This paper assesses the extent to which data unreliability could alter the assessment of the macroeconomic challenges ahead. The contributions of the indirect tax authority (ITA) in remedying the flaws are highlighted, and the architectural agenda is discussed. Fiscal sustainability and the government’s bold initiatives to secure it by restructuring the domestic claims have been assessed and key implementation issues in realizing the government’s plans noted. A survey of selected tax policy issues in Bosnia and Herzegovina is also included in the paper.


This paper assesses the extent to which data unreliability could alter the assessment of the macroeconomic challenges ahead. The contributions of the indirect tax authority (ITA) in remedying the flaws are highlighted, and the architectural agenda is discussed. Fiscal sustainability and the government’s bold initiatives to secure it by restructuring the domestic claims have been assessed and key implementation issues in realizing the government’s plans noted. A survey of selected tax policy issues in Bosnia and Herzegovina is also included in the paper.

II. Data Foundations for Policy Design1

1. This note examines the extent to which the uncertainties about the data affect staff’s core medium term policy prescriptions. Those prescriptions rest on understandings that the external current account deficit is large relative to GDP, that wage arrears and unemployment is high, that corporate fixed investment is low and that corporate loss-making is widespread and heavy. The data on all these matters are highly suspect. But are they so suspect as to render the foundations of staff advice insecure?

A. Nominal GDP and Unrecorded Activity

2. Official estimates of GDP exclude the grey economy. Key gaps are listed in Text Table 1. While there is much uncertainty about the value added in these areas, it is likely significant.

Text Table 1.

Bosnia and Herzegovina: Non-observed Economy

(percent of official GDP)

article image
Source: IMF Statistical Advisor

3. Preliminary estimates by the IMF Statistical Advisor in Bosnia suggest that the non-observed economy is currently about 40 percent of official GDP. Much of the unrecorded activities are due to the understated value added of registered businesses. However, the share of the non-observed economy has apparently been declining in recent years with the reduction in the presence of the international community, and better tax administration has increased the coverage of official data—particularly in 2002 where this factor created a significant upward bias in official nominal GDP growth estimates.

4. Further evidence is provided by velocity of circulation of cash outside banks. On the official GDP numbers, velocity is unusually low in BiH relative to other currency board cases—Estonia, Lithuania, and Bulgaria—in the transition area (See Text Table 2). This suggests that GDP could, other things being equal, be significantly underrecorded. This impression is further suggested by low velocity in BiH relative to Croatia and Slovenia. Only Kosovo has a significantly lower velocity than BiH, but multiple special factors likely account for that.

Text Table 2.

Velocity of Circulation of Cash Outside Banks for Selected Transition countries (2002)

article image

5. However, the statistical authorities, particularly in the Republika Srpska (RS), doubt these estimates of the non-observed economy. Their main concern is that coverage of the official data has improved in recent years, particularly in the informal Arizona market in Brcko district, so that the 40 perecnt estimate seems implausibly high in their assessment. What is undisputed is that GDP remains well below pre-war levels.

B. The External Current Account Balance

6. Assessments of the policy implications of the current account deficit hinge on the estimates of nominal GDP. On staff estimates of nominal GDP and the current account balance, the current account deficit in 2002 was 18½ percent of GDP. Were nominal GDP 40 percent higher, that deficit would instead be 12 percent of GDP. However, there is additional considerable uncertainty about the nominal value of the current account balance, with staff estimates below others.

7. The merchandise trade deficit is large, but quite how large is uncertain (Text Table 3). Staff monitor merchandise trade tracking partner country data as well as customs data. Over the past year, these two sources of data have increasingly converged, mainly as a result of the country-wide introduction of the ASYCUDA customs information system. Unlike staff, the Central Bank of Bosnia and Herzegovina (CBBH) makes adjustments to the customs data for underreporting and undervaluation of both import and exports, making considerably larger adjustments for imports than for exports. This accounts for the higher merchandise trade deficit in its publications, but the assumption that unrecorded trade is as imbalanced as recorded trade is a strong one—while both would reflect shared macroeconomic influences, the sector level influences could be distinct, and exporters face strong incentives to underreport their activity in Bosnia if they wish to avoid the authorities in the destination (often neighboring) countries. Nevertheless, note that staff estimates take no account of underrecording—implicitly assuming that this trade is in balance, and this largely explains why staff estimates of the trade deficit are smaller than others report. However, all three sources of data suggest that the merchandise trade deficit worsened between 2000 and 2002.

Text Table 3.

