the other hand, the interest rate elasticities on the alternate asset (iL) and on M4 itself had the wrong sign and were significant, suggesting that the demand for M4 goes up when it becomes more costly to hold it and when the reward to holding it falls (a theoretically implausible finding). A further problem with this specification was that the estimated residuals were highly serially correlated.

67. Several alternative specifications were tried as summarized in the attachment to this chapter. However, the implausible finding vis-à-vis the interest rate on the alternate asset was robust across these specifications. The conclusion drawn from the regressions was that if broad money demand was a standard function of activity and interest rates, the results using data in levels could be rationalized either by significant measurement error in the activity data being used, or by large, persistent, shocks to money demand (which would favor the first-differenced specification discussed in the attachment). With regard to the latter, the fluctuations in activity and interest rates have been relatively minor over the sample period in relation to the large observed changes in real balances. Absent autonomous shocks to money demand, and in the presence of plausible output and interest rate elasticities, these developments would be difficult to explain. Without such shocks, the price level would have needed to rise much faster than has been observed to achieve equilibrium in the money market.

68. Once we accept the notion of large autonomous shocks to money demand, it is natural to work with a trend, or a model in first differences. Unfortunately, the results using detrended and first-differenced data suggested insufficient high frequency variation in output and interest rates for identification of the demand function. Another possible source of misspecification across both types of models of M4 is that using the interest rate on bank lending may not be the best measure of the opportunity cost of money. In the case of broad money, however, an alternate measure is not obvious, given the lack of developed financial markets.

69. It would not appear that the underlying model driving the behavior of M4 can be stylized into a simple money demand framework. More complicated influences related to portfolio balance considerations are likely to be at play.

M2 Demand—Empirical Results

70. Next, kuna-denominated broad money (M2) demand was estimated using similar specifications. The variables were the same as those used in the estimation of demand for M4, except with respect to the choice of an interest rate on a reasonable alternate asset, which in the case of M2 was assumed to be holdings of foreign exchange denominated time and savings deposits. The empirical estimates for M2 were more in line with the money demand model than the results for M4: the own-interest rate elasticity had the correct sign and was statistically significant; the point estimate for the alternate asset-interest rate elasticity had the right sign, although it was small and insignificant.

article image

significant at the 5% level

significant at the 1%

However, once again, serial correlation of the error term was a problem, which alternate specifications consistent with the model could not solve. In addition, the results were extremely sensitive to the precise time series representation of the right hand side variables with respect to lags, with neither the signs or the significance of coefficients being robust. Thus while more plausible than for M4, the results for M2 also presented a number of statistical difficulties.

M1 Demand—Empirical Results

71. The last aggregate to be considered was narrow money. In this case the alternate assets were considered to be kuna time and savings deposits and foreign currency deposits (i.e., all quasi-money). The return on quasi-money was calculated as a weighted average of interest rates on the two denominations. Here, as in the case of M4, the sign on the coefficient on the own-interest rate was of the wrong sign; however, it was not significant. The variables of significance were activity and the return on kuna-denominated quasi-money. The latter, however, also had the wrong sign.

article image

significant at the 5% level

significant at the 1%

Estimating component shares in broad money

72. An alternative approach is tried in the second part of the analysis, which investigates the breakdown of M4 into its various components by modeling this breakdown as a function of the component-wise interest rate differentials. It is convenient to assume a recursive structure in how the different components of M4 produce liquidity services, that is, functional separability.25 An example would be if liquidity services were generated by a recursive series of constant elasticity of substitution (CES) functions of currency, demand deposits, kuna time and savings deposits, and foreign exchange deposits of the form:

S = f 0 ( C , f 1 ( D D , f 2 ( S . D , S F D ) ) )

where each of the functions fp is CES in its two arguments. This kind of structure motivates the following models of the breakdown of M4 into its components.

73. The fraction of M4 held as currency is modeled as a function of the average interest rate on noncurrency:

θ C a f n o f i M 4 - C = [ θ D D i k + ( 1 - θ C - θ D D ) i F ] / ( 1 - θ C )

74. The fraction of M4 minus currency held as demand deposits is modeled as a function of the differential between the interest rate on demand deposits and the average interest rate on M4 minus currency and demand deposits (i.e. total quasi-money):

θ D D / ( 1 - θ C ) a f n o f i M 4 - C - D D - i D = [ θ S D 0 i k + ( 1 - θ C - θ D D - θ S D ) i F ] / ( 1 - θ C - θ D D ) - i D

75. Similarly the fraction of M4 minus M1 (quasi-money) held as kuna savings and time deposits is modeled as a function of the differential between the interest rate on these deposits and the interest rate on other quasi-money, namely foreign exchange deposits:

θ S D 0 / ( 1 - θ C - θ D D ) a f n o f i M 4 - C - D D - S D 0 - i k = i F - i k

76. This recursive structure is restrictive in the sense that the demands for each component of M4 may, in reality, be unrestricted functions of the corresponding interest rates on all other components as well as the bank lending rate (iL). However, given limited data (and, therefore degrees of freedom) some strong restrictions are needed. It is also true that the overall demand for money could be a function of all the separate interest rates on the various monetary aggregates, as well as the credit rate, but also because of data limitations, it is convenient and standard to assume that the specification above is adequate.26 Sahay and Végh (1996) construct a continuous time model in which the rate of “dollarization,” which they define as the fraction of interest bearing assets held in foreign currency, is a similar function of the differential between nominal interest rates in the two currencies, and the rate of devaluation of the local currency.27

77. The first regression run was of the share of M4 held as currency on the weighted-average interest rate on non-currency. This regression is motivated by the assumption that the demand for currency relative to the other components of M4 should be inversely related to the interest premium on holding those components.28 The coefficient in the estimated equation was significant and of the right sign: as the interest rate on demand deposits and quasi-money (in kuna and foreign currency) increased the share of currency in M4 declined.

article image

significant at the 5% level

significant at the 1%

78. Second, a regression was run of the share of demand deposits in interest-bearing M4 (i.e. demand deposits plus quasi-money) as a function of the difference between the interest rate on quasi-money and the interest rate on demand deposits. Again the results show a significant coefficient on the interest rate premium which has the correct sign.

article image

significant at the 5% level

significant at the 1%

79. Third, the share of kuna-denominated time and savings deposits in M4 was modeled as a function of the differential between the interest rates on foreign currency- and kuna-denominated quasi monies. The coefficient on the interest rate differential is not significant, though it is of the correct, negative, sign.

article image

significant at the 5% level

significant at the 1%

80. If agents regarded kuna- and foreign currency-denominated deposits as good substitutes for one another this result would be surprising. One would expect that the share of kuna-denominated deposits would be significantly related, and highly sensitive, to the interest differential. The fact that neither of these features is present lends support to the view that agents do not hold foreign exchange primarily for transactions purposes, but rather as an asset for purposes of diversification.

C. Concluding Remarks

81. This study had two main parts. First, it estimated demand functions for various monetary aggregates which, had they been stable and theoretically plausible, would have been useful for understanding their behavior and for gauging monetary policy. Of the three aggregates, both the estimated demand for the broadest (M4) and the narrowest (Ml) aggregate yielded implausible results. The only monetary aggregate which displayed some of the expected properties was the domestic component of broad money (M2), but even here statistical problems were present and the results obtained were not especially robust. One possible explanation for the relative implausibility of the results for M4 is that demand for foreign currency deposits are not transactions based as in a standard money demand framework, bearing out the notion that the main function of such deposits relates more to portfolio diversification.

82. The second part of the study investigated the behavior of the relative shares of the different components of broad money. In this regard, the relative shares of some of its components are related in a statistically significant way to interest rates. The exception, the lack of sensitivity in the share of kuna deposits to the interest rate on these deposits relative to that on foreign exchange deposits, suggests that these two kinds of deposits may not be close substitutes. This argues against the presence of any kind of pervasive “currency substitution” in the sense of a movement away from domestic currency into foreign currency for the purpose of transacting. Indeed, while the demand for foreign currency has been growing, the demand for domestic currency has also been growing, in a non-inflationary manner.

83. Avenues for further research are severalfold. This chapter has focused on the demand for money, and has assumed that comovements among real balances, output and interest rates can be attributed to, or identified with, the money demand function. However, absent monetary policy which exogenously targets the money supply, some of these comovements are due to the nature of the money supply rule and money supply shocks. To formally deal with this problem, a structural vector autoregression (VAR) approach might be useful. The VAR would separately identify money demand, and money supply, and would facilitate an assessment of how monetary policy’s effectiveness in offsetting autonomous shocks to money demand, but such an approach must await a longer data series. A second, interesting, extension of this work would be an assessment of money demand more along the lines of Cagan’s analysis of hyperinflations, in which the expected inflation rate is taken to be the main determinant of money demand. This would be especially relevant to studying the historical period that predates the sample used in this chapter.

ATTACHMENT: Alternative Specifications of M4 Demand

84. This attachment details results for estimating M4 using four alternate specifications. The equation estimated is the same basic money demand equation presented in the main text:

ln ( M 4 t / P t ) = a ( L ) ln ( M 4 t - 1 / P t - 1 ) + b ( L ) ln ( Y t ) + c ( L ) i L t + d ( L ) i M 4 t + e ( d e p ) + ε t

where a(L), b(L), c(L), d(L), and e(L) are polynomials in the lag operator, and dep is the estimated expected depreciation.

85. First, the equation above was estimated using two-stage least squares, and lagged values of real balances, activity, and the interest rates, as well as a constant and a trend as instruments. This might be preferred to the method presented in the main text, given that money demand shocks (the errors in the equation) are likely to be correlated with output and interest rates if they have real effects. Only if the shocks are stationary and these latter variables are nonstationary can OLS yield consistent estimates. Of course, the instrument set described above, may be misspecified, if there is correlation between the money demand shocks and lagged right-hand side variables. The results were similar to those obtained using OLS. For example, the estimated semi-elasticity with respect to the bank lending rate using this method was 0.041 (again positive) and was significant at the 1 percent level. Furthermore, the residuals from this equation were not highly serially correlated.

article image

significant at the 5 percent level

significant at the 1 percent level

86. The second specification included the lag of real balances on the right-hand side, but the results were, for the most part, qualitatively and quantitatively unchanged. The implied elasticity with respect to activity became insignificant, while the credit rate elasticity was robustly positive and the deposit rate elasticity was negative. This specification led to a very good fit, since real balances resemble a unit root process: the coefficient on lagged real balances was 0.91 with a standard error of 0.04.

article image

significant at the 5 percent level

significant at the 1 percent level

87. The above specifications were motivated by the possibility that the error term in the money demand equation might be stationary, even though real balances and activity might be nonstationary or close to non-stationary. Another possibility is that the money demand shock, is, in fact, nonstationary, in which case the money demand equation should be estimated in first differences. When this third specification was used, log first-differences of real balances were found to closely resemble a random walk with drift: none of the differenced variables entered significantly into the equation. Adding lags of the dependent and right hand side variables to the specification did not significantly alter the fit or the point estimates. Put differently, there is insufficient information in the high frequency components of the data to identify money demand functions.

article image

significant at the 5 percent level

significant at the 1 percent level

88. A further possibility is that the money demand shocks are trend stationary. This motivated the inclusion of a trend in the levels regression in the fourth specification used, but did not change the sign of the coefficients on the interest rates while leaving the credit rate insignificant. It also drove activity out of the equation. The trend appeared to explain almost all of the significant movement in money demand, the deposit rate being the only other variable with a significant coefficient (robustly of the wrong sign). Thus, deviations from trend were found to have little useful informational content.

article image

significant at the 5 percent level

significant at the 1 percent level

89. All these specifications confirmed the basic implausibility of the earlier results obtained in the main text with respect to M4.

