Chapter

V Vulnerabilities During Euro Adoption

Author(s):
International Monetary Fund
Published Date:
April 2005
Share
  • ShareShare
Show Summary Details

These underlying characteristics point to several potential vulnerabilities that will need to be considered in developing strategies—on both policies and timing—for euro adoption. An inherent risk for any country joining a currency union is the possibility of an inappropriate central parity. Because estimates of equilibrium exchange rates are never precise, it will be important to consider for each country the relative costs of over- and undervaluation and in setting the parity so as to err on the side of a range of equilibrium estimates that minimizes the costs of possible errors. More specifically to the CECs, capital account volatility will remain a serious challenge throughout the process of preparing to adopt the euro. In fact, the analysis in this section suggests that the very process of adopting the euro could expose countries to additional sources of volatility. Finally, current low levels of bank credit to the private sector relative to GDP in each of the CECs raise the specter of rapid credit growth that could feed overheating and asset price bubbles. Countries will, therefore, need plans for how to deal with possible surges in credit growth, both during ERM2 and after euro adoption. From the IMF’s perspective, assessment of each of these issues will be a critical part of its surveillance activities.

Getting the Parities Right

One of the most challenging steps in the euro adoption process will be getting the central parity and euro conversion rates right. Two broad considerations must enter the decisions. First, what is the right parity, what strains over a period even extending well-beyond ERM2 would result from getting it wrong, and how should this influence the choice of parity? Second, how does the choice of central parity enter into the strategy for a country’s participation in ERM2 and efforts to meet the Maastricht criteria? The first of these is addressed in this section and in Section VII (under “Choosing Parities”). In this section, it is assumed that the market rate remains close to the parity during ERM2 and that the central parity becomes the conversion rate. Section VII considers how differences between the market rate and the central parity might affect the economy and policy choices during ERM2.

The starting point for setting central parities is identifying the underlying equilibrium real exchange rate—the relative price between domestic and foreign output that is consistent with internal and external equilibrium. For a given constellation of relative wages and prices, estimates of a real equilibrium rate can be mapped into a specific nominal parity. Identifying equilibrium exchange rates, however, is difficult in the best of circumstances, and the CECs present particular challenges. The problem is to distinguish between sustained changes in equilibrium rates, which are likely to stem from productivity-driven real convergence, and less durable changes in actual real exchange rates. In the CECs the latter are likely to result from changes in capital flows owing to transitory influences such as tight monetary conditions, credit-rating effects, or surges in privatization receipts, which can weaken competitiveness relative to some notional equilibrium. Obstfeld and Rogoff (2000) show that the more closed is the economy and the lower is the consumption elasticity of substitution between tradables and nontradables, the higher is the real appreciation associated with inflows responding to such influences. Short-term price rigidities, in turn, can lead to an initial overshooting of the (CPI-based) real exchange rate.

Notwithstanding these difficulties, a number of methods to identify equilibrium exchange rates can be applied to transition countries. Appendix 5.1 describes these. Also, to illustrate the strengths and weaknesses of each method and show how market rates in the CECs now relate to estimates of equilibria, each method is applied to recent data for the CECs. A few broad conclusions stand out.

  • No signs of flagrant over- or undervaluation as of end-2003 are found. All currencies seem to have been within or close to a range that would constitute a reasonable equilibrium. In the broadest terms, the forint and Czech koruna were probably at the top end of an equilibrium range, while the indicators on the Slovak koruna suggest it was in the low end of its equilibrium range. The recent depreciation of the zloty has moved it substantially from what was probably the high end of its equilibrium range, although PPP (purchasing power parity) indicators still suggest room for improvement.

  • Measures of real exchange rates and competitiveness have varied quite widely—over a span of 25 percent in some countries—and suggest a rather large range of relative competitive positions among the CECs.

  • In general, the indicators provide less guidance on the size of any under- or overvaluation than on movements toward or away from equilibria. This means that estimates of equilibrium exchange rates will be rather imprecise bases for identifying optimal central parities. At best, they will suggest a range—probably at least as large as 10 percent—of rates that would be manageable.

Erring too far on the downside or upside even within this range would expose the CECs to vulnerabilities for a period extending beyond ERM2. An undervalued central parity would feed overheating and stoke inflation—both directly, via import prices, and indirectly, via demand pressures.30 An overvalued exchange rate would lead to a combination of large external current account deficits and slack in the economy stemming from low profitability in the tradables sector and its adverse effects on activity and employment. During ERM2, a widening current account deficit would raise warning flags in the markets, and could translate into pressure on the currency and, in the extreme, a speculative attack.

Eliminating misalignments of parities would involve potentially stressful responses of output and prices. Three key points about this process stand out.

  • To the extent that equilibrium exchange rates trend upward due to B-S effects, overvaluation would diminish over time, reducing the need for depressing price or output effects. However, an overvalued rate could discourage investment in the tradables sector, thus slowing tradables productivity growth and associated B-S effects. In effect, initial overvaluation could inhibit an equilibrium trend real appreciation.

  • The more the correction is achieved by prices rather than quantities, the more rapid the adjustment and the lower the costs in terms of higher unemployment and forgone output. While the CECs have more flexible labor markets than the European core, the evidence on product market flexibility is less clear. At any rate, downward wage and price flexibility clearly has its limits. In a low-inflation environment with any downward stickiness in prices and wages, costs of adjusting to an overvalued rate are likely to substantially exceed those of adjusting to an undervalued rate.

  • Any misalignment would likely require a fiscal policy response to avoid undue costs from price or output adjustments. Paradoxically, fiscal tightening may be needed to redress the consequences of either overvaluation (large current account deficits) or undervaluation (surges in capital inflows and overheating).

On balance, it is likely to be more difficult and costly to correct an overvalued exchange rate than an undervalued one, suggesting that vulnerabilities would be minimized by erring toward a parity at the weak end of an equilibrium range.

Capital Account Volatility

For the foreseeable future, large net capital inflows are likely to continue. Indeed, during ERM2 several influences may even make them larger than in recent years.

  • Underlying differences in capital-labor ratios will remain large. Particularly if EU accession diminishes differences in legal and physical infrastructure, rates of return on investment in the CECs should remain high or even rise.

  • Macroeconomic conditions are likely to become more favorable to inflows, especially after a country enters ERM2: prospects for euro adoption will foster higher growth; structural reforms to conform to the acquis and macroeconomic changes to meet the Maastricht criteria should improve confidence and reduce risk premia; and for some CECs, disinflation will require maintaining high domestic interest rates.

  • Convergence plays during the run-up to euro adoption will be an added attraction, particularly for the higher inflation CECs where interest rate convergence will cover the greatest ground.

Already the CECs have had net inflows relative to GDP exceeding those prior to euro adoption in most of the noncore countries (Figure 5.1). Much of the difference in net inflows reflects the fact that FDI relative to GDP in the CECs is typically a multiple of that during the pre-euro years of the noncore countries. Despite the liberalization of capital accounts in the CECs (although in Hungary and Slovenia only in the past few years), inflows of portfolio and other capital have generally been lower as a share of GDP than in the pre-EMU euro-area countries. But the influences during ERM2 and afterward noted above should work toward closing this gap.31

Figure 5.1.CECs and Selected Noncore Euro-Area Countries: Capital Flows

(In percent of GDP)
(In percent of GDP)

Sources: IMF, World Economic Outlook; and IMF staff estimates.

Just by virtue of their size, large net inflows have the potential for disruptive volatility. Some features of inflows to the CECs—the large share of FDI, generally small derivatives markets, and the role of economic fundamentals in attracting them—should continue to protect CECs against volatility. Indeed, the standard deviation of annual net inflows in the CECs over the past decade was 70 percent of the average level of the inflows, compared with well over 100 percent in several noncore countries (Greece, Italy, Portugal, and Spain) in the decade before euro adoption. Beyond the factors mentioned above, this difference in volatility undoubtedly also reflects the flexibility of CEC exchange rates, which has absorbed at least some of the incipient capital account volatility. ERM2 will likely entail some reduction in this buffer. Thus, although adopting the euro will significantly reduce vulnerabilities from capital account volatility, the process of getting to that point—particularly the stay in ERM2—may well increase it.

This section examines three sources of risk specific to ERM2 that could produce capital account volatility: changes in market perceptions about the timing of euro adoption (terminal date risks); asymmetric shocks; and policy inconsistencies. The first is obviously a pre-euro-adoption risk only; the other two will pose substantially greater risks before euro adoption than after. Moreover, transmission of each of these shocks would differ markedly before and after euro adoption. Prior to euro adoption each would work through sudden or large shifts in CEC risk premia or in expected future exchange rates, changing the attractiveness of assets denominated in CEC currencies. After euro adoption, without the scope for changes in forward rates or exchange rate risk premia, capital flows may be large but the consequences of volatility will be reduced. The focus in this section, therefore, will be on the pre-euro-adoption period, with particular attention to the risks during ERM2.

Terminal Date Risks

Markets have already formed, and will continuously revise, expectations about the terminal date of euro adoption. The expected terminal date, together with a credible central parity, will anchor the path for interest rates and the market exchange rate to converge to euro-area interest rates and the conversion rate, respectively. Changes in expectations about the terminal date would have potentially large effects on capital flows and/or the market exchange rate. Such changes could result from revisions to perceptions about a country’s ability to meet the Maastricht criteria by a certain date.

These shocks will be largely, but not completely, under policymakers’ control. Consistent policies with adequate cushions vis-à-vis the Maastricht criteria, complete transparency, and good communications with markets—for example, on intentions about the timing of euro adoption, monetary policy frameworks for ERM2, and medium-term fiscal plans—will go a long way toward avoiding surprises or errors in expectations about policies. But sources of uncertainty outside official control, such as contagion from perceptions about progress of other euro candidate countries toward the Maastricht criteria, could alter market views of unchanged and known policies. Even news about how the Maastricht criteria will be assessed could severely buffet markets.