Merchandise Trade Balance

article image

8. Weaknesses in data on services and private and official transfers compound uncertainties about the size of the current account deficit. For the time being the staff is using CBBH data on private remittances and our own estimates of official grants.

  • The magnitude and trends in private transfers are a key uncertainty in the current account data and the recent revisions by the CBBH in the relevant estimates are welcome.2 The staff still uses the earlier CBBH estimates of the inflow of private transfers. The per capita receipt of remittances in Bosnia in 2002 implied by these estimates is lower than in neighboring countries which are in a comparable state of economic development (Serbia and Montenegro, Macedonia, and Albania) (Text Table 4). However, there is considerable uncertainty about (i) the number of Bosnian emigrants and remaining refugees; (ii) the average level of transfers actually made by emigrants and refugees; and (iii) the level of pensions reveived from abroad. On the first two, the CBBH has recently refined its estimates using banking sector data on private transfers from abroad. On the last, recall that many Bosnian citizens worked, prior to 1992, in parts of Federal Republic of Yugoslavia (FRY) which are now neighboring countries, contributing to and now being paid by their pension systems. Accordingly, pension receipts from abroad per capita and relative to GDP may be unusually high by international standards. The uncertainty about the size of these remittances is potentially part of the explanation of the size of errors and omissions in the staff and the authorities’ estimates of the balance of payments.

  • Data on official grants is also of poor quality (Text Table 5). A UNDP-financed technical assistance project aimed at monitoring and coordinating aid to Bosnia has been running for several years now but credible data have yet to be produced. The staff is currently using “partner country data” on development assistance to Bosnia as reported by donor countries and IFIs to the OECD’s Development Assistance Committee to inform estimates of these flows. Note, however, that these data, which are available only through 2001, suggest that aid to Bosnia has not fallen quite as fast as is estimated by the staff. The discrepancy arises because these data do not decompose flows into current and capital grants and it is also not clear to what extent donors are reporting aid to the OECD which in fact has no impact on Bosnia’s BOP. For this reason, the staff has so far refrained from using this data, only drawing on it.

Text Table 4.

Remittances in 2002

article image
Text Table 5.

Official Development Assistance

article image

9. While the current account deficit is large and has increased markedly over the past few years, it appears to be mostly financed by non-debt creating flows. Thus the external debt to GDP ratio is on a strong declining trend (Text Table 6). Moreover, the weaknesses in the external sector data are reflected in large positive errors and omissions in 2001 and 2002. To the extent that this reflects unrecorded inflows in the current account—possibly underestimates of pension receipts from abroad—the actual current account deficit may be smaller than presently estimated by the staff. If, in addition, GDP is underestimated, the current account deficit could in fact be closer to 10 percent of GDP. Note, however, that this is well below the official—including the CBBH—estimates, partly because the staff do not make adjustments for unrecorded trade.

Text Table 6.

Current Account Balance

article image

C. Employment and Wages

10. Employment and wage data are also fraught with uncertainty reflecting the high incidence of wage arrears and the large number of insolvent companies. The available data suggest that about 20 percent of people registered as employees in the personnel departments of companies receive no wages. Moreover, about 15-20 percent of the average accrued wage per worker goes unpaid each month. Thus, average accrued and paid wages diverge significantly. The monthly wage data published by both entity statistical agencies are average paid wages. Data on the accrued wage bill are collected annually, in the context of the collection of national accounts data, but are not published and are not regularly decomposed into accrued wage rates and employment.

11. At staff’s request, the authorities attempted estimates of accrued wage rates, derived from the national accounts data on the accrued wage bill. These can be compared with reported data for paid wages, but comparisons are risky. The published data calculates average monthly paid wages as the ratio of the total monthly paid wage bill to the number of employees paid in that month. Hence those workers who were not paid in that month are ignored in the calculation. In contrast, estimates of the average accrued wage are calculated as the ratio of the total accrued wage bill to the total number of employees. Thus these two measures are not directly comparable—in fact the published average paid wages are typically higher than the (unpublished) estimates of average accrued wages, which is counterintuitive but correct because the set of workers in each is different and low accrued wage workers tend disproportionately to go unpaid. (Text Table 7). A more informative comparison of paid and accrued wages requires that paid and accrued wages be compared for an identical set of workers. However this detailed assessment has not been carried out.

Text Table 7.