III. Ownership, Competition, and Efficiency in the Croatian Banking Sector29

A. Introduction

90. This paper attempts to assess the relationship between the nature of bank ownership, age, conditions of competition, and bank size on interest rate spreads, bank profitability, and measures of bank operating costs and bank asset growth. The specific issues to be examined are (i) the extent to which privatized banks are distinct in their behavior from either new private banks or the remaining state-owned banks and (ii) whether regional differences in competition conditions are a significant determinant of overhead cost or bank interest rate spreads. Finally, the paper will briefly review the implications of its findings for the likely results of continued banking system entry by new private banks and for policies to lower interest rate spreads in Croatia.

91. Descriptive statistics on the variables used in this study are presented in a discussion of general trends and measurement issues in Section B. This is followed by some cross section regression estimates that attempt to test narrower hypotheses of bank behavior through examination of the relationships among these variables. The hypotheses are laid out in Section C and the empirical results appear in Section D. The paper concludes with a brief discussion of the implications of the results for banking system evolution and policy in Croatia.

92. The Croatian banking system before the transition was characterized by two national banks and roughly twenty regional banks, each competing in largely non-overlapping regional markets. All banks were controlled by socially-owned (effectively state-owned) entities rather than private owners. Banks were owned and controlled by major borrowers to varying degrees. Starting in 1989, new private banks were allowed to enter the market. Two dozen new private banks entered the market by the end of 1994, mostly in the capital, Zagreb. Simultaneously, older Croatian banks and enterprises began a gradual process of privatization in which banks sought new private owners directly through new share issues at the same time as state-owned shareholders of banks also sought to privatize themselves. As a result of both processes, the new private banks were joined by eight privatized banks by the end of 1994. Eighteen other banks were in majority-state ownership at the end of 199430. New entry into banking was very heavily concentrated in Zagreb, while privatization of older banks occurred in both Zagreb and the other regions. At end-1995, Zagreb customers were served by twenty banks while customers in various regions of Croatia were generally served by one or two older regional banks, which may or may not have been privatized, and two national banks, one of which was privatized. In some cases, generally the more populous regions, a new private bank was also in the market.

93. Most of the descriptive statistics and almost all of the statistical analysis are based on balance sheet data for end-1995 and cash flows for the year 1995. This choice of data is driven by availability of useful information. Prior to 1994, Croatia was characterized by very high inflation and a unresolved situation regarding bank privatization. Although 1994 cash flow data were available, the very recent and sudden end of high inflation raises some question as to whether 1994 cash flows would be representative of “normal” conditions.31

94. Nominal magnitudes from bank balance sheets are presented as ratios to total assets in order to come up with measurements independent of scale as in other papers on banking system competition, structure, and performance.32 Two sources of data sometimes used in other analyses that were not available for this study are line of business cash flow data and stock market data.33 Data on costs and revenues by line of business (e.g., household versus government deposit-taking or commercial versus consumer loans) were not available at all. Information on the market valuation of assets exists but is quite limited; only a few banks have shares that are actively traded in Croatian stock markets and only one bank is fully listed on the Zagreb Stock Exchange. Therefore, this paper relies primarily on balance sheet data because of the absence of alternatives.

B. Competitive Structure and Financial Differences Among Croatian Banks

95. Tables 1 and 2 below show the share of banking system assets classified by type of bank ownership and regional competition conditions. Both classifications require some explanation, particularly in the case of the ownership classification. To capture the relationship, if any, between ownership and bank behavior, three categories of ownership are employed in the tables and estimates: “new private” banks are those private banks established since 1989 with a private ownership structure; “privatized” banks are those older banks with mixed but majority private ownership structures; and “state” banks, a majority of the shares of which are owned by majority-state enterprises.34

Table 1.

Banking System Assets by Ownership and Region: end-1991 through end-19951

article image

Data for 1993 are end-September rather than end-December observations.

Table 2.

Number of Banks by Ownership and Region: end-1991 through end-1995

article image

96. While the ownership of the new private banks is straightforward, the ownership of the older banks, whether privatized or not, requires some understanding of banking system ownership in the former Yugoslavia.

97. As noted above, the banks inherited from the old regime were owned by groups of socially-owned enterprises. This ownership structure was initially continued during the privatization process although sales of new bank shares to household or other buyers diluted the original ownership in many cases. In addition, enterprise privatization has been an ongoing process throughout the period under consideration with enterprises becoming majority privately owned through a gradual process of share sales from the Croatian Privatization Fund. As a result of both processes, the ownership structure of almost all of the old banks has been constantly changing, and the measurement of private bank ownership is dependent upon the measurement of privatization of the enterprises that own bank shares.

98. Privatized banks are those banks with a majority of shares owned by originally private firms, fully-privatized firms, majority-privatized firms, or private individuals as of end-1994.35 These banks include the largest bank measured by deposits (the second largest measured by assets) as well as seven other banks (see Tables 1 and 2). Only a single observation is used for this measure: thus there is no change in the composition of the privatized banks over time; the same eight banks contribute to the “privatized” data for all years. As a result, changes in the asset shares of this group over time reflect only the relative performance of these eight banks as they proceeded on the path to privatization and not changes in the composition of the group as banks move from “state” to “privatized” categories.36

99. The regional categories used to proxy competition conditions also require some explanation. Competition in the Croatian banking system may be characterized as follows: each regional bank competes primarily only with the two national banks and does not compete significantly with other regional banks or Zagreb-based banks other than two national banks. New entry has changed this situation to some degree, but the new entrants have been heavily concentrated in Zagreb—the largest market in Croatia but one which has between one-fourth and one-fifth of the country’s population. This has tended to make Zagreb a more competitive banking market than other regions of the country. Ideally, information on business activity by bank for each region could be used to better measure the degree of sub-market competition, but such data are unavailable and the Zagreb/regional distinction is used as a proxy for markets with relatively strong versus relatively weaker bank competition.

100. The relative average size of the banks and several other characteristics of banks in the various sectors are shown in Table 3 (end-1994 data). New private banks are on the order of one-tenth the size of state and privatized banks respectively. Capital among new banks of all sorts is remarkably high.37 We will see below that capital to a great extent plays the role of deposits for these banks. Old banks, whether privatized or state-owned, have much lower levels of capital adequacy. Note that capital adequacy levels in general seem acceptable, even for the old state banks. Although, at least for the worst banks, potential loan losses may in fact exceed existing capital.

Table 3.

Capital and Employee Productivity at end-1994

article image

101. Efficiency measurement in banking is a complicated issue. As a practical matter, cost ratios and similar input-based measures of efficiency are all that can be readily calculated from balance sheet data. Costs and inputs can be measured in a relatively straightforward manner and compared to assets of various types (as is done below); quality of services is harder to measure. Therefore, what is described here as efficiency refers to a narrow cost efficiency concept and not to a broad concept of economic efficiency that correctly accounts for service quality. Evidence from other economies suggests that this might not be appropriate. For example, Berger, Kashyip, and Scalise (1995) found that non-interest expenses relative to assets were rising in the United States in recent years as a result of increased competition and innovation; bank competition for customers took the form of an increased range of financial services and increased provision of customer services such as automatic teller machines and electronic banking.38 In spite of the increase in non-interest expenses relative to assets, one would not necessarily want to describe such an outcome as a decrease in “true” efficiency in a normative sense given the clear indications of increased service in this case.

102. Two personnel-based measures of banking overhead costs are assets per employee and personnel cost relative to assets. Data from 1994 on both employee productivity measures appear in Table 3. The evidence is mixed: new private banks have lower costs by the assets per employee measure but state banks have lower cost by the personnel costs to assets measure. It seems that the new private banks tend to pay higher wages, possibly to attract experienced personnel away from older banks or to provide greater employee incentives.

103. Data on broader cost measures appears in Table 4. All ownership categories have almost identical ratios of administrative costs to assets while new private banks have slightly higher ratios of total non-interest cost to asset than state or privatized banks. One item that does stand out in table 4 is the much lower administrative and non-interest cost ratios of the two national banks (one privatized and one still in state ownership). These data point to the fact that special attention should be given to the problem of economies of scale.

Table 4.

Banking System Cost/Asset Ratios and Interest Rate Spreads by Market Type, Age, and Ownership in 1995

article image

104. There are marked differences between various classes of banks in terms of interest rates and interest rate spreads. It may be seen in Table 4 that national banks have much lower interest rate spreads than either regional or Zagreb banks and that new private banks have higher spreads than state banks with privatized banks falling in between. Data show that new private banks held an advantage in net interest income as a ratio of total income. Table 5 presents data on interest income, interest rates, and profitability. New private banks tend to charge higher interest rates on short-term loans.

Table 5.

1994 Interest and Return on Assets Indicators

article image

105. The structure of profitability is largely consistent with the interest rate levels. Both new private and privatized banks are profitable by all measures. New private banks are the most profitable by the return on assets measure, but are less profitable by the return on capital measure due to high capital holdings of new private banks. It can be quite difficult for these new private banks to develop a retail deposit base. Opening up a branch network requires large initial investment in both equipment and personnel. Furthermore, the short-run opportunity cost of such a business strategy during 1995 was high. In the short run, it was more profitable for the new, capital-rich banks to lend at high interest rates through the inter-bank money market than to invest in branch networks.

C. Relationships Between Profitability, Efficiency, Competition, and Ownership

106. There are competing influences in the relationship between profitability, interest rate spreads, growth, and overhead cost structure on one hand and ownership, age, competitive environment, and scale on the other hand. The nature of bank ownership structures might be expected to be particularly important. New private banks might be expected to perform better than old, state-owned banks. This superior performance might be thought of as having a variety of causes. The new banks would not be burdened by an existing portfolio of suspect loans inherited from the former regime. Their private ownership would establish incentives that would make it less likely that they would engage in money-losing new lending for political reasons or reasons related to moral hazard inherent in their ownership structure. They would also be expected to avoid excessive overhead costs in the form of over-staffing, excessive wages or real estate or other “empire-building” activities divorced from profitable banking. Older banks that have been privatized might be thought of as an intermediate case, performing better than old, majority state-owned banks because of the improved incentives associated with the transformed ownership but less well than new private banks. Competing influences about the behavior of the new or private banks can be categorized in terms of three stylized types.

107. One hypothesis, combining several relationships might be characterized by describing the new or private banks as “lean and mean”. Support for this hypothesis would be finding a tendency for new private or privatized banks to have higher profits, lower interest rate spreads, and lower overhead-to-assets ratios. Conversely, the hypothesis suggests that older and majority state-owned banks might be expected to be relatively unprofitable, exhibit higher spreads, and have higher overhead cost ratios under this hypothesis. This is arguably the mainstream stereotype of banks (or other economic enterprises) in transition economies: the old regime is characterized by bloat and inefficiency for which cost control and aggressive pricing brought about through privatization and new private entry are the remedy.