Spreads on long bond interest rates in the CECs relative to the euro area, which will fall to a fraction of a percentage point by the conversion date, will be a bellwether for changes in market views on the pace or success of nominal convergence. In noncore euro-area countries, spreads against Bunds, which had varied around 300–400 basis points until three years before euro adoption, fell quickly and steadily once rapid progress in reducing inflation and fiscal deficits began (Figures 5.2 and 5.3). A precise mapping of interest rates and policy convergence is impossible, but rapid bond price convergence in the absence of policy convergence would raise questions about sustainability, while slower convergence would suggest market skepticism about the credibility of policies or the central parity. As of mid-2003 market views on the likely date of euro adoption by the CECs had shifted toward the end of the decade from 2007/08 previously. In June 2003, market participants expected Hungary, Poland, and the Slovak Republic to adopt the euro in 2008 or 2009, with an increasing bias toward 2009. The Czech Republic was widely expected to adopt the common currency only in 2010 (IMF, 2003a).

Figure 5.2.CECs and Noncore Euro-Area Countries: Interest Rate Differentials and Sovereign Ratings Prior to Euro Adoption1

(Percentage point differentials; end of period)

Sources: IMF, World Economic Outlook; and IMF staff estimates.

1t is the year before euro adoption for the noncore euro-area countries.

2 Difference between actual ESA-95 general government deficit as percent of GDP and the Maastricht deficit criterion.

Figure 5.3.CECs and Noncore Euro-Area Countries: Measures of Fiscal Convergence1

(In percentage points)

Sources: IMF, World Economic Outlook; and IMF staff estimates.

1t is the year before euro adoption for the noncore euro-area countries.

2 Difference between actual ESA-95 general government deficit as percent of GDP and the Maastricht deficit criterion.

3 Excluding second-pillar pension funds from the general government.

4 Including second-pillar pension funds to the general government.

As of October 2003, five-year bond yield differentials against Bunds ranged from 400 basis points in Hungary to close to zero in the Czech Republic. Excluding the latter, the CEC differentials are roughly similar to those that prevailed in the noncore euro-area countries some three to four years before euro adoption.32 At that point, the noncore countries had fiscal deficits of over 6 percent of GDP and inflation of about 5 percent. In contrast, fiscal and inflation convergence in the CECs varies widely, but the composition of progress toward the Maastricht criteria differs markedly from that in the noncore euro countries when interest rate differentials were at comparable levels (Table 5.1). Specifically, when average interest differentials in noncore countries equaled typical interest differentials in the CECs, fiscal deficits in the noncore euro-area countries were smaller vis-à-vis some of the CECs, but inflation rates and public debt ratios were generally higher. Therefore, while the possibility that interest rate convergence is getting ahead of deficit reduction efforts in the CECs exists, the risks are probably balanced by the better record on inflation and debt.

Table 5.1.CECs and Noncore Euro-Area Countries: Convergence and Interest Differentials
Czech RepublicHungaryPolandSlovak RepublicSlovenia
SE avg.1Actual2SE avg.1ActualSE avg.1Actual3SE avg.1ActualSE avg.1Actual
Date of comparable interest rate differential419992003 est.19942003 est.1995–962003 est.19962003 est.19952003 est.
Fiscal deficit (percent of GDP)5–0.9–6.6–7.2–5.9–4.6–5.6–3.3–3.6–5.8–1.8
Inflation (in percent)1.4–0.14.84.73.70.73.28.54.25.7
Debt-to-GDP ratio (in percent)83.931.389.359.185.149.184.242.886.026.7
Sources: Eurostat; country authorities; IMF, International Financial Statistics; and IMF staff estimates.

Southern European countries (SE): Greece, Italy, Portugal, and Spain. Indicators for Greece measured two years after the date shown for the SE average.

ESA-95 deficit and debt exclude government guarantees that have not been called.

Excluding second-pillar pension funds from general government.

Interest differential measured with the benchmark five-year domestic currency sovereign bond and the five-year Bund rate.

ESA-95.

Sources: Eurostat; country authorities; IMF, International Financial Statistics; and IMF staff estimates.

Southern European countries (SE): Greece, Italy, Portugal, and Spain. Indicators for Greece measured two years after the date shown for the SE average.

ESA-95 deficit and debt exclude government guarantees that have not been called.

Excluding second-pillar pension funds from general government.

Interest differential measured with the benchmark five-year domestic currency sovereign bond and the five-year Bund rate.

ESA-95.

Risks from Asymmetric Shocks

Asymmetric shocks are likely to be the biggest threat to macroeconomic stability outside the authorities’ control before and during ERM2. Prior to ERM2, countries will continue to have the tools with which they have responded to such shocks over the past several years. The Russia shock in 1998, which affected the CECs considerably more strongly than it did the euro area, is a good illustration of the scope for responding to asymmetric demand shocks. Most CECs, in the face of substantial potential effects on market confidence, raised interest rates, used scope within their flexible exchange rate frameworks to allow their currencies to depreciate, permitted the operation of fiscal stabilizers, and supported industrial restructuring that hastened the switch in production toward goods destined for Western markets. No country experienced a speculative attack. The experience of the Baltic countries during the Russia crisis, however, points to the viability of other strategies, again as long as they are clearly communicated to the markets and backed by consistent policies. Faced with the devaluation of the Russian ruble, the Baltic countries maintained their currency pegs unchanged and, in the end, survived the cyclical downturn, which proved to be shortlived, with their fixed exchange rates intact.

During ERM2, the CECs’ ability to withstand shocks will depend on the nature of the shocks, the consistency of the policy response, and the clarity of communications with the markets on how monetary policy decisions will be made. For most shocks, serious risks would arise primarily if policies are not sufficiently responsive and monetary frameworks do not provide sufficient latitude for and clarity about absorbing disruptions to financial conditions. In short, in the face of a significant shock with effects on capital account flows, credible mechanisms for allowing either exchange rates or interest rates to adjust are essential. And, in practice, most shocks can be addressed within any clearly defined monetary framework that leaves little doubt in the market about the credibility of the authorities’ response. ERM2 could weaken this credibility if it were perceived as limiting the scope for exchange rate changes beyond the ability of other macroeconomic policies to support.

The 1992–93 ERM crisis is the prime illustration of this risk. The shock of sharply tighter monetary policy in Germany—the anchor country of the ERM—to offset inflationary pressures created by a post-unification fiscal expansion brought to the fore in other ERM countries inconsistencies between the desire to maintain ERM parities to anchor inflation expectations and eroding competitiveness in the absence of inflation convergence. The free flow of capital brought these tensions to a head. After initial efforts to defend parities, it became apparent that intervention and interest rate hikes entailed domestic costs that would not be endured. Only those countries with broadly appropriate parities and the ability to mount a strong policy response to the shock survived with unchanged parities. The ERM crisis points to the rigorous conditions for a narrow band framework—willingness to subordinate economic policies entirely to the exchange rate objective—and the vulnerability of countries that fail to address overvaluation.

Box 5.1.Hungary—Experience with Increased Exchange Rate Volatility

Hungary’s 2003 experience with increased exchange rate volatility illustrates the risks of policy inconsistency and deteriorating economic fundamentals. Tensions in the monetary policy framework, which combines inflation targeting with exchange rate targeting, led in January 2003 to an attack on the strong edge of the exchange rate band. But by June 2003, weak economic outcomes and market doubts about the consistency of policies led to strong downward pressures on the currency. As a result, exchange rate volatility increased greatly.

  • In January, with the forint at the strong end of its ±15 percent band against the euro, the band came under a massive speculative attack. With considerable stimulus from an expanding fiscal deficit and rapid wage increases, speculators bet that the central bank (with needed legal consent from the government) would revalue the central rate to a stronger level or, absent a band, allow the exchange rate to appreciate, to meet the inflation target.

  • Instead, the Magyar Nemzeti Bank (MNB) attempted to fend off the attack by lowering the base rate (the interest rate on two-week deposits) by 200 basis points to 6½ percent; undertaking major intervention and accumulating large amounts of foreign exchange; widening the overnight interest rate corridor from ±100 basis points to ±300 basis points, thereby reducing the rate on overnight deposits from 7½ percent before the attack to 3½ percent after; and imposing quantity limits on the two-week deposit facility, which forced funds to be shifted to the overnight facility and caused money market rates to fall significantly below the base rate.

  • In the aftermath, the MNB sold foreign exchange to mop up liquidity and help unwind long forint positions; and, effective February 24, 2003, removed the limit on the two-week deposit facility and narrowed the interest rate corridor back to ±100 basis points (while keeping the base rate at 6½ percent).

  • In June 2003, the market exchange rate weakened rapidly after the government depreciated the central parity by 2.26 percent against the euro while leaving the currency’s ±15 percent band unchanged. With growing downward pressures on the currency, the MNB raised its main policy interest rate by a total of 300 basis points to 9½ percent. Following a period of increased volatility, the exchange rate stabilized in the range of 250–260 forint per euro until late November 2003.

  • In late November, renewed downward pressures on the currency mounted amid increasing market concerns about the expanding external current account deficit against a background of net FDI outflows and questions about the government’s ability to deliver its fiscal targets. With the currency weakening past 275 forint per euro and under considerable pressure, the central bank raised interest rates by 300 basis points to stand at 12½ percent. Considerable volatility ensued before the forint stabilized in a range of 260–265 per euro by the end of the year.