Accrued and Paid Net Wages in 2002 (Km)

article image

12. Interpretation of wage data is also complicated by widespread underreporting, particularly in the private sector, as a means to avoid labor taxes. As a result a sectoral breakdown of wage data suggest that government wages are generally higher than private sector wages. However, a World Bank report, based on data from a Living Standards Measurement Survey (LSMS) of households conducted in 2001, found—upon controlling for other factors such as educational background—that there was a substantial wage premium from working in the private sector. The discrepancy between the wage data reported by households to the LSMS surveyors, and the data reported by enterprises to the statistical agencies is a measure of the degree of underreporting of wages, though it is not completely clear that the household respondents to the LSMS surveyors were reporting accrued wages or paid wages. The authorities have also noted that Pension Fund data provides evidence of underreporting—an implausibly high proportion of employees are reported to be earning only the minimum wage.

13. Similar problems arise in measuring employment. Because of pervasive wage arrears and corporate difficulties, four different types of employees may be distinguished:

  • those who actually work and get paid;

  • those who actually work but do not get paid;

  • those who do not work but get paid (though not necessarily the full amount of wages);

  • and those who do not work and do not get paid.

Many workers are known to report to work every day at defunct companies, but do no actual work—although they may increase the wage arrears liabilities of such firms. Also, many actually work but do not get paid because of financial problems at their companies. The published employment data do not distinguish between these four types of employees.

However, staff are provided with unpublished data on paid employees from a sample survey covering most large and medium size reporting units. But with data on employment in small enterprises being particularly poor, these estimates of paid employment are subject to wide margins of error. If total employment is defined as only those who worked and were paid, estimates of employment are overstated, while unemployment estimates are understated.

D. Unemployment

14. Unemployment is measured with significant uncertainty (Text Tables 8 and 9). There are two sources of data—official data on registered unemployed from the entity employment agencies, and data from the LSMS carried out in 2001. While they differ in terms of the reported unemployment rate, they both clearly indicate that there is a serious unemployment problem. This coincides with the anecdotal evidence provided by the authorities, trade unions, and business representatives.

Text Table 8.

Official Unemployment Rate

article image
Text Table 9.

BiH: LSMS Unemployment and Participation Rates in 2001, ILO Definition

article image

15. The official data on registered unemployment are particularly poor. A substantial number of workers in the grey economy are known to have registered as unemployed in order to enjoy health benefits. A World Bank labor market report calculates that “only slightly less than a quarter of registered unemployed qualified as unemployed under the standard ILO-OECD definition.” With these caveats in mind, the registered unemployment rate in both entities has hovered around 40 percent in the recent past. In the Federation there has been a slight upward trend since 2000. On the other hand, in the RS the unemployment rate rose strongly up to 2001, and has since improved somewhat.

16. The LSMS data are obtained from a survey of households, and may therefore better capture grey economy activities. However, the sample was assembled under difficult circumstances, and may be unreliable. The LSMS unemployment rate, calculated using standard ILO-OECD concepts, was reported to be 16 percent in 2001.

17. Concern with sample quality in the LSMS data centers on the observation that the participation rates reported are implausibly low, particularly for women—views shared by business representatives and entity statistics agencies. In addition to sampling problems, the low participation rates could reflect two other factors:

  • Refugees and displaced persons may have been wrongly classified as not seeking employment.

  • Responses from rural women may have been distorted—many would not have considered their small scale trading and farming activities as jobs, and would likely have reported themselves as not working, and (likely) not seeking work either.

18. Thus, a better measure of the unemployment situation may be obtained by including at least some discouraged workers as in the labor market, even though they are reportedly not actively seeking work. Reclassifying all discouraged persons as unemployed would cause the unemployment rate to increase to 31 percent, while the participation rate increases to a more plausible 61 percent, but with female participation rates still implausibly low (Text Table 10). On account of a judgement that including all of these workers as in the labor market may be excessive, notwithstanding the still low implied participation rates, the staff suggest that unemployment may be in the low-to mid-20s.

Text Table 10.

BiH: LSMS Unemployment and Participation Rates Incl. Discouraged Persons, 2001

article image

19. The unemployment problem is made more acute by the fact that over two-thirds of the unemployed—defined using standard ILO definitions—are long-term unemployed (Text Table 11). The LSMS data indicate that for BiH as a whole almost three-quarters of unemployed persons have been unemployed for more than one year. It is unclear what this pattern would look like were the unemployed defined to include at least some of the discouraged workers.

Text Table 11.

BiH: Unemployment Duration in 2001, from LSMS data

article image

20. Overall, it seems clear that there is a serious and persistent unemployment problem in Bosnia, regardless of which measure is used.