108. A second hypothesis would have it that new private banks (but not necessarily privatized banks) are exploiting a niche to engage in what might be characterized as “cream skimming”—taking the most profitable customers on both the deposit and lending sides of the balance sheets by providing higher quality services (and therefore incurring higher costs). These banks might also be expected to be relatively profitable, but, unlike banks under the first hypothesis, they might be expected to have higher interest rate spreads and perhaps modest growth rates reflecting the preference for high rates of return over aggressive expansion. This hypothesis would be consistent with the view that existing bank ownership links have channeled excessive credit to bank shareholders while pushing otherwise creditworthy customers to the margins of the banking system where they would face higher rates due to greater competition for funds. Support for this hypothesis would be finding a positive relationship between new private banks on one hand and higher profits, and relatively high interest rate spreads, on the other.

109. A third hypothesis, closely related to the “cream skimming” hypothesis, might be characterized as “aggressive full service banks” also taking the most profitable customers on both the deposit and lending sides of the balance sheets by providing higher quality (and therefore incurring higher costs). These banks might also be expected to be relatively profitable, but they might be expected to have higher overhead-cost-to-assets ratios. However, unlike banks under the cream skimming hypothesis, they would be expected to have lower interest rate spreads and high rates of asset growth reflecting the preference for aggressive expansion over high rates of return. Such banks might typically be foreign banks (largely absent from Croatia in 1995) or domestic banks seeking low risk foreign or domestic “blue chip” clients. Support for this hypothesis would be finding a positive relationship between new or privatized banks and higher profits, higher overhead-to-assets ratios, and relatively low interest rate spreads.

110. Independent of the differences between these three hypotheses, two factors that need to be controlled for are the influence of economies of scale and competitive environment. First, Croatia is a middle income country with less than five million people but more than fifty banks of widely varying sizes; it is not obvious that all are at or above the minimum efficient scale. Absent a control variable for economies of scale, large but inefficient banks could have their inefficiency masked by the effect of the scale economies. Second, competition in the more competitive banking markets could result in an equilibrium with lower profits, spreads, and overhead cost relative to other markets independent of ownership and age effects. The regional variables described above are used as proxies to capture the effects of regional differences in competition.

D. Empirical Examination of Competition, Scale Economies, and Efficiency

111. This section presents the results of regressions that attempt to assess some of these hypotheses. Two measures of profitability (profits to assets and profits to capital), two cost ratios (non-interest expenses-to-assets and administrative expenses-to-assets), the ratio of interest income to total assets less the ratio of interest expenses to deposits (the “spread” for the purposes of these regressions), and rate of growth are the dependent variables. The independent variables are the natural log of total assets (to capture the influence of economies of scale), the ratio of non-interest expenses to assets, a dummy variable for new private banks (established in 1989 or later), a dummy variable for older privatized banks, and a dummy variable for banks operating in Zagreb. All regressions use 1995 flows and end-1995 stocks for all variables.

112. The sample used for the regressions reported on below excluded those banks with deposits-to-asset ratios under 0.2. Because these banks are not taking significant amounts of deposits relative to their assets, concepts such as interest rate spreads are of questionable relevance and cost ratios need to be considered in the light of having few of those costs associated with depositor services. Some of these banks operating with little or no deposits are in a very early phase of their start-up while others appear to be engaged in something more like investment banking than intermediation between depositors and borrowers. Another anomalous bank, one created to manage the Croatian assets and liabilities inherited from a bank based in Serbia, was also excluded due to its anomalous nature. These deletions reduces the total sample from 53 banks to 48 banks. However, these omitted banks are relatively small and account for a small share of Croatian banking assets.

Profitability

113. The first measure of profitability employed is the ratio of 1995 profits to end-1995 assets. This measure of profitability was regressed on the logarithm of total assets to capture the effect of economies of scale, the ratio of deposits to total assets (to capture the effect of dependence on borrowed resources) and dummy variables for local banks based in Zagreb, new private banks, and privatized banks (Table 6). A second regression was run on the same variables but with the profit to capital ratio as the dependent variable (Table 7). In both cases, restricted regressions were run excluding the least statistically significant variables. The restrictions were statistically insignificant (as measured by the F-statistics) in both cases and the new private dummy variable became positive and significant (by conventional t-statistic measures) while the privatized dummy variable remained positive but fell short of the traditional t-statistic levels for statistical significance.41

Table 6.

Determinants of Profitability Dependent Variable: Profits/Assets

article image

indicates statistical significance at the five percent level

Table 7.

Determinants of Profitability Dependent Variable: Profits/Capital

article image

indicates statistical significance at the five percent level

114. These regressions indicate that private banks, particularly new private banks, are more profitable than state banks when scale effects and the influence of deposit-to-asset ratios are taken into account. Three features of the insignificant variables are worthy of note. First, differences in profitability between new private and privatized banks, which should have been picked up by the new bank dummy, do not appears to be substantial given that the coefficient is low relative to the standard error. Second, the dummy variable for Zagreb, although of a low level of significance, has the opposite of the expected sign. Rather than suppressing profits, the more intensely competitive Zagreb market appears to be associated, albeit weakly, with higher profits. The third feature worth noting is the economic significance of the coefficients on these new private and privatized dummy variables. The dummy on new private banks suggests an additional profit effect for new private banks equal to 2.0 percent of assets and 6.8 percent of capital, almost exactly the size of the mean profit to assets and profit to capital ratios for all banks (2.1 percent and 6.7 percent respectively). The estimated profit effects for privatized banks (1.5 percent relative to assets and 4.7 percent relative to capital) are smaller and statistically insignificant but are also large in economic terms. This suggests that the low statistical significance may be more the result of low sample size rather than small economic effect.

Non-interest expense and administrative cost ratios

115. Conventional wisdom about the banking system in Croatia suggests that the newer private banks are relatively efficient due to low overhead expenses (arguably too low for effective loan evaluation according to some), while the older banks, particularly those that have not yet been privatized are burdened with excessive overhead. The information in Tables 4 and 5 provides some limited support for that view, but the differences across ownership types are small and some measures show higher costs for the new private banks. The apparent lack of difference across ownership types may reflect the offsetting effects of economies of scale tending to reduce the costs of the generally larger state-owned and privatized banks while poor cost control in those banks raises costs.

116. Two dependent variables (non-interest expenses to assets and administrative costs to assets) are regressed on the natural log of total assets (to capture the influence of economies of scale), the ratio of deposits to assets (to capture the higher costs associated with retail banking), a dummy variable for banks operating from Zagreb (other than the two national banks), a dummy variable for new private banks established in 1989 or later, and a dummy variable for older privatized banks.

117. In the case of both non-interest expense and administrative cost ratios, the results are similar. Both regressions show significant economies of scale, and a weaker but positive relationship between the ratio of deposits to assets.42 The regressions on administrative expenses showed a marginally significant positive relationship between both new private banks and location in Zagreb on one hand and higher costs on the other (perhaps reflecting either a need to provide more services in a more competitive environment or higher wages in Zagreb). Otherwise, the ownership, age and locations dummies showed little relationship to the cost ratios in the presence of the other independent variables.

118. These results tend not to support the idea that new private banks or more competitive banking markets will reduce costs in the Croatian case or that older or state-owned banks are suffering from bloated overhead costs. The new entrants may be more efficient in the sense that they are providing a higher level of service with the same cost ratios as other banks. However, it does not seem to be the case that the new private or privatized banks are providing lower cost banking services once the influences of scale and dependence upon deposits are taken into account.

Interest rate spreads

119. Regressions with interest rate spreads as the dependent variable (Table 9) suggest that the regional variables do not have much explanatory power for this variable and the economies of scale is only marginally significant. Nevertheless, it is noteworthy that the relationship between bank size (log of assets) and spreads is positive once other factors are controlled for by the regression in spite of the generally lower spreads at the very large national banks and the relatively high spreads at the small new private banks (see Table 4 above). There is also a positive, albeit statistically marginal, relationship between higher interest rate spreads and privatized banks. However, the size of the coefficient on non-interest expenses gives a strong hint as to why the non-interest expenses variable had no strong correlation with profit ratios: the restricted regression suggest that more than 89 percent of these costs are passed on to customers in the form of lower deposits rates or higher credit rates43.

Table 8.

Determinants of Non-Interest and Administrative Expense

article image

indicates statistical significance at the one percent level

Table 9.

Determinants of Interest Rate Spreads

(Dependent variable: interest rate spread)

article image

indicates statistical significance at the five percent level

indicates statistical significance at the one percent level

120. Similar results, although with weaker statistical significance were obtained with the administrative costs ratio as the independent variable in place of the non-interest expense ratio. These regressions also suggested that there was a strong relationship between private ownership and higher spreads and that most of the non-interest costs were passed on through interest rate spreads. As with the non-interest and administrative cost ratios, it is worth distinguishing between the marginal statistical significance of the ownership dummy variable coefficients and the economic significance of the coefficient values. The parameter values in the restricted regression for both the new private and privatized dummy variables suggest that the effect of either ownership status is in excess of one fourth of the mean interest rate spread.

Growth in assets

121. Regressions on asset growth in 1995 produced only one significant variable, the dummy variable for new private banks. The growth variable is calculated as the change in assets from end-1994 to end 1995 relative to the average of end-1994 and end-1995 assets. The coefficient on the dummy variable for new private banks suggests an additional 49 percent growth in assets for these banks in both the restricted and unrestricted regressions.

Table 10.

Determinants of 1995 Asset Growth

(Relative to the Average of end-1994 and end-1995 assets)

article image

indicates statistical significance at the one percent level

E. Concluding Remarks

Characterizing the private banks

122. These results suggest that Croatian private banks, particularly new private banks, are more profitable, faster growing, and probably have higher spreads than those remaining in state hands. Profits of private banks tend to be higher by an amount roughly equal to median profitability regardless of whether profits are measured relative to assets or capital. The estimates of the effect of ownership status on growth suggest that new private banks grow at a rate that is an additional 49 percent higher than other banks. There was no significant relationship between ownership status, age, or market location in terms of non-interest or administrative expenses once other variables were controlled for, although there were strong diseconomies of scale which would fall heavily on the new private banks because they tend to be smaller than other banks.

123. In terms of the three hypotheses about the behavior of private banks laid out above, the “lean and mean” hypothesis seems clearly inconsistent with these results with its implications of lower administrative and non-interest costs, and lower interest rate spreads. Similarly, the lower spreads argue against the “aggressive full service bank” hypothesis. The “cream skimming” hypothesis looks most consistent with the data; private banks do have substantially higher profits and interest rate spreads than their state-owned competitors.

124. The results on growth suggest that the higher spreads have not prevented growth of the new private banks, although perhaps growth would be still faster if new private banks were to switch to an “aggressive full service” approach. However, as noted above, such an approach may not be cost effective given the costs and time needed to build a retail banking network that could fund faster growth.

Implications for the Croatian banking system

125. New entry by either new private banks or by foreign banks might be thought of as a source of competitive pressure that could reduce the high interest rate spreads observed in Croatian banking. In Croatia, this new entrant role would potentially fall on new private banks as foreign banks were largely absent in Croatia during the sample period. However, it does not seem to be the case in Croatia that new entrants were putting downward pressure on interest rates by undercutting the incumbent banks on pricing; new private banks in Croatia appeared to have higher spreads.