Hungary: Nominal Exchange Rate

Risks from Policy Inconsistency

ERM2 will place a high premium on the consistency of policies. The 1992–93 ERM crisis provides some indication of the rigor the market will bring to the evaluation of policy consistency. But strains on the Hungarian forint in 2003 suggest that the CECs could also face strong market scrutiny (Box 5.1). At issue in Hungary was the increasingly obvious conflict between an ambitious inflation target and an exchange rate pressing against the top end of a wide band. Tensions in monetary policy from this conflict resulted in an attack on the strong edge of the band in January 2003. But by June of that year, weak economic performance and market doubts about the consistency of policies led to strong downward pressures on the currency and a sharp increase in exchange rate volatility towards the end of the year.

Risks from policy inconsistency loom over all countries regardless of their monetary frameworks. The CECs could face a particular challenge, however, in avoiding policy inconsistency while steering their economies toward the Maastricht criteria. Specifically, the combination of an ambitious inflation target and an actual or perceived exchange rate band—asymmetric or not—may well be prone to problems of credibility in a way similar to the Hungarian experience. As in all instances with potentially conflicting objectives, supporting policies—in this case, fiscal and structural—will be critical to avoiding disruptive market pressures.

Post-Euro Risks: Capital Account Inflow

Once in a currency union, exchange rate risks will be eliminated. Yet the fundamental attractiveness of investment in the CECs—high rates of return on capital—is likely to remain. In fact, eliminating exchange risk, even if lower risk premia are broadly matched by lower interest rates, may enhance the appeal of the CECs as a destination for investment. But without exchange rate volatility in the picture, swings in market sentiment and, relatedly, capital inflows should be substantially less than before euro adoption.

An important remaining risk, however, is the potential for large inflows to build up liabilities—particularly in the banking system, but also through other markets—that could produce a relatively unconstrained increase in residents’ indebtedness. To the extent that such debt were not reflected in higher productive investment, the scope for property price bubbles and general overheating would exist.

The experiences of several existing euro-area members yield some interesting lessons. Among the noncore countries, the boom in demand brought about by the convergence-related decline in interest rates and strong capital inflows proved to be short-lived, leaving in its wake higher foreign liabilities of banks. The contribution of credit-financed spending to total demand was significant.

A substantial portion of the increase in investment was housing related and did not directly contribute to the repayment capacity of borrowers. Thus, the increase in indebtedness of the private sector needed to be unwound by restrained consumption and investment, leading to a sharp decline in GDP growth. This correction process is evident among countries that had a sharp buildup in private indebtedness. The risk therefore for post-euro-adoption CECs is that exchange rate volatility will be replaced by larger variations in demand and output.

Credit Booms: Risks and Responses

Rapid credit growth looms on the horizon for each CEC. Growth of permanent incomes and ample investment opportunities will inevitably combine with strong risk assessment techniques of newly privatized banks to produce rising growth of bank credit to the private sector (BCPS). Recent BCPS growth rates either in the aggregate or for specific classes of borrowers indicate that the process has already started in some CECs (Figure 5.4).

Figure 5.4.CECs: Bank Credit to the Private Sector1

(In percent of GDP)

Sources: Eurostat; and IMF staff calculations.

1 Credit data for the Czech and Slovak Republics are unadjusted for loan write-offs and changes in classification of financial institutions.

Some of the changes that will accompany euro adoption could hasten this process. The effects of euro adoption on growth and investment will of course be important demand-side impetuses to bank credit. In addition, however, the fall in risk premia will push down the cost of foreign borrowing by banks in the CECs, allowing them to fund a rapid increase in credit from abroad. Lower bank lending to the government attendant on fiscal retrenchment will free funds for lending to the private sector. After euro adoption, interest rates will be fully determined by euro-area conditions, while cyclical and structural influences might keep inflation high relative to the rest of the euro area. Very low or even negative real interest rates, especially vis-à-vis nontraded goods, could add to risks of a credit boom.

This section assesses the likelihood of rapid bank credit growth in the CECs and the contribution that euro adoption might make to it. Drawing on estimates from a model of credit growth in the euro zone, the section considers the risk that rapid credit growth will lead to overheating of the economy or to banking sector distress. Policy responses to forestall excessively rapid credit growth are considered.

Prospects for Credit Expansion in the CECs: What Does the Euro-Area Experience Suggest?

Rapid increases in bank credit began in several of the noncore euro-area countries about five years before euro adoption and continued after euro adoption. Growth rates were dramatic, pushing BCPS relative to GDP substantially above the euro-area average in Portugal and, to a lesser extent, Ireland (Figure 5.5). Some changes in policies and perceptions owing to prospective euro adoption undoubtedly contributed to these booms: governments reduced borrowing requirements, prospects for growth improved (reflected for example in rising real estate prices) and interest rates fell (Figures 5.6, 5.7, and 5.8). But other policy changes also played a role. For example, deregulation of domestic financial sectors (eliminating most interest rate caps, bank entry restrictions, and external capital controls) in the early 1990s set the stage for more dynamic banking sectors and an increase in foreign borrowing by banks (Figure 5.9).

Figure 5.5.Selected Noncore Euro-Area Countries: Bank Credit to the Private Sector

((n percent of GDP)

Sources: Eurostat; IMF, International Financial Statistics; and IMF staff calculations.

Figure 5.6.Selected Noncore Euro-Area Countries: Components of Bank Credit

(In percent of GDP)

Sources: IMF, International Financial Statistics; and IMF staff calculations.

Figure 5.7.Selected Noncore Euro-Area Countries: Residential Price Indices

(Index: 1992 = 100)

Source: Bank for International Settlements.

1 GDP weighted average of Austria, Belgium, Finland, France, Germany, Italy, and Netherlands. Data for Greece are not available. 2002 data for Austria not available, so the same growth as in 2001 has been assumed.

Figure 5.8.Selected Noncore Euro-Area Countries: Real Interest Rates1

(In percent, period average)

Sources: OECD; and IMF staff calculations.

1 Short-term rates refer to 3-month interbank rate; long-term rates refer to 10-year benchmark bond yield. Deflated by CPI inflation during the preceding 12 months.

Figure 5.9.Selected Noncore Euro-Area Countries: Banking Sector Assets and Liabilities

(In percent of GDP)

Sources: IMF, International Financial Statistics; and IMF staff calculations.

Real interest rate developments after euro adoption in some of the noncore countries continued to support private credit growth. In these countries, nominal bank lending rates remained at euro-area levels, while inflation tended to increase, reflecting the full influence of B-S induced real appreciations under exchange rate fixity and buoyant cyclical positions. Therefore, real interest rates stayed low (particularly vis-à-vis nontradables) or dropped further. Overheating pressures may even have set up a cycle of ongoing easing of monetary conditions, by lowering real interest rates and further fueling growth in credit and domestic demand (Walters, 1990; Rebelo, 1992).

Private sector credit ratios in Ireland and Portugal well above euro-area levels have raised concerns about the future health of their domestic banking sectors. Standard indicators of banking sector soundness, however, remain strong overall (Table 5.2). Non-performing loans relative to total loans have remained broadly stable since 1999 despite the strong credit growth. Average risk-based capital adequacy ratios remain above—and in some cases well above—the 8 percent minimum recommended by the Bank for International Settlements (BIS), although the ratio has fallen in Portugal. An added risk factor in both Ireland and Portugal is the exposure of banks to the real estate market: property-related loans account for more than 40 percent of total BCPS, compared with 28 percent in the euro area. Housing price increases in Portugal have been quite modest, since supply has tended to expand with demand. But in Ireland, real house prices have risen by over 130 percent since 1993.

Table 5.2.Selected Noncore Euro-Area Countries: Banking Sector Indicators, 1997–2003(In percent)
1997199819992000200120022003
Greece
Share of NPLs in total loans116.58.711.27.25.65.55.1
Risk-based capital adequacy ratio10.310.216.213.612.510.612.1
Ireland
Share of NPLs in total loans11.01.01.01.00.9
Risk-based capital adequacy ratio10.810.710.612.313.9
Portugal
Share of NPLs in total loans14.53.32.42.01.92.12.3
Risk-based capital adequacy ratio11.511.110.89.29.59.89.8
Spain
Share of NPLs in total loans11.01.00.91.01.0
Risk-based capital adequacy ratio12.612.412.912.512.5
Sources: IMF Country Reports; and national authorities.

NPLs, nonperforming loans.

Sources: IMF Country Reports; and national authorities.

NPLs, nonperforming loans.

The likelihood that the CECs will also experience credit booms is high, but how these will compare with the booms in some of the noncore euro-area countries is unclear. Some factors will lessen the risks. Nominal and real interest rates are already below those that prevailed in the noncore countries four to six years ahead of their adopting the euro. Heightened competition in domestic banking sectors (particularly where the foreign presence is large) has been one factor helping to push down bank lending rates close to levels in several noncore euro-area countries, with mortgage rates coming down the fastest (Figure 5.10). Interest subsidies on mortgages in Hungary and—until recently—in the Czech Republic as well as tax deductibility of mortgage costs in the Czech Republic have further lowered effective borrowing costs. Persistent FDI inflows could continue to substitute for some bank borrowing by the corporate sector and help maintain low interest rates.

Figure 5.10.Euro Area and CECs: Interest Rates on Bank Loans

(In percent)

Source: Eurostat.

A number of other factors, however, suggest that the credit expansion in the CECs could exceed that in the euro-area noncore countries. Credit-to-GDP ratios in the CECs are lower, so the potential for catch-up is greater (Table 5.3). B-S effects are expected to keep inflation somewhat above that in the euro-area noncore countries on average, implying that already low real interest rates could drop below levels in the euro-area noncore group. Lower per capita incomes combined with expectations of fast income convergence after EU accession could also fuel a more rapid catch-up. That widespread credit expansion has not yet taken place in the CECs, despite the low level of interest rates, may reflect a legacy of consumer, business, and bank caution with credit instruments. Overcoming this reticence could start a rapid adjustment in BCPS. Eliminating controls on housing purchases by foreigners (which have so far suppressed residential property prices in the CECs) over the next decade in accordance with accession agreements could also spur strong increases in house prices and mortgage borrowing. On the supply side, CEC banks have so far relied primarily on deposits as their main source of funding, but like the southern noncore group, could increase funding from foreign sources (Figure 5.11).