E. Fixed Investment

21. There are no official estimates of fixed investment at the BiH level. The estimates in staff documents are based on gross capital formation estimates for the Federation. These, however, suffer from poor coverage. To illustrate, we prepared alternative estimates of BiH investment using data on foreign direct investment, external loans and capital transfers for reconstruction, total construction output, and consolidated general government investment, as well as conservative assumptions on corporate investment (Text Table 12).

Text Table 12.

BiH: Gross Fixed Investment

(percent of GDP)

article image

22. These indicate an investment-GDP ratio typically somewhat above the estimates currently used by staff, though the difference has been declining—in 2002 the alternative estimate of investment was only slightly higher than the estimates in staff documents. And if GDP is significantly undermeasured, the ratios may be lower than shown unless the share of underrecorded investment in underrecorded GDP is higher than the share of recorded investment in recorded GDP.

F. Corporate Profitability

23. While data on this is poor quality, it all suggests widespread loss-making in the corporate sectors in both entities. Moreover, in numerous discussions with the authorities, trade unions, and business representatives they have concurred that the data are broadly representative of the actual situation in the corporate sector. Anecdotal evidence suggests that losses are particularly large amongst State-Owned Enterprises (SOEs).

  • World Bank survey data of 105 state owned and private companies indicate that over the period 1999-2001 the percentage of firms in the entire country reporting profits fell from 55 percent to 40 percent (Text Table 13). In the Federation, the percentage fell from 54 percent to 41 percent, while in the RS it fell from 57 percent to 39 percent.

  • These data also indicate that aggregate profit rates (percent of sales) have been significantly negative in both entities (Text Table 14). For BiH as a whole, the weighted aggregate profit rate has varied from a low of -15 percent to a high of -10 percent over the period 1999-2001.

  • Data from the accounting reports of over 7000 state owned and private companies in the RS (about 70 percent of all companies in that entity) indicate that aggregate operating profit rates have been negative from 2000 through the first half of 2003, albeit with an improvement in 2002, possibly reflecting the appreciation of the Serbian dinar vis-a vis the KM and other currencies (Text Table 15).

  • Recent monthly data on paid unit labor costs and manufacturing prices suggests a continued decline in profit rates in the Federation but an improvement in the RS since end-2001, the latter consistent with that reported by the 7,000 companies.

  • A joint DFID/World Bank assessment of financial statements of 40 and 50 strategic SOEs in the Federation and RS respectively, representing some 7 percent of the output in each entity, indicates that most made losses in 2002—aggregate profit rate was -7 percent for the Federation companies and -16 percent for the RS companies.

  • As noted above, wage and employment data indicate a significant buildup of wage arrears, suggesting profitability problems at current accrued wage rates. Data on paid and total employment indicate that every month about 20 percent of workers are not paid.

  • Notwithstanding problems with tax administration and possible underreporting of profits, revenue from profit taxes are very low in both entities, considering that profit tax rates are 30 percent and 10 percent in the Federation and RS, respectively (Text Table 16). This lends further support to the conclusion of low profitability.

Text Table 13.

Percentage of Firms Reporting Profits

article image
Text Table 14.

Aggregate Profit Rate

(In percent of sales)

article image
Text Table 15.

Republika Srpska: Operating Profits

(In percent of sales)

article image

Output Price/ULC

Citation: IMF Staff Country Reports 2004, 054; 10.5089/9781451804812.002.A002

Text Table 16.

Revenue from Profit Tax

(percent of GDP)

article image

24. Overall, therefore, there appears to be widespread persistent corporate loss making in both entities. This appears to be the key domestic counterpart to the large current account deficit, as lossmaking lowers domestic savings and constrains exports. The situation appears to be getting better in the RS, where wage growth has been more restrained in recent years, but there is so far no evidence of an improvement in the Federation.

G. Conclusion

25. While key policy recommendations appear to be robust to the data problems, the risk of surprises—pleasant or unpleasant—is high. Even on the most optimstic estimates of nominal GDP and the nominal current account balance, the external deficit is high, as are wage arrears and unemployment. Statistical and anecodotal evidence also both suggest that corporate investment and profitability are weak. Hence, the statistical basis for the staff’s core policy recommendations, while far from secure, suggests that the main assumptions underlying that advice hold.

26. This discussion nevertheless makes clear the urgent need to improve data quality. This will require political commitment, increased budgetary resources for statistics, and support for the Office of the High Representative (OHR) mandated integration of all the statistical services under a competent Director.


Contributed by Daniel Kanda and Geert Almekinders.


The revised estimates of private transfers, along with other revisions to the balance of payments, were posted on the website of the CBBH in January 2004.