126. Their profitability and rapid growth both suggest that new private banks are successful and aggressive competitors in some sense (particularly notable given their high concentration in Zagreb, Croatia’s most competitive banking market). However, given that they have higher spreads and no lower overhead cost ratios than other banks, their competitiveness would appear to be founded on providing more or better services or serving previously under served markets rather than gaining customers through more aggressive pricing and better cost control.

127. A reduction in interest rate spreads does not seem likely to be the direct, near-term result of new entry by new private banks thus far. However, the absence of a direct effect of new entry in the sense of having new entrants undercutting the spreads of incumbent banks isn’t to say that there has been no indirect effect on spreads as a result of new entry. Older banks may have reduced their spreads relative to what they would be otherwise in order to prevent an even faster erosion of market share.

128. If the new private banks or privatized banks exhaust the high return potential of under-served clients willing to pay high interest rates, their own business strategy might change. Absent a different strategy on the part of new private banks, other influences would seem to be called for in order to reduce interest rate spreads. Bank rehabilitation of problem banks is one possibility; this is already underway but results are not yet evident. The recent rehabilitation and possible future breakup of the largest state-owned bank and the recent rehabilitation of two other large state-owned banks could substantially change the structure of the Croatian banking market and the financial structure of its major participants. The first stages of bank rehabilitation in early 1996 have already reduced interbank interest rates, and it is certainly possible that spreads and lending rates will fall as rehabilitation proceeds. Moreover, foreign banks are only just now starting to enter the Croatian market. Their entry could reduce concentration in the banking sector and bring down interest rate spreads through greater competition. The horizontal breakup of the larger state-controlled banks, for example through privatization, could further invigorate competitive forces in ways that had not been apparent before.

ANNEX I: Summary of the Croatian Exchange and Trade System (Position as of December 31, 1996)

Exchange Arrangement

129. The currency of Croatia is the Kuna, the external value of which is determined in the interbank market. The exchange rates in the interbank market are determined by authorized banks that transact with each other at freely negotiated rates. The National Bank of Croatia (NBC) may set intervention exchange rates at which it will transact with banks outside the interbank market for purposes of smoothing undue fluctuations in the exchange rate. On December 31, 1996, the average interbank market rate for the U.S. dollar was HrK S.S396 per US$1.

130. There are no taxes or subsidies on purchases or sales of foreign exchange.

131. Croatia accepted the obligations of Article VIII, Sections 2, 3, and 4 of the Articles of Agreement on May 29, 1995.

Administration of Control

132. Foreign exchange transactions are governed by the Law on the Foreign Exchange System, Foreign Exchange Operations, and Gold Transactions, which was enacted on October 7, 1993. The NBC formulates and administers exchange rate policy and may issue foreign exchange regulations under this law. A Trade Law (codifying domestic trade and foreign trade legislation in a comprehensive law) was passed on January 31, 1996, and came into force on February 17, 1996. Companies wishing to engage in foreign trade must register with the commercial courts. The representative offices of foreign companies must be registered with the Ministry of Economy.

133. Foreign exchange transactions must be conducted through authorized banks; currently 52 commercial banks in Croatia are licensed to conduct foreign exchange transactions. Restricted licenses are given to banks that may open accounts for resident natural persons and may buy and sell banknotes and checks (currently 8 banks).

Prescription of Currency

134. Settlements between residents and nonresidents may be effected in any convertible currency. Croatia does not maintain any bilateral payments agreements involving exchange restrictions.

Resident and Nonresident Accounts

135. Resident natural and juridical persons may, in principle, open and operate foreign exchange accounts only in Croatia. However, the NBC has the authority to allow resident juridical persons to keep foreign exchange in accounts with foreign banks in order to cover the costs of business operations and meet the requirement of regular foreign trade activities abroad. costs of business operations and meet the requirement of regular foreign trade activities abroad. The law also makes specific provisions for resident juridical persons engaged in capital project construction abroad to maintain accounts with foreign banks, subject to a license issued by the NBC.

136. Nonresidents may open foreign exchange accounts with fully licensed banks in Croatia. These accounts may be credited freely with foreign exchange and debited for payments abroad or conversion into domestic currency; reconversion of domestic currency into a foreign currency is permitted. Juridical persons may credit these accounts with foreign banknotes up to a limit of US$20,000 without special permission from the NBC.

137. Nonresident natural and juridical persons may open accounts in domestic currency with the proceeds from sales of goods and services or with foreign exchange transferred from abroad. They may purchase foreign exchange with funds held in these accounts without restriction.

Imports and Import Payments

138. Croatia does not maintain any import quotas—the existing quotas were abolished by decree of July 12, 1996, although such quotas are in principle allowed for under the Trade Law under conditions similar to WTO rules. All imports except a list of products whose importation is controlled by international agreement for noneconomic reasons (such as, arms, gold, illegal drugs and narcotics, and artistic and historic work) are free from licensing requirements, and a small number of other products (notably iron tubes and bars). The importation of these items is allowed on a case-by-case basis.

139. Imports are subject to a customs tariffs of up to 25 percent, with a few exceptions. The exemption for duty-free imports by travelers is US$100. Goods imported by travelers and postal shipments up to the value of US$500 are subject to a simplified customs procedure with a unified tariff rate of 8 percent. For imports exceeding that value, the regular import tariffs and taxes are applied. Returning citizens may bring into the country household effects duty free in an amount related to the period spent abroad for household effects and for private business purposes without restrictions, but on a case-by-case basis, under the approval of Ministry of Finance. Under certain conditions, goods imported by nonresidents for investment purposes are exempt from import duties. Also, raw materials and intermediate products used in the production of exports are exempt from all import duties and taxes, provided that the value added of the export product is at least 30 percent of the value of the imported items and that export proceeds are received in convertible currency. Payments for legal imports are not restricted.

140. Advance payments for imports are permitted, where down payments are required by suppliers, in accordance with customary international practices.

Payments for invisibles

141. Payments for invisibles related to legal imports by juridical persons may be made freely. Payments of leasing fees are permitted provided that temporary imports have been registered with the Customs Office. Natural persons may also purchase foreign exchange in the interbank market for the payment of goods and services abroad and for deposit in a foreign exchange account for the purpose of future payments. Resident juridical persons (including tradesmen and natural persons engaging in independent activities) may purchase foreign exchange only for authorized payments abroad, except to make payments for activities related to scientific, humanitarian, cultural, or sport events. Payments of royalties, insurance, and legal obligations and contracting of life and casualty insurance policies with foreign companies are also permitted.

142. Resident natural persons may take out of the country foreign currency equivalent to DM 1,000, including for short cross-border trips. There exist no restriction on the frequency with which such amounts can be taken out. An additional amount equivalent up to DM 2,000 may be taken out, provided that it is withdrawn from foreign currency accounts or purchased from banks for travel expenses. In both cases the NBC may allow higher amounts to be taken out on a case-by-case basis. The exportation of Croatian currency by both residents and nonresidents is limited to HrK 2,000 a person.

Exports and Export Proceeds

143. Exports are free of restrictions except for certain products for which permits must be obtained (list D products: e.g., weapons, drugs, and art objects); several basic foodstuffs to ensure adequate domestic supplies; unrenewable resources (oil and natural gas); hides and wood.

144. Export proceeds must be collected and repatriated in full to Croatia within 90 days of the date of exportation; this period may be extended with the permission of the NBC. If payment terms in excess of 90 days have been agreed with foreign importers, the credit arrangement must be registered with the NBC.

Proceeds from Invisibles

145. Proceeds from services are, in principle, subject to the same regulations as those applying to merchandise exports. The importation of Croatian currency by both residents and nonresidents is limited to HrK 2,000 a person.

Capital

146. Resident juridical persons, including commercial banks, may borrow abroad. They are required to register the loans contracted, including commercial credits, with the NBC. Financial credits may be extended to nonresidents by resident juridical persons, only in accordance with the provisions of the Law on Foreign Trade Credit Operations. Natural persons are permitted to obtain loans from nonresidents in domestic or foreign currency. The foreign exchange positions of commercial banks are subject to a limit (maximum up to 30% of the bank capital).

147. Foreign direct investment by nonresidents may take the form of joint ventures or full ownership and must be registered with the commercial courts. Repatriation of capital and transfers abroad of profits are not restricted. In principle, domestic and foreign investment is treated equally (e.g. “national treatment”). If the foreign equity capital participation exceeds 20 percent, inputs used in the project are exempt from import duties. The profit tax rate is uniform and amounts to 35 percent. Foreign direct investment abroad by residents must be registered with the Ministry of Economy within a 30-day period commencing from the signature of the contract. Such investment must generally be undertaken through loans abroad or through reinvestment of profits. Inward portfolio investment is not restricted, except in central bank short-term securities on the primary market. In general, outward portfolio investment is restricted.

148. Nonresident natural persons may acquire real estate in Croatia through inheritance as long as their country of residence extends reciprocal treatment to residents of Croatia. Nonresident natural persons not engaged in economic activities in Croatia may purchase real estate only under the same conditions. Nonresident natural or juridical persons engaged in economic activities in Croatia may also purchase real estate under these conditions and may sell it to resident or nonresident juridical persons. Residents may acquire real estate abroad on the basis of reciprocity of treatment, but they are not permitted to purchase foreign exchange in the exchange market for this purpose; the use of balances in foreign exchange accounts for this purpose is also prohibited.

Gold

149. The National Bank of Croatia may export gold and gold coins without any restrictions. Unprocessed gold may be exported under the approval of the National Bank of Croatia.

150. Gold coins may be exported by authorized commercial banks, under the approval of the National Bank of Croatia.

151. Importation of gold is subject to the approval of Ministry of Economy.

Changes From July 1, 1995 through December 31, 1996

152. December 14, 1995 the Law on Issuance and Sale of Securities was passed and came into force on January 1, 1996.

153. December 14, 1995 the Law on Investment Funds was passed and came into force on January 3, 1996.

154. February 17, 1996 the Trade Law became effective.

155. June 8, 1996 the Law on Foreign Credit Operations became effective.

156. July 1, 1996, the Law on Customs Tariffs (including unification of various and tariffs in a single tariff structures) became effective.

157. July 12, 1996, a Decree Law Export Quotas (including abolition of import quotas) became effective.

158. July 31, 1996 an exchange of bonds in the context of an agreement with London Club creditors was effected.

ANNEX II: Summary of the Croatian Pension System: Current Features and Issues for Reform44

A. Introduction

159. Pension reform is high on the agenda of several industrialized and transition countries. One route that countries are taking is a rationalization of the existing pay-as-you-go (PAYG) pension system through selected parametric adjustments on either the revenue or the expenditure side or both, such as the contribution rate and the pension benefits formula. Croatia has been considering a broader approach, encompassing a more fundamental systemic reform of its pension system by (a) downsizing and rationalizing the existing PAYG system and (b) introducing privately-managed, obligatory and voluntary funded pension schemes.

160. Under an unfunded regime like the PAYG system, current revenue of the pension plan is expected to finance current obligations, and the rate of the payroll tax—the standard financing source for public pension plans—is adjusted periodically to ensure that revenues and expenditures match. Such an approach formalizes an implicit intergenerational contract for younger working generations to pay for the older generations’ retirement benefits. The existing commitments to current and future retirees is an unfunded debt.