Table 5.3.Credit-to-GDP Ratios in the CECs and Noncore Euro-Area Countries, End-2002(In percent)
VECMCottarelli, Dell’Ariccia, and Vladkova-Hollar (2003)
ActualPredicted1Deviation2Predicted1Deviation2
Czech Republic30.978.5–47.669.3–38.4
Hungary34.973.3–38.470.5–35.6
Poland31.163.9–32.870.4–39.3
Slovak Republic32.574.3–41.859.9–27.4
Slovenia335.881.3–45.663.8–28.0
Average CECs33.074.3–41.266.8–33.7
Greece61.284.1–22.9
Ireland104.6105.0–0.4
Portugal133.185.947.2
Spain101.191.49.7
Euro area94.998.5–3.6
Sources: Eurostat; IMF, International Financial Statistics;Cottarelli, Dell’Ariccia, and Vladkova-Hollar (2003); and IMF staff calculations. VECM, vector error correction model.

Equilibrium value predicted based on estimates of the long-term cointegrating relationship, using actual data for 2002.

Deviation of the actual from the predicted level.

As of 2002, Slovenia had not issued any tolar-denominated long-term government debt. The predicted credit ratio in the VECM is based on the average real 10-year government bond yield in the other CECs.

Sources: Eurostat; IMF, International Financial Statistics;Cottarelli, Dell’Ariccia, and Vladkova-Hollar (2003); and IMF staff calculations. VECM, vector error correction model.

Equilibrium value predicted based on estimates of the long-term cointegrating relationship, using actual data for 2002.

Deviation of the actual from the predicted level.

As of 2002, Slovenia had not issued any tolar-denominated long-term government debt. The predicted credit ratio in the VECM is based on the average real 10-year government bond yield in the other CECs.

Figure 5.11.CECs: Net Foreign Liabilities, Credit to Private Sector, and Deposits

(In percent of GDP)

Source: IMF, International Financial Statistics.

Because historical data from the CECs would not be relevant for predicting future credit developments, any estimates of possible credit booms in the CECs must draw on other countries’ experience.33 As the closest available approximation to forward-looking behavior, quarterly aggregate data for the euro area during 1991–2002 were used to estimate a dynamic model of BCPS.34 The framework—a vector error correction model (VECM)—provides estimates of both the equilibrium value of BCPS and the dynamic adjustment path to this equilibrium (Box 5.2). The estimated parameters are used to make predictions for the individual CECs. Two considerations prompted this approach. First, several factors suggest that banking sectors in the CECs will increasingly come to resemble those in the euro area: euro-area banks’ large ownership share in the CECs’ banks; transposition of the banking sector elements of the EU’s acquis; increasing linkages with the euro-area interbank market; and deepening cooperation between EU (including CEC) financial supervisors. Second, while the experience of the noncore euro area may be more apt to this exercise, ex ante it is not obvious which of the non-core countries the CECs’ credit paths will most closely replicate.

For the euro area as a whole, the model suggests that the actual BCPS-to-GDP ratio at end-2002 was only marginally below the predicted long-term value. But for some individual euro-area countries, the deviations were large: actual BCPS exceeded fitted equilibrium ratios in Spain and Portugal by 10 and 47 percentage points, respectively, but for Ireland the deviation was minimal (see Table 5.3). Applying the same parameters to the CECs suggests that BCPS values were on average 41 percentage points below the equilibrium and varied within a narrow range.35 This result is broadly similar to forecasts for the CECs by Cottarelli, Dell’Ariccia, and Vladkova-Hollar (2003), based on a panel estimation from non-transition developing and industrial countries.

Results from the second component of the model—estimates of dynamic convergence paths of credit ratios to their equilibrium—can be used to simulate the adjustment of credit ratios in individual countries. For the noncore euro area, the gaps between actual and simulated BCPS ratios largely confirm the earlier heuristic analysis (Figure 5.12). For Greece, the credit ratio has increased more slowly than predicted by the model; for Portugal and Spain, the increase was much more rapid than forecast; and for Ireland, the increase was about as predicted.

Figure 5.12.Euro Area: Bank Credit to the Private Sector

(In percent of GDP)

Sources: IMF, International Financial Statistics; Eurostat; and IMF staff calculations.

1 Predicted values are obtained from the VECM.

Box 5.2.Estimation of Credit-to-GDP Ratio

The vector error correction model (VECM) of the demand for credit includes the three variables: creditratio (nominal bank loans to the nongovernment sector relative to GDP); rlti (the long-run real interest rate on government bonds); and lngdp (the log of per capita income measured in PPP terms). The second of these, rlti, is a proxy for the cost of credit. The 10-year government bond yield is deflated by annual inflation three years ahead, since no estimates of inflation expectations for this period are available. Deflating by contemporaneous inflation during a period of sustained disinflation would likely have biased downward measured real interest rates. The third variable, lngdp, is a proxy for the strength of economic activity and the overall financial condition of households and corporations. Per capita income, measured in PPP terms, can be thought of as proxying a borrower’s ability to service debt and therefore take on new loans. This accords with banks’ actual lending practices where, in a situation of imperfect information, banks rely on observable measures of repayment ability, such as current income. Without such market imperfections, the importance of this variable in predicting credit would be reduced. Euro-area data are taken from the following sources: bank loans to the nongovernment sector are from the ECB; 10-year government bond yields, HICP inflation, nominal GDP, and per capita GDP measured in PPPs are from Eurostat.

While some empirical studies of credit volume use indicators of financial liberalization and banking sector competition, these are not available as time series for the countries in our study, including several members of the euro area. These and other supply-side factors are likely to influence the dynamic adjustment in credit, rather than the equilibrium credit-to-GDP ratio.

Based on the maximum eigenvalue and trace tests, we find a single cointegrating relationship between these three variables that is significant at the 1 percent level:

0 = creditratio – 32.62 lngdp + 1.99 rlti.

This long-run relationship indicates that the credit ratio is positively related to per capita income and negatively related to the real rate of interest. The coefficient on the income term can be interpreted as a semi-elasticity: its estimated magnitude implies that a 10 percent increase in per capita income raises the credit-to-GDP ratio by about 3 percentage points in the long run. A rise in the real interest rate by 1 percentage point lowers the equilibrium credit ratio by nearly 2 percentage points. Detailed regression results are reported in Appendix 5.2.

Dynamic simulations for the CECs suggest a potentially rapid acceleration in real BCPS in the next few years.36 Growth rates are predicted to peak at 30–45 percent—substantially above the highest historical growth rates in the CECs (20 percent in Hungary), but below the peak growth rate seen in the noncore countries (66 percent in Ireland) and similar to recent rates observed in other transition countries (Bulgaria, Latvia, and Lithuania) (Figures 5.13, 5.14, and Table 5.4). By construction, the VECM model yields the sharpest increase in credit in the first year of the simulation (2003) when the credit-to-GDP ratio is furthest from its long-term equilibrium. Growth rates return to fairly low levels within a few years.

Figure 5.13.CECs: Real Bank Credit Growth to the Private Sector1

(In percent per year)

Sources: IMF, International Financial Statistics; Eurostat; and IMF staff calculations.

1 Nominal credit deflated by CPI; actual values for 1997–2002, predicted values for 2003–2020.

Figure 5.14.CECs: Baseline Simulations of Bank Credit to Private Sector1

(In percent of GDP)

Sources: Eurostat; and IMF staff calculations.

1 Actual values for 1996–2002, simulated values for 2003–2020.

Table 5.4.CECs and Noncore Euro-Area Countries: Credit Developments Relative to Risk Thresholds1
Historical maximum value (1990–2002)
Credit boom and crisis criteriaThresholdEuro areaGRCIRLPRTESP
Credit/GDP ratio (Gourinchas, Valdes, and Landerretche)2
Absolute deviation from trend34.8 percent of GDP2.56.6*6.5*11.0*7.4*
Relative deviation from trend424.9 percent3.012.17.59.710.6
Credit/GDP ratio (Kaminsky and Reinhart)59.3 percent of GDP4.78.323.6*17.2*7.5
Credit/GDP ratio (Borio and Lowe)64 percent of GDP2.56.6*6.5*11.0*7.4*
Real growth in credit710 percent per year7.638.5*66.4*25.9*15.0*
Memorandum item
Average annual real credit growth (in percent)3.94.818.811.94.1
Historical maximum value (1993–2002)
Credit boom and crisis criteriaCZHHUNPOLSLKSVN
Credit/GDP ratio (Gourinchas, Valdes, and Landerretche)2
Absolute deviation from trend36.0*5.5*3.06.4*1.6
Relative deviation from trend412.98.58.714.55.0
Credit/GDP ratio (Kaminsky and Reinhart)55.84.05.14.94.0
Credit/GDP ratio (Borio and Lowe)66.0*5.5*3.06.4*1.6
Real growth in credit73.120.6*16.7*9.027.0*
Memorandum item
Average annual real credit growth (in percent)–4.0–2.15.8–4.28.2
Predicted maximum value (2003–20)
Credit boom and crisis criteriaCZHHUNPOLSLKSVN
Credit/GDP ratio (Gourinchas, Valdes, and Landerretche)2
Absolute deviation from trend35.4*5.3*4.25.2*5.5*
Relative deviation from trend47.68.37.28.17.6
Credit/GDP ratio (Kaminsky and Reinhart)513.0*12.9*10.4*13.7*15.2*
Credit/GDP ratio (Borio and Lowe)65.4*5.3*4.2*5.2*5.5*
Real growth in credit733.5*41.2*29.6*44.0*39.2*
Memorandum item
Average annual real credit growth (in percent)4.78.58.13.44.4
Note: CZH, Czech Republic; ESP, Spain; GRC, Greece; HUN, Hungary; IRL, Ireland; POL, Poland; PRT, Portugal; SLK, Slovak Republic; SVN, Slovenia.