161. Under a funded regime, workers save in and invest for retirement via pension accounts. Reserves could be accumulated with a view to equaling the expected benefits payable after retirement to those who work and save in the pension plan now. Contributors, upon retirement, would get back only what they put in plus the return on their investment.

162. Croatia—with the help of the World Bank—envisages a possible switch to a multiple pillar system, under which (1) the current PAYG pension system (first pillar) would be both downsized and rationalized, and complemented by (2) a compulsory, privately-managed savings scheme in which current workers invest for their retirement (second pillar), and (3) an additional voluntary pension scheme under which workers could invest additional amounts for their retirement (third pillar). Under the three pillar concept: the compulsory savings scheme (second pillar) would have the primary responsibility for handling the retirement savings of the workers; the downsized publicly-managed and tax-financed first pillar would have the primary goal of redistribution (to keep old people out of poverty); and voluntary saving plans (third pillar) would exist for people who want more protection than what is offered by the first and second pillars of the pension system.

163. As soon as the two funded pillars have been established, the working population would have to provide for the already retired population under the PAYG system and, at the same time, save in their individual retirement accounts. Put differently, with unchanged pension contributions, a transitional financing gap would emerge in the PAYG system, because individual contributions would be diverted from the PAYG system to the compulsory and voluntary individual retirement accounts. Until the transition from a single pillar PAYG system to the three pillar pension system is completed, the PAYG pension system would be faced with the obligation of having to meet pension obligations that exceed the reduced availability of payroll tax revenues. This transitional gap depends on a number of factors, including the size of the second and third pillar relative to the PAYG system and the speed at which the funded pension schemes are phased in.

164. This paper provides an overview of the Croatian pension system in terms of its current features and issues for reform in light of the multi-pillar reform proposals currently under consideration. Section B contains a description of the starting point, the existing PAYG pension system. Section C highlights the effect of aging on the existing PAYG pension system. Section D discusses some issues related to the envisaged downsizing and rationalization of the PAYG system. Section E provides a short overview of issues related to the establishment of compulsory and voluntary funded pension schemes which will be privately managed. Sections F and G shed some light on issues related to the transition from the single-pillar PAYG pension system to the multi-pillar pension system being considered in Croatia. Sections H and I briefly review the role of privatization proceeds in the reform process and the impact on government and private savings. Section J contains concluding remarks.

B. The Existing Pension System

165. The PAYG system is managed by the pension and disability fund (“the pension fund” hereafter), which comprises three subfunds: the workers’ fund (771,000 beneficiaries in 1996); the fund of self-employed (18,000 beneficiaries in 1996); and the farmers’ insurance funds (55,000 beneficiaries in 1996). These three subfunds provide three types of benefits: retirement pensions (450,000 beneficiaries in 1996), disability pensions (197,000 beneficiaries in 1996), and survivor pensions (196,000 beneficiaries in 1996).

166. The average monthly retirement pension paid out by the workers’ fund in September 1996 was HrK 887, the average disability pension was HrK 905, and the average survivor pension was HrK 747. The benefits paid out by the fund of the self-employed were about 25 percent below those of the workers’ fund. The benefits paid out by the farmers’s insurance fund were considerably lower, with the average old age pension for former farmers at HrK 153 in September 1996.

167. The pension fund is not allowed to run deficits except to the extent that it can finance them by drawing down bank deposits that were accumulated because of previous surpluses. In 1995, the pension fund had a small deficit equivalent to 0.2 percent of GDP: its total revenues amounted to 12.1 percent of GDP and total benefits to 12.3 percent of GDP. In 1996, the pension fund is estimated to have run a surplus of 0.1 percent of GDP, with revenues and expenditures reaching 12.9 and 12.8 percent of GDP, respectively.

168. The pension fund obtains the bulk of its revenues from employee and employer contributions. The percentage of its revenues raised from contributions, however, declined from 93 percent in 1995 to 88 percent in 1996, reflecting an increase in transfers from the central government from HrK 280 million to HrK 1,200 million. Transfers are budgeted to increase again in 1997, so that the share of contributions in total revenues is expected to decline to 83 percent in 1997. Contribution rates are set at 12.75 percent of wages for both workers and employers in the case of the workers’ fund, and at 25.5 percent of the self-employed’s insurance base in the case of the self-employed fund. Individual farmers contributing to the farmers’ fund pay 12.27 percent of the insurance base plus 0.76 percent of their annual cadastral income plus 0.76 percent of their average cadastral income. A small percentage of the revenues of the pension fund also accrues from its portfolio of shares in enterprises.

169. In terms of benefits, old age pensions are currently determined by the ten highest consecutive years of earnings. For the calculation of the pension base, earnings are adjusted upward by average wage growth in the years up to the one preceding the year in which the entitlement is established. The accrual factor is set in such a way that it aims to produce an 85 percent replacement rate for a full career worker. For men, the pension base is multiplied by a coefficient of 0.35 plus an additional 0.02 per each year of contribution up to a maximum of 0.85. For women, the initial coefficient is 0.40 plus an additional 0.03 per each year of contribution or other qualifying periods, also up to a maximum of 0.85. The standard retirement age for men is 60 and for women 55, but one can choose early retirement. For example, men can retire at 55 (with reduced benefits) after 35 years of work, while women can retire at 50 after 30 years of contributions. Under the early retirement option, the pension benefit is assessed in the same way as the regular old age pension, but a reduction factor of 1.33 percent is applied for each year of early retirement. Farmers are not entitled to early retirement.

170. There is a guaranteed minimum pension, provided that neither the retiree nor other members of their household have other sources of income sufficient to support them. In 1995, about 1/6 of the total retirees were entitled to a supplement to reach this minimum guaranteed threshold, which was increased to HrK 615 per month in early 1996. Certain groups of people—such as former members of the Yugoslav army, retired Parliamentarians, World War II veterans, invalids and families of victims of the last war, and refugees from other republics of the former Yugoslavia—also receive so-called beneficiary pensions. These beneficiaries receive pensions either without having contributed to the system or beyond the amount that would result from the application of the above-described arrangements. Starting in 1997, these beneficiary pensions are to be fully financed by transfers from the state. The increase in the minimum guaranteed pension in 1996, the decision to fully finance beneficiary pensions by state transfers, and the sharp increase in beneficiary pensions related to the last war explain the bulk of the increase in transfers from the central government budget in 1996 and 1997.

C. Aging Population

171. The PAYG pension system is particularly sensitive to demographics. Croatia has a relatively old population. In 1990, the percentage of the population over 60 years of age stood at 17.8 percent (see table 1). This percentage, which was only slightly below the average of 18.2 percent seen in the OECD countries, made Croatia one of the transition countries with the oldest populations. Moreover, as is the case with many other European countries, Croatia is undergoing a demographic transition. If current projections prove to be correct, Croatia’s population will experience significant aging in the next few decades as a result of declining fertility rates and the convergence of life-expectancy to the higher levels observed in most middle to upper income countries.

Table 1:

Percentage of the Population Over Sixty, 1990–2150

article image
Source: International Bank for Reconstruction and Development (1994).

Simple average.

172. Based on the most recent world population projections published by World Bank, the Croatian old age dependency ratio (measured as the ratio of the population of normal retirement age to the working-age population) is expected to increase from 36 percent in 1990 to 41 percent in the year 2000, 49 percent in the year 2010 and almost 68 percent in 2035 (see chart 1).45 Fortunately, the rate of increase in the dependency ratio is projected to slow down somewhat between 1995 and the beginning of the next century, thus providing some short-term breathing space to the pension system. However, the rate of increase in the dependency ratio is projected to accelerate in the first half of the next decade, and growth in this ratio will then remain at a substantially higher level until the year 2020. Under current pension arrangements, this would require a dramatic increase in the contribution rate.

CHART 1
CHART 1

Croatia: Old Age Dependency Ratio and Contribution Rate 1/ 2/

Citation: IMF Staff Country Reports 1997, 035; 10.5089/9781451817232.002.A001

Sources: International Bank for Reconstruction and Development (1994) and staff calculations1/ The old age dependency ratio is defined as the population of normal retirement ages (60 for men and 55 for women) as a percentage of the working-age population (men aged 15–59 plus women aged 15-54).2/ The pension contribution rate is calculated as described in section C of the paper.

173. If the average pension is B, the payroll tax rate t, the number of contributors L, the contributions base (the average wage) W, and the number of pensioners P, balancing the cash-flow of a PAYG plan requires the following: B*P = t*W*L; or t = (P/L)*(B/W). Thus, the required contribution varies directly with (P/L), a variable that can be proxied with the old age dependency ratio, and the average replacement rate (B/W). Application of this equation to Croatia—assuming unchanged average retirement ages, replacement rates, and the increases in (P/L) implied by the World Bank’s population projections—indicates that the required contribution rate will have to be about 5, 25 and 70 percent higher than now in the years 2000, 2010 and 2035, respectively, to preserve the balance of the PAYG pension system. Based on current law, it would thus have to be raised from 25.5 percent to 26.7 percent by 2000, to 32 percent by 2010 and to about 44 percent by 2035. These calculations clearly suggest that the contribution rates required to balance an unreformed pension system would start to become untenable from about 2005 onwards.

174. The necessity to pay for old age, disability and survivors pensions would eventually become a major burden for the economy. As the pace of privatization and the role of private sector firms increases, the negative effects on labor supply engendered by ever higher contribution rates would aggravate the tensions already emerging in the PAYG pension system’s financial situation. Therefore, the government’s decision to overhaul the current pension system is welcome. It also appears that the government would be getting an early start as near-term estimates by the World Bank suggest that the aging of the Croatian population would start to put the most serious pressure on an unreformed pension system around 2005.

D. Reform of the PAYG System

175. Under the reform proposal, a one-pillar PAYG system would be replaced by a three-pillar system. The first pillar would not vanish, but would be reduced in size and rationalized. Thus, workers would continue to contribute to the mandatory and public PAYG system and benefits would continue to be received under the PAYG system at the time of retirement, but they would be less generous. Accordingly, the accrual factor of the PAYG plan would need to be adjusted.

176. Measures that are being considered in the context of a rationalization of the PAYG system in Croatia—or elsewhere for that matter—have one thing in common: the choice is between cutting benefits (of the pensioners) and increasing the contributions (of workers). In line with the general approach of downsizing the role of the PAYG system, the reform options enumerated below are all based on the former. One advantage of concentrating on expenditure-cutting rather than revenue-raising measures is that it avoids increases in contributions which, by increasing the indirect cost of labor, may have a negative impact on employment. With Croatia already suffering from a relatively high rate of unemployment, this consideration is particularly relevant.

177. The main reform options are discussed below.

  • Increasing the retirement age. Under the current pension regime, the normal retirement age in Croatia of 60 and 55 years for men and women, respectively, which is low for a country with a population almost as old as that of the OECD countries. For comparison, average normal retirement ages in the OECD countries in 1990 were 64.4 and 62.9 years for men and women respectively, with several countries having retirement ages of 67 for both men and women47. Although life-expectancy in OECD countries is also significantly higher than in Croatia, these figures clearly suggest that there is substantial room for increasing the retirement age in Croatia. In this respect, recent simulations conducted by the IMF staff for industrial countries suggest that relatively moderate extensions in the retirement ages can have powerful positive effects on the financial position of pension systems.48 Recognizing this, the Croatian authorities have been considering an increase in the retirement age to 65 years for men and to 60 years women. A rapid increase has obvious advantages from a financial perspective, but political and social considerations advice a more gradual approach to raising retirement ages.