Values exceeding the corresponding thresholds are marked with *.

Gourinchas, Valdes, and Landerretche (1999). The trend is estimated using a Hodrick-Prescott (HP) filter; thresholds of lending booms defined to capture exactly 100 cases of lending booms in the sample of 91 countries over the period of 1960–96.

Absolute deviation is the difference between the actual credit-to-GDP ratio and the HP trend value, in percentage points of GDP.

Relative deviation is the ratio of the actual value of the credit-to-GDP ratio and the HP trend value, in percent.

Kaminsky and Reinhart (1996). The threshold for annual changes in the credit-to-GDP ratio that maximized the signal-to-noise ratio in predicting banking crisis in a group of industrial and market countries. Average for sample countries.

Borio and Lowe (2002). Absolute deviations from the HP trend of credit-to-GDP ratio; threshold is the best predictor of crisis at the one-year horizon, minimizing the noise-to-signal ratio.

Honohan (1997). Threshold for impending banking crisis.

Note: CZH, Czech Republic; ESP, Spain; GRC, Greece; HUN, Hungary; IRL, Ireland; POL, Poland; PRT, Portugal; SLK, Slovak Republic; SVN, Slovenia.

Values exceeding the corresponding thresholds are marked with *.

Gourinchas, Valdes, and Landerretche (1999). The trend is estimated using a Hodrick-Prescott (HP) filter; thresholds of lending booms defined to capture exactly 100 cases of lending booms in the sample of 91 countries over the period of 1960–96.

Absolute deviation is the difference between the actual credit-to-GDP ratio and the HP trend value, in percentage points of GDP.

Relative deviation is the ratio of the actual value of the credit-to-GDP ratio and the HP trend value, in percent.

Kaminsky and Reinhart (1996). The threshold for annual changes in the credit-to-GDP ratio that maximized the signal-to-noise ratio in predicting banking crisis in a group of industrial and market countries. Average for sample countries.

Borio and Lowe (2002). Absolute deviations from the HP trend of credit-to-GDP ratio; threshold is the best predictor of crisis at the one-year horizon, minimizing the noise-to-signal ratio.

Honohan (1997). Threshold for impending banking crisis.

The credit paths coming from these simulations should be seen as extreme scenarios. The VECM was estimated on the aggregate euro area, where deviations of actual credit ratios from equilibrium never exceeded 10 percentage points during the sample period (although for individual countries the deviation was as great as 50 percentage points on the upside). Using these estimates of the dynamic adjustment in the CECs, where the disequilibrium is considerably greater, is likely to bias upwards the projected pace of adjustment. This is especially true if institutional differences between the euro area and the CECs remain (for example, creditors’ rights continue to be less well-protected in the CECs, or banks continue to feel uncertain about the adequacy of credit assessments).

Further estimates and simulations of the model suggest that euro adoption per se would add little to the underlying impetus for rapid credit expansion in the CECs. To assess the implications of approaching euro adoption on credit, the VECM was reestimated including a variable measuring the number of quarters until euro adoption. The coefficient on this term, however, was small and statistically insignificant. The effects of euro adoption may be seen instead through larger B-S effects after euro adoption leading to higher nontraded-goods inflation and lower real interest rates than assumed in the baseline scenario. But even a simulation assuming that real interest rates converge to zero over 24 quarters (compared with 1½ percent over 12 quarters in the baseline) differs little from the baseline (Figure 5.15). Nevertheless, euro adoption could still impact bank credit growth through more general channels such as improvements in macroeconomic and structural policies.

Figure 5.15.CECs: Alternative Simulations of Bank Credit to Private Sector1

(In percent of GDP)

Sources: Eurostat; and IMF staff calculations.

1 Actual values for 1996–2002, simulated values for 2003–2020.

Financial Sector Risks and Remedies

A critical concern with rapid credit expansion is the risk of banking distress or even a banking crisis (that is, an episode when a significant segment of the banking sector is illiquid or insolvent). Before euro adoption, a credit boom could lead to twin banking and balance of payments crises (Kaminsky and Reinhart, 1996). With a substantial share of bank credit denominated in foreign currencies in some of the CECs, the risk that even perceptions of banking sector fragilities could generate pressures leading to a depreciation that aggravates weaknesses in banks’ balance sheets is significant. After euro adoption the risks involved in credit booms are somewhat reduced. Nevertheless, rapid credit growth could still generate overheating, asset price bubbles, and ultimately an impaired banking sector, and quelling it without an independent monetary policy would be challenging. Moreover, adjustment in the aftermath of overheating or asset price bubbles may well be difficult without an exchange rate instrument to effect needed changes in relative prices. Thus, whether a country with a large impending increase in bank credit to the private sector is helped or hurt by joining a currency union is far from clear.

The question of how great are the risks of rapid credit growth, nevertheless, remains. Several studies show that banking crises tend to be preceded by credit booms.37 This link presumably reflects a tendency for banks’ screening of borrowers to worsen when funding is easy and competition among banks is strong. In this sense, it may stem from the interaction of rapid credit growth with other factors—surges in capital inflows not directly related to the credit opportunities banks face (Hardy and Pazarbasioglu, 1998), asset price inflation (Borio and Lowe, 2002), financial deregulation (Eichengreen and Arteta, 2000), or overexposure to credit risk.

Of course not all periods of rapid credit expansion herald future banking distress. Gourinchas, Valdes, and Landerretche (2001) find that financial development typically occurs in bouts, characterized by short periods of intense financial deepening. Consequent rapid output growth is likely to scale down the impact on credit ratios. That a banking crisis is far from a certain consequence of a lending boom is evident in estimates of probabilities of banking crises. For example, Gourinchas, Valdes, and Landerretche (1999) find a relatively low probability (10–21 percent) of a banking crisis following a lending boom, while Tornell and Westermann (2002) estimate the probability at 6–9 percent, only slightly greater than during periods of tranquil credit growth (4–4½ percent).38

Although the forward-looking simulations of credit growth in the CECs suggest vulnerabilities to banking distress, some special circumstances of the CECs mitigate the risks. The simulated maximum predicted BCPS growth in each of the CECs generally exceeds thresholds of credit behavior found in other studies to signal the presence of lending booms and heightened probability of a banking crisis (see Table 5.4). These thresholds are measured in terms of: (1) deviation of credit-to-GDP ratios from their long-term trend; (2) change in credit-to-GDP ratios; and (3) real credit growth rates. However, two caveats should be borne in mind. First, the simulations, based on a VECM for the euro area, are likely to be upper bounds on future credit growth. Slower movements toward equilibrium credit ratios would reduce outlying observations in Table 5.4. Second, each of the four noncore euro-area countries examined had credit growth rates in excess of at least one of the thresholds at some point in the past decade without experiencing a banking crisis. In fact, each of the CECs has already had credit growth in excess of a threshold, and only the Czech Republic has experienced widespread bank insolvency. This suggests that close ties with the EU, generally sound macroeconomic frameworks, and reasonably strong bank supervision can reduce the risks in any given rate of credit growth.

In light of possible challenges from rapid credit growth to financial, exchange rate, and price stability before euro adoption (and to financial and price stability after euro adoption), countries will need to consider policy responses to credit booms. The first line of defense against disruptive effects from rapid credit growth must come from strong supervisory and prudential oversight. While Financial Sector Stability Assessments (FSSAs) conducted recently in the CECs by the IMF in connection with the FSAP note progress in this regard, they also emphasize that a number of weaknesses need to be addressed to improve the CECs’ ability to weather a period of potential credit exuberance. Recommendations include moving supervisory frameworks to a consolidated basis and relying more on a risk-based approach; strengthening cross-border supervision, increasing autonomy for supervisors, including strengthening their legal protection; and expanding enforcement powers. Recent empirical work also points to the importance of regulations and supervisory practices that require accurate information disclosure by banks, empower private sector monitoring of banks, and foster incentives for the private sector to exert corporate control (Barth, Caprio, and Levine, 2004). The FSSAs also pressed for measures related to these objectives: enhancing risk management capacities of banks; ingraining principles of good corporate governance in bank management; and establishing broad-based credit bureaus and better defining creditor rights.

Macroeconomic policies will need to play a key role in containing any excesses in BCPS growth. Until euro adoption, increases in interest rates, even in the absence of obvious inflation problems, may be necessary. However, large capital inflows, appreciation pressures, and rising demand for foreign currency borrowing, which often accompany rapid credit growth, would limit the scope for a monetary policy response. Indeed, using interest rate policy to stem rapid credit growth would not be without conflict even before ERM2 and euro adoption.

Participation in ERM2 and euro adoption will limit but not foreclose members’ financial policy options for avoiding a credit boom. Bank supervision will remain under the ambit of the national authorities, allowing countries to adapt if necessary their own prudential standards.39 EU accession requires the CECs to harmonize reserve requirements on banks’ liabilities with those of the European System of Central Banks (ESCB) but Greece was able to devise a system of higher quasi-mandatory reserves that was acceptable to the ECB and the EC. This indicates that reserve requirement harmonization is not a de facto constraint until euro adoption. Temporary controls on capital inflows, which are consistent with EU and OECD rules under specific conditions, could in principle allow some breathing room in the event of a foreign-financed bank lending boom.