  • Linking benefits to lifetime earnings. In general, a lengthening of the earnings period applied for calculating the pension base (of currently a worker’s best ten consecutive years) should result in a reduction of the pension base.

  • Changing the benefits adjustment formula. If pensions are linked to wages, they will probably grow faster than if they were linked to prices. Linking pensions to prices protects the real value of pensions, but would probably exclude pensioners from participating in the growth of the economy and, as a consequence, the replacement rate would probably fall. To address the problem of a drop in the replacement rate, a review, say every five years, could take place to allow for a supplementary adjustment if this were felt to be warranted by the increase in wages and salaries. An alternative option (and one that has been considered by the Croatian authorities) would be to adopt a mixed system of indexation, such as the Swiss one, whereby pensions are linked to both wages and prices through a formula. With this intermediate option, pensions would grow over the medium-term less rapidly than wages but faster than prices, allowing pensioners to thereby benefit partially from productivity growth in the economy. Basically, the question of price versus wage indexation of pensions is centered on the issue of the extent to which pensioners should also enjoy the benefits of growth in labor productivity taking place after they have left the labor force.

  • Curtailing pension credits for years spent in child rearing, higher education and others.

  • Reducing the accruals rate. Unless the change is retroactive, it will effect pension benefits and the finances of the PAYG pension system only very gradually.

  • Increasing the actuarial decrement for early retirement The size of the actuarial decrement for early retirement is a key variable in a worker’s retirement decision. If the factor is too small, individuals will be encouraged to retire early because the discounted value of their expected pensions with early retirement is larger than with normal retirement. The 1991 reform in Germany introduced a 3.6 percent factor per year, which is being considered in Croatia, representing a marked increase compared to the 1.33 percent decrement under the current PAYG pension system.

178. In addition to these parametric changes that aim at rationalizing the benefits of the current PAYG system, the Government of Croatia has also focused on the collection side. In 1996, efforts to improve the collection of contributions owed to the pension fund included: strengthening the Institute for Payments system of collections such that social security contributions are in fact deducted before wage payments can be executed; legal characterization of social contributions as tax due and owed to the central government, with the same penalties as would apply to failure to pay income and profit taxes; enforcement of social security contributions by the dedicated unit within the Ministry of Finance; and late interest charges on delinquent payments that are significantly higher than commercial bank credit rates. In addition, means-testing provisions have been enforced.

E. The Two Funded Pillars

179. The PAYG system could be complemented by two funded pillars. The second pillar would be an obligatory funded and privately-managed pension scheme that would be based on individual accounts that are supervised and regulated by public authorities. Upon retirement, the payments out of the individual accounts would be in the form of annuities according to life tables. The third pillar would be a voluntary retirement savings scheme in the form of higher contributions to individual accounts, also supervised by public authorities.

180. One critical prerequisite for moving toward funded pension schemes is the existence of an adequately developed financial infrastructure, offering a basic set of financial instruments in a reasonably efficient way. Besides bank instruments, this should include government debt instruments allocated by a well-functioning, competitive market. The existence of independent pension fund management companies and life insurance companies associated with the pension system can improve the functioning of the financial market and, more generally, help facilitate financial market development. At the same time, private pension funds’ ability to earn reasonable rates of return is enhanced with well functioning financial markets, while ensuring that pension savings can be invested in financial instruments which are adequately regulated and appropriately safe.

F. The Transition

181. As soon as the two funded pillars have been established, the working population—for a number of years—would have to provide for the pensions of the already retired population under the PAYG pension system and, at the same time, save in their compulsory retirement accounts as well as, if desired, in the voluntary retirement accounts. They would have to pay, from current income, both for their own pension plan investments and the (unfunded) benefits promised to current retirees. In other words, with an unchanged rate of pension contributions, a fiscal gap would be created by diverting individual contributions from the PAYG system to compulsory and voluntary retirement accounts. This gap would change over time, but eventually disappear.49

182. Croatia is envisaging only a partial shift towards a funded system. The PAYG pension system would stay, but would be scaled down. There are various options in terms of the speed at which the role of the current PAYG system could be diminished and the funded pension plans be phased in. The key factors are (1) the size of the pensions to be provided under the new PAYG pension system and thus the share of pension contributions that will remain with this system and (2) at which age a switch to the funded pension scheme would take place.

183. Under a very gradual approach, only new labor market entrants would be required to make contributions both to the PAYG system and to the funded system. Under this scenario, contributions for new labor market entrants could be, for example, 15.5 percent of wages to the first pillar and 10 percent to the second pillar.50 Everybody else would continue to contribute the full 25.5 percent to the PAYG pension system. The total contribution rate would therefore remain at 25.5 percent. It would take about 45 years until all workers would be contributing under the multi-pillar system.51 Upon retirement, those who had contributed to the funded pension schemes would receive the basic pension under the new PAYG system and additional annuities based on the savings accumulated in their retirement account. Those who had not contributed to the funded scheme, but only to the PAYG system, would receive the basic pension under the new PAYG system plus transition benefits under the PAYG system. Over time, those who receive a transition benefit would diminish and disappear (after some 85 years).

184. For a more rapid shift, a plan could be adopted under which, for example, all workers below a cut-off age (say 40 years) would be included in the multi-pillar system. This plan would be along the lines of the proposal currently under consideration in Croatia. Under this scenario, contributions for workers below the age of 40 could be 10 percent of wages to the second pillar and 15.5 percent to the first pillar. For those currently above the age of 40, contributions would continue to be made in an amount of the full 25.5 percent of wages to the PAYG pension system. Within 20 years (assuming that retirement ages are kept at their present levels), all workers would be contributing under the multi-pillar system. Transition benefits under the PAYG system, however, would need to be paid for as long as under the previous scenario because of the rights that those under 40 of age at the time of the start of the reform would have accumulated from the previous PAYG system.

185. Under the most rapid approach, all workers would start contributing to the multi-pillar system from the beginning of the reforms. Again, transition benefits would need to be paid for the same length of time as under the two previous scenarios, i.e., until the last pensioner who entered the labor force before the reform process started, and has thus accumulated rights from the previous PAYG system, dies off.

F. Transition Costs

186. The calculations that follow are meant to be helpful as illustrations of the magnitude of the possible fiscal consequences of various transition scenarios.

187. In 1996, the Croatian pension fund had budgeted for payroll tax collections equivalent to 11 percent of GDP. Diverting 1 percentage point of today’s 25.5 percent contribution rate for every worker from the PAYG system to individual retirement accounts would result in a revenue decline for in PAYG system of about 0.4 percent of GDP. Assuming a cut-off age of 40, about half of the total wage bill would be affected by the switch52, and the budgetary costs of diverting 1 percentage point of today’s 25.5 percent contribution rate would be 0.2 percent of GDP. Consequently, 10 percentage points would probably “cost” about 2 percent of GDP.

188. During the transition, the fiscal gap of the PAYG system would change according to the speed of the switch to the three-pillar pension system. The dynamics depend on two factors: the speed at which the transition benefits paid out by the PAYG system disappear; and the extent to which pension contributions are diverted from the PAYG system to the funded pension plans. These two elements are in fact interrelated.

189. Under the very gradual approach, where part of the contributions for only new entrants to the labor market are sent to the funded pension schemes, the revenue shortfall in the PAYG system would be small in the beginning, but increase over time. At the same time, pension obligations under the PAYG system would remain unchanged for some time and start to decline only after the first worker retires under the multi-pillar system and begins to draw a pension from his individual accounts (after some 40 years).

190. On the other hand, if all workers were to switch immediately, the PAYG pension system would suffer the full revenue loss right from the beginning of the transition period, but pension obligations under the PAYG system would also start to decline early. Assuming contributions for all workers to the funded pension system were equal to 10 percentage points of the total of 25.5 percentage points, a deficit of 4 percent of GDP would immediately emerge in the PAYG pension system. The fiscal gap would begin to decline as retirees start to draw a pension from their funded pension schemes, accompanied by a fall in transition benefits paid out by the PAYG pension system. The gap reaches zero when the last transition benefit under the PAYG system is paid out (after some 85 years).

191. Recent discussions between the Croatian authorities and the World Bank, in the context of a possible Public Sector Adjustment Loan, have focused on a transfer to the second pillar funds, from the beginning of the reforms, of between 5 and 8 percentage points of the 25.5 percent contribution paid by workers aged 40 or below. If, for example, 8 percentage points of the contribution were transferred to the second pillar, the initial fiscal gap in the PAYG system would be equivalent to about 1¾ percent of GDP. This gap would increase during the years after pension reform is introduced because more and more workers would divert their contributions to their individual account; at the same time, PAYG benefits would remain unchanged. The effect on the fiscal deficit would peak in 20 years because, at that time, all workers would have moved to the multiple pillar system and the new pensioners would begin to draw a pension from their individual retirement accounts. Thus, the revenue shortfall in the PAYG system would stop growing; and benefits paid out by the PAYG system would start to decline.

G. Using Privatization Proceeds

192. Croatia is contemplating the use of privatization proceeds to help cover the financial burden resulting from the transition to a multi-pillar pension system. To finance the transition costs by earmarking these proceeds for the pension system could be done in a direct and an indirect way. In the second case, assets of the privatization fund would be sold and the proceeds used to close the initial financing gap. Alternatively, the assets owned by the privatization fund could be handed over directly to the obligatory pension plans.

H. Private and Government Savings

193. During the transition, when a financing gap would emerge in the public PAYG pension system, the private pension funds would, concurrently, run a surplus. Thus, if the Government decides to raise resources to cover the temporary gap in the PAYG system by issuing debt, there could be strong demand for this debt by the pension funds, since they need to invest in long-term instruments.52 At the same time, saving by the private sector would, ceteris paribus, rise in line with the pension funds’ surplus that results from introducing the second and third pillar. Thus, the deficit created in the government PAYG pension system and the consequent reduction in government saving would be offset by an increase in saving by the private sector.

194. The implications for national saving would, however, become more uncertain once the ceteris paribus assumption is relaxed. The final effect of pension reform on national saving would depend on its effect on the private sector’s perception of its future tax liabilities, on whether or not the private sector shows some degree of Ricardian behavior, on the extent to which the government takes fiscal measures to offset the temporary gap in the PAYG system, and on the particular form those measures take.53

I. Concluding Remarks

195. Population aging is a widespread phenomenon and will—as in many other countries—contribute to fiscal stresses on Croatia’s existing PAYG pension system. A substantial reform of Croatia’s pension system is unavoidable, although the most serious pressures from these demographic trends are unlikely to arise until about the middle of the next decade.

196. One route that a number of countries have taken is a reform through selected parametric adjustments of the PAYG system on either the revenue or the expenditure sides or both.

197. Croatia is considering a different approach. It plans to downsize, but not eliminate, the PAYG system, and complement it with fully funded pension schemes. Such an approach should significantly reduce the vulnerability of Croatia’s pension system to the aging phenomenon. However, there would also be significant temporary fiscal costs from shifting to a multi-pillar pension system, in the form of a transition gap in the public PAYG pension system that would need to be dealt with. In this regard, a large part of the transition costs could be covered by privatization proceeds if the privatization process were to gather sufficient momentum. Further, the temporary gap of the PAYG system would have as a corollary temporary surpluses of the funded pension schemes. A key question is the extent to which pension reform could generate an increase in private saving and how to channel it towards the financing of the deficit of the PAYG system.

198. Generally, pension measures take a long time to mature in the sense that their full financial implications may take more than a generation to become apparent. It is, therefore, welcome that Croatia is giving attention now to how to cope with the aging phenomenon; but care will need to be taken to ensure that the transition costs to a new pension system are manageable.

STatistical Appendix: Summary Of the Croatian Tax System

SUMMARY OF THE CROATIAN TAX SYSTEM AS OF DECEMBER 31, 1996

article image
article image
article image
article image
article image
article image

Table 1.

Croatia: Quarterly GDP at Constant 1990 Prices

article image
Source: Central Bureau of Statistics.
Table 2.

Croatia: Gross Domestic Product at Current Prices 1/

(In millions of kuna)

article image
Source: State Institute for Macroeconomic Analysis and Forecasting.

These estimates are produced independently by SIMAF. The Central Bureau of Statistics is in the process of preparing current price estimates for GDP.

Table 3.

Croatia: Gross Domestic Product at Constant 1994 Prices 1/

(In millions of kunas)

article image
Source: State Institute for Macroeconomic Analysis and Forecasting.

These estimates are produced independently by SIMAF and, therefore, are not consistent with official estimates from the Central Bureau of Statistics.

Table 4.

Croatia: Gross Domestic Product Deflators 1/

(1994=100)

article image
Source: State Institute for Macroeconomic Analysis and Forecasting.

These estimates are produced independently by SIMAF. The Central Bureau of Statistics is in the process of preparing current price estimates for GDP.

Table 5.

Croatia: Trends in Industrial Production 1/

article image
Source: Central Bureau of Statistics.

Seasonally unadjusted indices, 1990=100.

Table 6.

Croatia: Mining and Industry - Indices of Production, Stocks, Consumption, Employees and Productivity

article image
Source: Central Bureau of Statistics.

Consumption of electricity and fuels includes only fuels used for production and technology purposes. Fuels used as raw materials for further production are not included.

Productivity indices show the change in the period from January to the month for which the data is given in comparison with the same period in the previous year.

Table 7.

Croatia: Agricultural Production

(1988=100)

article image
Source: Central Bureau of Statistics.
Table 8.

Croatia: Tourism Data – Overnight Stays

article image
Source: Croatian Economic Trends.
Table 9.

Croatia: Number of Nights Spent According to Accomodation

(In thousands)

article image
Source: Central Bureau of Statistics.
Table 10.

Croatia: Nights Spent by Tourists According to Country of Origin

article image
Source : Statistical Yearbook, Central Bureau of Statistics.
Table 11.

Croatia: Composition of Employment 1/

article image
Source: Central Bureau of Statistics, Monthly Statistical Report.

The data cover the former socially – owned enterprises, privated and partially privatized enterprises, as well as the general government sector. Annual data are the average of March and September.

The last available data is for November 1996; provisional data is presented for that month. The number of employed persons for November 1996 is taken from a monthly survey which covers 70% of employed from each activity. The employed persons in the Ministry of Defense. Ministry of Interior and independent farmers are not included.

Table 12.

Croatia: Trends in Employment and Unemployment 1/

(End of period)

article image
Source: State Institute for Macroeconomic Analysis and Forecasting.
Table 13.

Croatia: Trends in Wage Bills and Non-wage Compensation

(In thousands of kunas, prices of December 1990)

article image
Source: Croatian Economic Trends.
Table 14.

Croatia: Trends in Average Monthly Net Wages and Salaries 1/

article image
Source: Croatian Economic Trends; and staff estimates.

Comprises the formerly socially-owned industrial sector (“economy”) and the general government sector (“non–economy”).

Series break starting in January 1995 when amounts previously included in non-wage compensation are shifted to wages.

Table 15.

Croatia: Average Gross Monthly Pay per Employee

(In Croatian kuna)

article image
Source: Central Bureau of Statistics.
Table 16.

Croatia: Indices of Nominal Net Wages and Salaries per Employee

(1993=100)

article image
Source: Bureau of Statistics.
Table 17.

Croatia: Indices of Real Net Wages and Salaries per Employee

(1993=100)

article image
Source : Central Bureau of Statistics.
Table 18

Croatia: Health Insurance of Workers

(Number of persons insured)

article image
Source: Central Bureau of Statistics, and Croatian Health Insurance Institute

As of 1991 health insurance for agricultural workers is also included.

Table 19.

Croatia: Child Care Supplements

article image
Source: Central Bureau for Statistics, and National Workers’ Retirement and Disability Fund
Table 20.

Croatia: Disability and Retirement Insurance 1/

article image
Source: Central Bureau of Statistics, and National Workers’ Retirement and Disability Fund

From 1991 onwards, coverage is slightly broader than previously.

Table 21.

Croatia: Price Developments

article image
Source: Central Bureau of Statistics.
Table 22.

Croatia: Retail Inflation Rates

(Average percentage change)

article image
Source: Central Bureau of Statistics.
Table 23.

Croatia: Indices of Prices

(1989=100)

article image
Source: Central Bureau of Statistics.
Table 24.

Croatia: Electro-Energy Balance Sheet

(In millions of Kwh)

article image
Source: Electricity Institute Zagreb.

Including hydro- and nuclear-generated electricity.

Table 25.

Croatia: Energy Balance Sheet

(In millions of tons coal equivalent)

article image
Source: Electricity Institute Zagreb.

Hydro=hydro-generated electricity/0.35.

Nuclear fuel=nuclear generated electricity/0.336.

Table 26.

Croatia: National Public Enterprises

(End of Year)

article image
Source: State Institute for Macroeconomic Analisis and Forecasting.
Table 27.

Croatia: Number of Enterprises in the Economic Sector and Number of Legal Entities Undergoing Bankruptcy

(End of Year)

article image
Source: State Institute for Macroeconomic Analysis and Forecasting

October.

Table 28.

Croatia: Number of Employees in Enterprises Undergoing Bankruptcy

(End of Year)

article image
Source: State Institute for Macroeconomic Analysis and Forecastingt

October

Table 29.

Croatia: Government Employment

article image
Source: Croatian Ministry of Finance

Estimate

Table 30

Croatia – Budgetary Central Government Revenue

article image
Source: ministry of Finance.

includes the Road Fund

GDP shares are at an annual rate.

Table 31.

Croatia: Budgetary Central Government Expenditure and Net Lending

article image
Source: Ministry of Finance

Includes the Road Fund

GDP shares are at an annual rate

Table 32.

Croatia: Budgetary Central Government Expenditure by Function

article image
Source: Ministry of Finance.

Includes the Road Fund.

GDP shares for January-September 1996 are at an annual rate.

Table 33.

Croatia: Consolidated Fiscal Accounts

article image
Sources: Ministry of Finance; and staff estimates.

Revenues and expenditures adjusted for inter-governmental transfers.

Includes since 1994 the Croatian Roads fund.

Composed of the Pension fund, Health fund, Employment fund, Child Benefits fund and, since 1994, the Water Management fund. Does not include the Privatization fund.

Table 34:

Croatia: Monetary Survey (End of period, millions of kuna)

article image
Source: National Bank of Croatia.
Table 35:

Croatia: Monetary Authorities Balance Sheet

(End of period, millions of kuna)

article image
Source: National Bank of Croatia.
Table 36:

Croatia: Deposit Money Banks’ Accounts

(End of period, millions of kuna)

article image
Source: National Bank of Croatia.
Table 37:

Croatia: Interest Rates of the National Bank of Croatia

(In Percentage, On Yearly Basis)

article image
Source: National Bank of Croatia.

From December 1992 through October 7, A291993 data present weighted averages of monthly interest rates on statutory reserves deposited with the NBC.

Since November 1993 table shows weighted averages of monthly interest rates from auctions of NBC bills.

Table 38:

Croatia: Deposit Money Bank Interest Rates – Domestic Currency Rates

article image
Source: National Bank of Croatia.
Table 39:

Croatia: Deposit Money Banks Interest Rates – Foreign Currency Rates

article image
Source: National Bank of Croatia.
Table 40:

Croatia: Balance of Payments, 1993–1996

(In millions of U.S. Dollars)

article image
Sources: Croatian authorities and staff estimates.

Does not include debt that was excluded from London Club agreement, and claims on international reserves of the former SFRY.

Includes deposits by non-residents, mainly international organizations.

Includes trade credits under 3 months, unreported foreign interest income, workers remittances and repatriation of savings in kind.

Excludes short-term trade related debt.

Table 41.

Croatia: Merchandise Exports and Imports, 1993–96

(In millions of U.S. dollars)

article image
Source: Central Bureau of Statistics
Table 42.

Croatia: Composition of Exports (SITC)

(In millions of U.S. dollars)

article image
Source: Central Bureau of Statistics.

Starting with 1992 data include trade with the countries of the former Yugoslavia.

Table 43.

Croatia: Composition of Imports (SITC)

(In millions of U.S. dollars)

article image
Source: Central Bureau of Statistics.

Starting with 1992 data include trade with the countries of the former Yugoslavia.

Table 44.

Croatia: Exports by Destination

(In millions of U.S. dollars)

article image
Source: Central Bureau of Statistics.

Starting with 1992 data include trade with countries of the former Yugoslavia.

Table 45.

Croatia: Imports by Origin

(In millions of U.S. dollars)

article image
Source: Central Bureau of Statistics

Starting with 1992 data include trade with countries of the former Yugoslavia.

Table 46.

Croatia: Exchange Rates and International Reserves

article image
Source: National Bank of Croatia.

Croatia introduced a new currency, the Croatian kuna, on May 30,1994, at the rate of one kuna per 1,000 Croatian dinars. Exchange rates prior to May 1994 are foreign currency units per 1,000 Croatian dinars.

Real effective exchange rates are calculated relative to seven currencies using retail or consumer prices. An increase in the rate denotes a real appreciation.

Table 47.

Croatia: External Debt Stock and Principal Repayments, 1993–2001 1/

(In millions of U.S. dollars)

article image
Source: National Bank of Croatia; and staff estimates.

Debt stock data for 1993–1995 have been adjusted to reflect the shares in unallocated principal debt to the Paris Club and the London Club creditors that have been recognized by Croatia.

Scheduled payments as of end-October 1996.

Excludes estimates of interest arrears, includes IMF.

Includes post-cutoff commercial credits (signed after December 2,1982) guaranteed by Government agencies.
2

Prepared by Christian Mulder.

3

Estimates for the first half of 1996 indicate somewhat higher numbers.

4

Some of the financing for higher imports of consumer durables and investment goods in 1995 and for sustaining them in 1996 appeared to be in errors and omissions and resulted from a drawdown of foreign savings held abroad. As declines in this form of capital inflows could be expected to be associated with an offsetting decline in imports, there may also be a self-correcting element to BOP adjustment.

5

This section builds on the “Report on the Balance of Payments Statistics Mission, February 15–29, 1996” by Ricardo Puig and Elizabeth Sumar (June 3, 1996) and on “Republic of Croatia Balance of Payments Compilation Manual,” draft version and non-official translation (April 5, 1996).