Tightening prudential controls is another possible response to rapid credit growth. One possibility is to link minimum capital ratios, provisioning rates, and/or maximum loan-to-value ratios for collateralized lending to indicators of financial stress. Alternatively, these prudential norms may be adjusted according to the supervisory agency’s risk perceptions, which themselves could be based on stress indicators.40 To achieve a longer horizon for risk assessments, banks could be required to apply cycle-average loan default probabilities (rather than current default rates, which tend to be procyclical), or to accumulate higher provisions in periods of low loan losses (Borio, Furfine, and Lowe, 2001).41 These measures are not without pitfalls, however: adjustments in capital ratios are not effective when capital adequacy ratios (CARs) are well above legal limits (as in most CECs, Table 5.5), and dynamic provisioning (based on projected rather than actual losses) could decrease transparency.

Table 5.5.CECs: Banking Sector Indicators, 1997–2003(In percent)
Risk-Based Capital Adequacy Ratio1997199819992000200120022003
Czech Republic
Share of NPLs in total loans119.113.48.14.9
Risk-based capital adequacy ratio9.512.113.614.115.414.314.5
Hungary
Share of NPLs in total loans14.23.83.63.6
3.3Risk-based capital adequacy ratio14.213.713.913.011.6
Poland2
Share of NPLs in total loans110.210.513.315.017.921.120.9
Risk-based capital adequacy ratio12.511.713.212.915.113.813.8
Slovak Republic
Share of NPLs in total loans127.231.623.720.120.59.26.4
Risk-based capital adequacy ratio8.06.612.712.519.821.321.6
Slovenia
Share of NPLs in total loans13.84.23.93.7
Risk-based capital adequacy ratio13.511.911.911.5
Sources: IMF Country Reports; and national authorities.

NPLs, nonperforming loans.

Poland uses a more stringent criteria for classifying NPLs than international practice.

Sources: IMF Country Reports; and national authorities.

NPLs, nonperforming loans.

Poland uses a more stringent criteria for classifying NPLs than international practice.

Moving further in the direction of policies that control the symptoms—but not the underlying causes—of rapid credit growth are taxes on financial intermediation, quantitative limits on credit, and controls on capital inflows. Typically these types of policies have been used to contain current account deficits, control inflation, or prevent currency appreciation, rather than to protect the banking sector. For example, Greece introduced market-based credit controls to help achieve the Maastricht inflation criterion ahead of euro adoption. Fighting strong capital inflows, Croatia introduced taxes on foreign borrowing by banks in 1999 and bank-by-bank credit ceilings in 2003. Nevertheless, such measures could also serve prudential purposes.

Macroeconomic Booms

Potential credit booms are part of a broader picture of the likely pickup in growth of incomes, demand, and output as the relatively low-income CECs become more integrated with Western Europe. Ideally these effects would accrue smoothly, beginning even before EU accession and continuing after the CECs adopt the euro. In fact, the process is likely to be more choppy, with periods of rapid growth fueled by buoyant market expectations about the outlook interspersed with slowdowns related to cycles abroad, policy restraint, or changes in market sentiment. The process, therefore, is not without risks—of turning into episodes of overheating that could produce strains on resources, inflation, and asset price bubbles. An important challenge for policymakers will be to distinguish between excessive increases in demand and the sustainable gains from integration and ultimately currency union.

EU accession and the prospect of euro adoption will reinforce the divergence of domestic saving and investment through greater integration of goods and financial markets and convergence of nominal interest rates. Through these channels effective borrowing costs will be reduced, access to financial intermediaries will increase, and expected rates of return and relative prices will converge to EU levels. The accompanying increase in investment and consumption will be financed partly by inflows from wealthier countries, the counterpart of which will be widening current account deficits. Consistent with this prediction of the effect of EU and EMU-related integration, Blanchard and Giavazzi (2002) show that euro-area current account positions have become more positively correlated with per capita income during the 1990s: poorer countries have tended to run deficits (and accumulate foreign liabilities), and richer countries have tended to run surpluses (and accumulate foreign assets). Moreover, this pattern is not explained by differences in the behavior of the fiscal deficit.

Whether foreign inflows finance higher consumption or investment will depend on the structural characteristics of the country. Where a large share of households is still liquidity-constrained due to earlier financial repression, deepening financial markets will tend to boost consumption and lower saving. Where the low per capita income reflects capital scarcity rather than low TFP, integration should increase investment. These two effects could also occur simultaneously.

A simple optimizing model of consumer and firm behavior illustrates the importance of these basic structural characteristics for the paths of saving and investment as integration proceeds.42 The simulation assumes that nominal interest rate convergence generates a permanent 1 percentage point decline in the real interest rate. In the base case, 65 percent of individuals are assumed to be liquidity constrained. The drop in interest rates induces a pickup in investment and in consumption by individuals with access to capital markets. Under the base scenario, investment as a share of GDP gradually increases to a peak of 0.4 percent of GDP above baseline in the fourth year and stabilizes at this new long-run level (Figure 5.16). Consumption’s share in GDP increases in the first year to about 0.3 percent of GDP above baseline and then falls, eventually declining below baseline in the fourth year as individuals with access to credit increase saving to service their higher debt. The deviation in the current account deficit (which reflects these changes in saving and investment) peaks in the third year at 0.7 percent of GDP and then declines gradually. In this base case, higher investment is the main determinant of the expansion in the current account deficit.

Figure 5.16.Simulated Effect of a 1 Percentage Point Reduction in Real Interest Rates on Consumption, Investment, and the Current Account

(Deviation from baseline in percentage points of GDP)

Source: IMF staff simulations.

Yet a wide range of equilibrium investment and consumption paths are consistent with EU- and EMU-related cross-border financial integration and domestic financial deepening. The size of the consumption boom, for example, is sensitive to the share of liquidity-constrained consumers. Where a larger share of individuals already has access to credit, the initial increase in consumption is greater, and the subsequent decline is slower. On the other hand, if financial intermediation becomes more efficient (for example, because foreign bank presence increases) the share of liquidity-constrained consumers would fall gradually, causing a slower—but ultimately larger—increase in consumption compared with the scenario where half the population is liquidity constrained throughout. In this scenario, the widening of the current account deficit peaks several years after the interest rate drop, and rising consumption contributes slightly more than rising investment. The widening of the current account deficit is also sensitive to preference and technology parameters. For example, greater substitutability of capital for labor on the production side generates a larger increase in investment and a faster fall in consumption compared with the baseline.

A danger in this process is that the demand boom exceeds any equilibrium path and creates overheating pressures, unsustainable indebtedness, or asset price bubbles. For example, increases in aggregate demand in the converging country could boost inflation. When nominal interest rates are fixed by the ECB, real interest rates would decline further, triggering an additional, disequilibrating, demand response. Such a spiral may come to a sudden halt if creditors—concerned about mounting levels of indebtedness—are no longer willing to offer new financing, and debtors are forced to abruptly increase saving. In the context of the monetary union, this would precipitate a drop in growth, the size of which would depend on the downward flexibility of prices and wages. Similarly, if newly available mortgage financing were to lead to unsustainable asset or real estate price inflation, adjustment to appropriate levels would produce stresses on incomes, demand, and growth.

The size of the initial boom in domestic demand is an imperfect indicator of excessive convergence-related spending. That quite buoyant reactions of consumption and investment can be generated as equilibrium responses to interest rate convergence and elimination of liquidity constraints suggests that the best indicator of a sustainable boom is whether an orderly correction in consumption occurs. However, waiting for such a correction to occur does not offer much guidance for policymakers who, in the context of strong growth and a rising current account deficit, need to decide whether to introduce remedial measures. This suggests that, particularly in the early years of euro-area membership, the authorities will need to watch diverse indicators for signs of overheating, though none will provide unambiguous signals.

A clear role for fiscal policy exists when convergence-related spending is judged to be excessive. Under EMU, fiscal policy in the converging countries will be one of the most powerful macroeconomic tools for mitigating private sector demand booms. Judging the appropriate degree of fiscal restraint, however, will be as difficult as judging whether demand booms are excessive. Fiscal policy should not attempt to fully offset the increase in private spending coming from integration and financial deepening: eliminating an equilibrium widening in the current account deficit would close off one of the main benefits of integration—namely, the ability to reallocate consumption and investment through time. But a fiscal withdrawal would be highly appropriate for heading off a spiraling increase in inflation and private sector indebtedness. In this context several non-core euro-area members have relied on fiscal policy to varying degrees to mitigate convergence-related increases in private consumption and expansions in their current account deficits (Box 5.3).

Appendix 5.1. Methods of Exchange Rate Assessment

Choosing the Right Exchange Rate: Considerations and Methods

By comparison with mature market economies, equilibrium exchange rates in the CECs are likely to be subject to considerable change and to be difficult to pin down with much confidence. The particular challenge is to distinguish equilibrium changes in real exchange rates stemming from productivity-driven real convergence from real appreciations driven by capital inflows responding to tight monetary policy, convergence plays, and credit rating effects that can weaken competitiveness relative to some notional equilibrium. Obstfeld and Rogoff (2000) show that the more closed the economy, and the lower the consumption elasticity of substitution between tradables and nontradables, the higher is the real exchange rate appreciation associated with inflows responding to such influences. Short-term price rigidities, in turn, lead to an initial overshooting of the (CPI-based) real exchange rate.

This appendix briefly reviews possible methods of assessing equilibrium exchange rates. The main conclusion is that all have significant shortcomings. The choice of central parities must, therefore, take into account the high degree of uncertainty about the level of the exchange rate that will be conducive to exchange market stability and adequate competitiveness.

Real effective exchange rate measures

Measures of real effective exchange rates (REERs) are typically based on CPIs and unit labor costs (ULCs) relative to trading partners. A critical difference between the two stems from B-S effects, which should produce steady but sustainable increases in CPI-based REERs, but affect ULC-based REERs for manufacturing, where traded goods weigh heavily, relatively little. Measured ULC-based REERs could, however, display a steady appreciation if changes in the quality or composition of exports are not properly captured in measures of a “unit of production.”43 The strength of REER analysis is that it provides an assessment of changes in competitiveness over time with readily available and reasonably reliable data. Its weakness is that it provides insight on the level of competitiveness only if an equilibrium base period can be identified, and even then only if the subsequent path of the equilibrium is known. Efforts to get around this problem typically focus on identifying a “benchmark” period, characterized as a year when current account and export growth were deemed satisfactory, and judging divergences of the REER from the level in that period. Another caveat is that comparator countries are weighted by actual trade shares. In the CECs, where European countries dominate as trade partners, such measures put too little weight on actual and potential third-country competitors.

Box 5.3.Harnessing Macroeconomic and Credit Booms—Stylized Facts in the Noncore EMU Countries

Developments during the past decade in several non-core euro-area countries (Greece, Ireland, Portugal, and Spain) illustrate the links between economic integration and macroeconomic booms. Broadly speaking, macroeconomic balances since the mid-1990s followed similar patterns, although the intensity of credit, demand, and output booms varied rather widely (see figure).

Selected Noncore Euro-Area Countries: Macroeconomic Developments

(In percent of GDP unless otherwise indicated)

Sources: Eurostat; IMF, World Economic Outlook; and IMF staff calculations.

  • Investment ratios rose steadily for 5–6 years before leveling off.

  • Private saving also fell for 5–6 years before leveling off. In Portugal, Ireland, and Spain, household dissaving, at least partly fed by booms in bank credit to households, especially mortgages, was the primary cause (see Figure 5.4). In Greece, where the increase in BCPS has been smaller than in Portugal or Ireland, a decrease in retained corporate earnings was an important factor.

  • Inflation, after falling to pre-EMU lows, rose to a reasonably stable 3–5 percent by 2000. Property price inflation varied markedly—exceeding the euro-area average significantly in Ireland, but more modestly in Portugal and Spain (see Figure 5.7).

  • Increased public saving largely matched falling private saving, except in Ireland, where the increase in public saving was substantially larger.

  • Current account balances—a summary of domestic saving and investment trends—varied widely: reflecting the largest increase in investment among the four countries, Portugal’s current account position weakened to a peak deficit of 10 percent of GDP; Ireland, with the strongest fiscal position, kept its external position in balance.

  • In each country, GDP growth rose steadily for 5–6 years, through the first year or two of euro-area membership. Thereafter, growth in all countries except Greece fell rather sharply, possibly reflecting the global downturn as well as any endogenous slowing of the demand booms. Although following roughly the same pattern, Ireland’s growth rate was consistently a multiple of that in the other three countries.

No definitive conclusions can be drawn from this short period when many influences were at play. Nevertheless, the experiences suggest some possible lessons. First, while increasing economic and financial integration may have contributed to widening macroeconomic imbalances, no signs of a spiraling disequilibrium are apparent: this may reflect the timing of the global downturn and/or self-correcting mechanisms as indebtedness rose. Second, increased fiscal saving seems to have played a role in containing imbalances. Indeed, in Ireland, where public saving increased most, the current account imbalance was lowest. Third, the large increase in public saving does not appear to have dampened the gains from integration in Ireland, where growth significantly exceeded rates in the other countries.

As expected in transition countries experiencing B-S effects, Hungary and Poland saw significant CPI-based real appreciations during the second half of the 1990s, while their ULC-based REER fell or remained constant (Figure 5.17). In contrast, the Czech Republic and Slovak Republic saw large ULC-based appreciations during the same period. However, with a large margin of uncertainty about the extent of any disequilibrium in the early 1990s and about changes in the quality and composition of production, these developments do not make a clear statement about current competitiveness. Compared with recent “benchmark” years—when current account and export developments might be viewed as sustainable (Table 5.6)—Poland’s ULC-based REER has depreciated recently, whereas those of the Czech Republic and Slovenia have appreciated somewhat, and Hungary’s ULC-based REER has appreciated significantly. The Slovak ULC-based REER has remained roughly unchanged from the benchmark year.

Figure 5.17.CECs: Real Effective Exchange Rates1

Sources: National authorities; IMF, International Financial Statistics; and IMF staff estimates (INS for EMU countries).

1 End of period. For each country, the benchmark year (defined in Table 5.6) and the two preceding years are marked.

2 Trade-weighted (using all trading partners with a share higher than 1 percent).

3 From the Slovenian authorities.

4 Using weights based on export markets. The countries used are the United States, Canada, 12 euro-area countries, Norway, Sweden, and the other CECs.

Table 5.6.CEC Countries: Summary Exchange Rate Assessment, 2003(In percent, unless otherwise indicated)
Benchmark Year1ULC-REER Compared with Benchmark Year2Wage Costs/Value Added in Manufacturing3PPP Exchange Rate Ratio4Krajnyák and Zettelmeyer (1997) Dollar wage ratio6
ActualActual/norm5
Czech Republic20003.8615075–9781
Hungary200019.4685486–11367
Poland1998–23.9664789–12068
Slovak Republic20030.0504474–9758
Slovenia20015.4n.a6887–111n.a
Sources: National authorities; OECD; and IMF staff estimates.

The benchmark year is when the exchange rate was considered to be appropriately valued (by IMF country desks), considering factors including the size of the current account deficit, export growth, and GDP growth.

Unit-labor-cost-based real effective exchange rate.

Ratio of wage bill per employee to value added per person. Equals one minus the profit share.

Ratio between the market exchange rate and the PPP exchange rate (both relative to euro area).

Norm is the PPP exchange rate ratio consistent with a country’s GDP per capita. See Figure 5.18.

Ratio of actual dollar wage to “equilibrium” dollar wage.

Sources: National authorities; OECD; and IMF staff estimates.

The benchmark year is when the exchange rate was considered to be appropriately valued (by IMF country desks), considering factors including the size of the current account deficit, export growth, and GDP growth.

Unit-labor-cost-based real effective exchange rate.

Ratio of wage bill per employee to value added per person. Equals one minus the profit share.

Ratio between the market exchange rate and the PPP exchange rate (both relative to euro area).

Norm is the PPP exchange rate ratio consistent with a country’s GDP per capita. See Figure 5.18.

Ratio of actual dollar wage to “equilibrium” dollar wage.

Profit share indicators

The ratio of wage costs per employee to value added (in current prices) per person in manufacturing provides a measure of (one minus) the profit share in the tradables-intensive sector of the economy.44 This measure improves on ULC-based REERs by taking into account variations across countries in the price of tradable output (Lipschitz and McDonald, 1991). Nevertheless several caveats must be borne in mind. First, profit shares in manufacturing are not a good guide to differences in the rate of return on capital if there are significant differences in production technology. Comparisons of profit shares between countries at roughly similar stages of development should be more meaningful, although even here different product mixes can distort level comparisons. Analysis of changes in profit shares are meaningful because changes in production technology typically occur slowly. Second, the aggregate indicators could hide large differences in profit shares within manufacturing industry.45

This measure of profit share suggests that despite variations over a range of over 15 percent since the mid-1990s, no significant secular changes have occurred in the CECs (Figure 5.18). To the extent that comparisons of the levels of these measures are meaningful (that is, production technologies are similar), 2003 data suggest that the Slovak Republic had a substantially higher profit share than the other countries, while Poland and Hungary were closely aligned. Data for these calculations were not available for Slovenia.

Figure 5.18.CECs: Ratios of Wage Costs to Value Added1

Sources: OECD, STAN Database; OECD, Analytical Database; and IMF staff estimates.

1 Wage bill per employee in manufacturing, as ratio of value added per person employed. Equals one minus the profit share.

2 Using IMF staff estimates for total employment.

PPP exchange rates

In contrast to REERs, the ratio between the nominal exchange rate and the PPP exchange rate aims to assess levels of current exchange rates against their long-term equilibria. The PPP exchange rate compares the cost (in national currency) of a similar basket of goods (typically that of GDP) in two countries. For countries at similar levels of development, ratios of the market rate to the PPP rate above 1 indicate overvaluation and below 1 indicate undervaluation. For such countries, this is a particularly powerful tool because it measures over- or undervaluation directly, rather than indirectly via the presumption that any change in the real value of a currency is a movement toward or away from a static equilibrium.

Using the PPP exchange rate ratio to analyze exchange rates of countries at different stages of development is more complicated. Countries with lower GDP per capita have lower wages in the service (or nontradables) sector and therefore lower prices in this sector expressed in a common currency—the ratio of the market rate to the PPP rate should be below 1 and rising over time as their economies grow (De Broeck and Sløk, 2001).46Figure 5.19 bears out this relationship.

Figure 5.19.Euro Area and CECs: PPP Exchange Rate Ratio and GDP per Capita1

(1998 unless otherwise indicated)

Sources: OECD; IMF, World Economic Outlook; and IMF staff estimates.

1 The slopes of the illustrative regression lines resulted from panel estimation on data for 1980–2001 for the EMU countries and for 1993–2001 for the CEC countries. Regression 1 assumes no fixed-country effects; regression 2 assumes fixed-country effects. For both, the estimator was OLS on first differences.

2 Both relative to the euro area. A higher ratio indicates a more appreciated market rate. Euro area average = 100.

3 Euro-area average = 100.

In the CECs, as in the Southern euro-area countries in earlier decades, the PPP exchange rate ratio has tended to rise as convergence has proceeded (Figure 5.20). In most CECs, the ratio appears to have remained roughly in line with GDP per capita through 2003 (see Table 5.6). It has remained relatively low in the Czech Republic and Slovak Republic, while the significant depreciation of the zloty in 2002–03 has brought the Polish PPP exchange rate ratio to a level more in line with its GDP per capita. Slovenia’s PPP exchange rate ratio appears consistent with its relatively high GDP per capita.

Figure 5.20.Euro Area and CECs: Market Exchange Rate Compared with the PPP Exchange Rate1

Sources: IMF, World Economic Outlook; and IMF staff calculations.