6

Customs data are quite comprehensive and appear to be well recorded-with the exception of small scale trade-and thus the main source of errors and omissions relating to the use of customs data must be sought in the financing of imports. One indication of the sizable underestimation of small scale exports is that the number of passenger cars entering Croatia far exceeds the number of cars exiting (CBS, transportation and communications survey).

7

A survey conducted for the first eight months of 1995 indicated an average payment delay of about 30 days for imports that accounted for over two thirds of total imports. A repeat survey for the first six months of 1996 indicated a further increase in the share of imports for which delayed payment was allowed to 72 percent in the first quarter of 1996 and to 75 percent in the second quarter of that year.

8

The amount of trade credit extended for less than three months, which is not registered with the National Bank of Croatia and thus not part of the reported short-term financing, is also estimated to have increased rapidly—broadly in line with the increase in trade credit extended for three months or more but less than one year.

9

Pliva is a large pharmaceutical concern; Zagrebacka Banka is a large commercial bank.

10

With retail margins in Croatia still relatively high, import taxes substantive, and distances short, there is substantial cross-border shopping. About 55,000 cars crossed the various borders per day in 1995, carrying a total of about 28 million passengers round-trip (6 times the population).

11

Foreign currency savings accounts were frozen in 1992.

12

The balance of payments would greatly benefit if payments made from non-resident accounts to residents accounts and cash withdrawals by nonresidents were more accurately measured and recorded under service inflows. Refining procedures to this effect should be relatively straightforward and be accorded a high priority.

13

The sizable and mostly regulated net foreign assets of the banking system provides an important cushion against not only short-term outflows, but also against diminished inflows from the repatriation of savings held abroad.

14

Croatians residing abroad for over 1 year are still usually registered as residents. A more exact registration for banking purposes would be helpful in order to assess appropriately the risks associated with the build-up of large short-term non-resident deposits. This is especially relevant because the community of foreign Croatians is large and relatively well to do.

15

Notwithstanding some concentration in the summer and holiday seasons and a downturn during “Operation Storm” in the late summer of 1995, the monthly pattern of foreign currency inflows has been is remarkably stable. This also suggests substantial continuity in these flows, which could be explained by a constant earnings stream, rather than one-off events.

16

The customs office collects data on imports for processing, as well as data on the corresponding reexports. While imports for processing were unchanged, reexports fell by 3.5 percentage points of overall exports in the first ten months of 1996 compared to 1995.

17

This figure is based on the export-weighted growth of imports of goods and services in countries that account for at least 95 percent of Croatia’s trade.

18

The restructuring of several industries (for example, finished textiles, textile fabrics, steel, and paper products) may also have had a negative effect on exports in 1996. This restructuring was partly in response to increased competition from the lifting of the economic embargo against the Republic of Yugoslavia, and the performance of these exports in 1997 is more uncertain.

19

One of the most difficult and largest items to project in the Croatian BOP is the foreign inflows in the form of cash and deposits by residents, which under current methodology are allocated to both tourism earnings (25 percent) and other sector non-loan inflows (75 percent). The impact on the overall BOP is, however, diminished by the large fraction of deposits which is redeposited abroad by banks. It is assumed in the baseline projection that the gross inflows remain broadly unchanged—thus the growth in tourism is only driven by the increase in nonresident inflows, in line with overnight stays as discussed above, and not by an increase in resident flows—but that net inflows decline as the percentage which banks redeposit abroad is conservatively projected to increase somewhat from the already high level estimated for 1996. It should be noted that an equivalent effect for the overall BOP can result if gross inflows decline and banks do not increase the percentage redeposited abroad.

20

The debt service ratio projection is rather sensitive to the maturity structure assumed for new loans, especially the percentage of short-term loans.

21

Prepared by Sonal Desai.

22

Monthly data is available for all series starting in June 1994 and ending in November 1996.

23

Such an approach is often motivated by the Baumol-Tobin model of the transactions demand for money.

24

This split between deutsche mark and dollars has been roughly stable for several years.

25

In very general settings the relative demand for two goods cannot always be characterized as a function of only their relative price. For example, to model the relative demand for two components of a consumer’s expenditure it is often necessary to assume some kind of separability in the utility function: either Hicksian or functional. For further discussion of this issue see Varian (1984), Microeconomic Analysis. New York: W.W. Norton and Co.

26

If the chosen model is misspecified the parameter estimates are likely to be inconsistent, as the coefficient on the interest differential will capture the covariance between that differential and the omitted variable.

27

Ratna Sahay and Carlos A. Végh (1996) “Dollarization in Transition Economies: Evidence and Policy Implications,” in P. Mizen and E.J. Pentecost, eds., The Macroeconomics of International Currencies: Theory, Policy and Evidence. Cheltenham, U.K.: Edward Elgar.

28

In this part of the empirical analysis, the interest rate on any monetary aggregate which foreign currency deposits is denominated in kuna and includes implicitly the constructed measure of expected kuna depreciation. Allowing expected depreciation to enter separately from the pure interest rate differential did not significantly improve the results.

29

Prepared by Thomas Dorsey. This paper is an extension of a part of a larger paper prepared for the Second Dubrovnik Conference on Economies in Transition: Velimir Šonje, Evan Kraft, and Thomas Dorsey, “Monetary and Exchange Rate Policy, Capital Inflows, and the Structure of the Banking System in Croatia,” June 1996.

30

Five additional majority state-owned banks entered the market between end-1991 and end-1995. Most were new subsidiaries of older majority state-owned banks.

31

Retail price inflation in Croatia fell from 38.7 percent in October 1993, to 1.4 percent in November 1993, and to 0.5 percent in December 1993. The first five months of 1994 were all characterized by falling prices.

32

This scaling is done to focus on the proportional effect of asset and liability allocation, not to offset heteroskedasticity (although it may serve that function as well). As these data are drawn from balance sheets, the total assets (or equivalently total liabilities) should not be interpreted as asset market values. This approach has been used even in cases in which the banks under consideration were failed banks. See in this regard David C. Wheelock and Paul W. Wilson, “Explaining Bank Failures: Deposit Insurance, Regulation, and Efficiency,” Review of Economics and Statistics, 1995.

33

See for example, Mary S. Schranz, “Takeovers Improve Firm Performance: Evidence from the Banking Industry,” Journal of Political Economy, 1993.

34

There are “new” state banks, but these are generally subsidiaries of, or successors to, other state banks that represent a continuation rather than a transformation of the ownership structure.

35

The measurement of privatization for banks existing at end-1994 is taken from an inventory of banks published in the Croatian financial journal Banka International, December 1995, pp. 38–43. Although the process of privatization began years earlier, these banks had not formally become majority private until 1994. Given that financial flow data was for 1995, end-1994 and end-1995 privatization data would be equally relevant. However, even though two additional banks had become majority privately-owned by end-1995, changing the definition of “privatized” to reflect end-1995 ownership status to add these two banks to the privatized category status did not substantially affect the empirical results.

36

The number of privatized banks based on ownership structure at the time of measurement was zero through 1993, 8 at the end of 1994, and 10 at the end of 1995.

37

Money assets such as reserve deposits with the NBC have been removed from total assets to approximate risk-weighted capital ratios (since these items are zero-weighted) for this table only. All other references to asset ratios use total book value of all assets.

38

Allen M. Berger, Anil K. Kashyip, and Joseph M. Scalise, “The Transformation of the US Banking Industry: What a Long Strange Trip It’s Been,” Brookings Paper on Economic Activity 1995:2, 1995.

39

The spread is calculated as the ratio of interest income to assets less the ratio of interest expenses to deposits. Three banks for which the interest rate spread is not meaningful (e.g., because of near zero deposits) are excluded from the calculation.

40

Announced interest rates are taken from the financial weekly Privredni Vjesnik.

41

Although somewhat subjective, the general criteria for the variables excluded from the restricted regressions shown in the table are the omission of variables with very low t-statistics and, in the case of the highly correlated Zagreb and New Private dummies, the omission of the less significant variable of the two when both are marginally significant.

42

The deposit-to-assets ratio is included in the restricted regression for non-interest expenses in spite of its low t-statistic to maintain parallel treatment to the restricted regression for administrative expenses.

43

Ideally, the potential interdependence between non-interest costs and interest rate spreads arising from accounting identities should be addressed through simultaneous equations methods. Broadly similar results were obtained with three-stage least squares estimates of a system similar to the second columns of tables 8 and 9. However, the usefulness of such estimates is severely limited because of the lack of truly exogenous instruments other than ownership and regional dummies (e.g., variables not included in balance sheet or profit-and-loss identities) and the attendant need for very strong assumptions behind the identifying restrictions.

45

Prepared by Christian Schiller with Heliodoro Temprano-Arroyo.

45

International Bank for Reconstruction and Development, 1994, World Population Projections 1994-95, (Baltimore: The Johns Hopkins University Press).

47

See the statistical appendix of International Bank for Reconstruction and Development, 1994, Averting the Old Age Crisis, (New York: Oxford University Press).

48

Sheetal K. Chand and Albert Jaeger, 1996, Aging Populations and Public Pension Schemes, Occasional Paper 147, (Washington: International Monetary Fund).

49

The fiscal deficit of the transition to a new pension arrangement in reality reflects the financial materialization of the outstanding implicit debt of the PAYG system. Holzmann has estimated, for a number of countries and for a reasonable range of parameter assumptions, that the present value of the total unfunded future liabilities of PAYG systems is 15 to 25 times current annual pension expenditures. With pensions in Croatia currently running at about 11 percent of GDP, applying these results to Croatia would imply an unfunded debt of the PAYG pennon system in Croatia on the order of 160 percent to 275 percent of GDP in present value terms. See R. Holzmann, 1993, “Reforming Old-age Pension Systems in Central and Eastern European Countries in Transition”, Journal of Economics, pp. 191–218.

50

Recall the total contributions are 12.75 percent of wages from workers, matched by employers to give a total of 25.5 percent.

51

This would be the time needed for the normal retirement of the youngest men already working at the time of the introduction of the pension reforms, assuming no change in retirement ages and assuming that workers enter the labor force from the age of 16 onwards.

52

Information on the distribution of wages by age was not available. A cut-off age of 40 is assumed to affect about half of the total wage bill: the World Bank population projections indicate that 56.4 percent of the working age population was below 40 in 1995. However, the average wage of workers below 40 is usually less than the wage of the more senior workers.

52

In Chile, for example, when the new fully funded pennon scheme was introduced in 1981, the pension funds were required to invest virtually all their portfolio in government paper. In the meantime, these restrictions have been eased somewhat.

53

A comprehensive list of the main references in the literature on the impact of pension systems on saving can be found in the bibliographic notes (p. 400) of Averting the Old Age Crisis, referenced in footnote 3.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Staff Country Reports
Croatia: Selected Issues and Statistical Appendix
Author:
International Monetary Fund
Volume/Issue:
Volume 1997: Issue 035
Publisher:
International Monetary Fund
ISBN:
9781451817232
ISSN:
1934-7685
Pages:
135
DOI:
https://doi.org/10.5089/9781451817232.002
CHART 1

Croatia: Old Age Dependency Ratio and Contribution Rate 1/ 2/