1 The market exchange rate as a ratio of the PPP exchange rate (relative to euro area). A higher ratio indicates a more appreciated market rate.

Market shares

Euro-area market shares have generally developed consistent with this assessment (Figure 5.21). They have generally been increasing over time, as they should have, given the too low trade integration with Western Europe in the early 1990s. Since 2000, however, Hungary’s market share growth has decelerated, while the Slovak Republic’s share has outpaced that of the other CECs. That Slovenia’s market share has increased little since the early 1990s—although starting from a relatively elevated level—appears at odds with the development of its competitiveness indicators.

Figure 5.21.CECs: Market Share in Euro-Area Imports

(2000 = 100)

Source: IMF, Direction of Trade Statistics.

Macro-model-based and/or econometric estimates of equilibrium exchange rates

These approaches attempt to explain the time-varying path of the equilibrium REER as a function of economic fundamentals. Their application to transition countries remains challenging due to structural changes in the relationships and the absence of long data series.

The IMF’s macroeconomic balance approach and related calculations of the fundamental equilibrium exchange rate (FEER) are based on a structural model of the economy, focusing on the trade equations.47 This approach compares the underlying external current account with a norm or target. The underlying current account is derived by adjusting the actual current account for “transitory” elements, including the cyclical position and the impact of (all) past real exchange rate changes. The norm is derived from medium-term saving and investment balances or from current account positions needed to achieve a certain net foreign asset position. The gap between the underlying current account and the norm is then mapped into a gap between the actual and equilibrium exchange rate.48

The key caveat is that the approach—developed for mature market economies—assumes that excess current account deficits are caused by misalignment of relative prices. In transition economies, however, factors including consumption smoothing, volatile and bulky capital flows, and structural changes in saving behavior can produce temporary large current account deficits not necessarily due to a misaligned exchange rate.

In an application of this approach to Hungary, IMF (2003e) finds that, taking into account the “pipeline” effects of the recent real appreciation of the forint, the current account deficit exceeded a notional equilibrium in early 2003, implying an overvaluation. Šmídková, Barrell, and Holland (2002) employ the related concept of the fundamental real exchange rate (FRER) to the Czech Republic, Hungary, Poland, and Slovenia.49 Their results suggest that at the end of 2001 exchange rates for all countries except Slovenia were overvalued by 5–10 percent (compared with the upper band of a corridor derived using average historical volatility of exogenous variables).

Econometric investigations of the behavioral equilibrium exchange rate (BEER) estimate a reducedform relationship between the exchange rate and economic fundamentals as a benchmark for measuring overvaluation. Studies by Halpern and Wyplosz (1996) and Krajnyák and Zettelmeyer (1997), for example, estimate equilibrium dollar wages as a function of productivity and human capital proxies in a panel set-up. The main weakness of these studies is the absence of a theoretical model behind the estimated relationships. Moreover, while including a large number of countries improves the statistical properties of the estimation, the increasing heterogeneity of the group complicates the interpretation of the size of coefficients.

Applying data through 2003 to the model of Krajnyák and Zettelmeyer suggests that while both actual and equilibrium wages were rising in all CECs, the ratio of actual to equilibrium remained below 1, implying undervaluation (see Table 5.6). Although the levels of ratios should be interpreted with caution, given the complications due to heterogeneity, the ratios relative to other CECs remain meaningful. The calculations would suggest that, compared with equilibrium, actual dollar wages increased particularly rapidly in 2002–03 in the Czech Republic and Hungary. As a consequence, by 2003 the Czech wage ratio exceeded that in Poland, while Hungary’s was on par. In the Slovak Republic, though increasing, the ratio remained below that in other countries.50

Summary of Results

Taking the results of the price-, cost-, and profit-share-related perspectives on the “right exchange rate,” some conclusions can be drawn (see Table 5.6). While all the CECs show some rise in price and/or cost ratios, most are still not far from rates justified by productivity-based income convergence. As of 2003, Hungary appeared the least competitive among the CECs on the criteria of PPP exchange rate ratios and wage shares. This position had previously been taken by Poland. However, Poland’s competitiveness improved significantly through 2003, while competitiveness worsened in the Czech Republic and, in particular, in Hungary. The Slovak koruna appears undervalued through 2003, while the Slovenian tolar does not appear significantly misaligned in 2003. Euro-area market shares have generally developed consistent with this assessment. Looking ahead, while equilibrium exchange rates will strengthen in line with productivity developments and other economic fundamentals, actual rates will also be affected by financial capital flows, and distinguishing these two factors will remain key.

Appendix 5.2. Estimating Credit Behavior in the Euro Area

This appendix discusses the estimation of a model of bank credit to the private sector (BCPS) in the euro area using a vector error correction model (VECM). Data availability strongly influenced the specification of the model. Since separate data on credit to households and firms in the euro area are not available before 1997, the model therefore focuses on the aggregate behavior of these two sectors, assuming implicitly that they both exhibit similar responsiveness to the variables in the model.

Observed bank credit is the result of both demand and supply factors, introducing the possibility of an identification bias in the estimated coefficients. However, factors affecting the determinants of bank loan supply, including intermediation spreads and banks’ profitability, are likely distinct from those that determine credit demand, suggesting that credit demand can be correctly identified. Supply factors may nonetheless be an added constraint of credit behavior, especially in the short run. However, these data are not available on a harmonized basis for the entire euro area and therefore were not included.

The ratio of bank credit to GDP—rather than the level of nominal credit—was chosen as the “dependent” variable to allow the model estimated on the euro area to be applied out of sample to smaller countries. Moreover , the literature on banking crisis predictors commonly uses credit expressed as a share of GDP. Following Calza, Gartner, and Sousa (2003), the other variables in the system—the natural log of per capita GDP expressed in terms of PPP, and the real long-term interest rate—are included to measure economic activity and the cost of credit. That GDP enters separately in a model of the credit-to-GDP ratio would capture the extent to which financial deepening occurs as economies grow richer. Moreover, in the presence of information asymmetries, current income—rather than expected profitability or income arising from the use of the loan—would be expected to play a strong role in credit decisions because banks would need to condition their lending decisions on observable indicators of ability to repay (Kiyotaki and Moore, 1997). It is also possible that higher current income is associated with lower credit demand, as firms and households use unexpected positive shocks to profits and income to repay loans or to finance spending from internal sources (Friedman and Kuttner, 1993). A positive relationship between credit and income would therefore suggest that credit’s role in easing liquidity constraints dominates.

The real interest rate is included to capture the cost of borrowing. Ideally, the cost of alternative, non-bank sources of credit would be included, but even in more advanced financial systems, households have limited access to alternative borrowing sources. Given that about 75 percent of loans in the euro area mature in more than one year, and some 60 percent mature in more than five years, a measure of the cost of long-term credit is used. While bank lending rates would have been the most appropriate choice, data limitations dictated using the yield on 10-year government bonds. To obtain the real cost of credit in a period of generalized disinflation, an accurate measure of expected inflation over the duration of the lending period is ideally needed. Since no such indicator is available, these nominal interest rates were deflated by actual inflation 12 quarters ahead. The empirical results were not very sensitive to some lengthening or shortening of the inflation period used.

A number of structural factors have also been shown to influence BCPS.51 These include the extent of financial liberalization, banking sector concentration, credit-to-deposit ratios, restrictiveness of bank entry conditions, and the public debt ratio. Financial systems of current euro-area members were substantially liberalized in the early 1990s, opening the way for the subsequent takeoff in credit. However, liberalizations did not occur on the same timetable in all countries. It is therefore not possible to use one country’s financial liberalization index as representative of the entire euro area. Also, available data sets (including Abiad and Mody, 2003) do not include all (or even most) euro-area members, or any of the CECs. While recognizing the importance of this and other structural variables, the absence of data prevented their inclusion in the study.

Model Specification and Estimation Results

A vector of yt of n potentially endogenous I(1) variables can be modeled as an unrestricted vector autoregression involving up to k lags of yt:

This equation can be reformulated as a vector error correction model (VECM) of the form:

where Π = –(IA1 – … – Ak) and Γi = –(IA1 – … – Ai), (i = 1, …, k – 1). This way of specifying the system contains information on both the short-and long-run adjustment to changes in yt, through estimates of Γi and Π, respectively.

The model is specified in terms of the following variables: creditratio, lngdp, and rlti, where creditratio is bank loans to the private sector (denominated in euros) expressed in percent of GDP, lngdp is the natural log of per capita GDP measured in terms of purchasing power standard (denominated in thousands of euros), and rlti is the yield on the 10-year government bond deflated by actual inflation 12 quarters ahead, expressed in percent. Loans to the private sector include those allocated to households, nonfinancial corporations, nonprofit institutions, and non-monetary financial corporations. The last two groups account for only 5–7 percent of total credit to the private sector since 1997 (the period for which these data are available). The estimation is carried out on quarterly data for the period 1991 Q1 to 2002 Q4. Data on bank loans are taken from the ECB. All other data are from Eurostat. Through 2000 Q4, data cover the original 11 members of the euro area; thereafter, Greece is included. All aggregations of individual country data are performed by the data-issuing institutions.

Summary of Results

To determine the time-series properties of the individual data series, we perform unit root tests on the levels and first differences. The results are shown in Table 5.7.

Table 5.7.Unit Root Tests
Test Statistics1
VariableNull HypothesisAlternative HypothesisAugmented Dickey-Fuller2Phillips-Perron3
creditratioI(1)I(0)0.2570.499
I(2)I(1)–4.055**–4.146**
lngdpI(1)I(0)0.2820.168
I(2)I(1)–5.989**–6.075**
rltiI(1)I(0)–0.724–0.864
I(2)I(1)–5.700**–5.689**

Reject null at 1 percent level.

Constant included, no time trend.

Using largest statistically significant lag.

Using bandwidth chosen by Newey-West criterion.

Reject null at 1 percent level.

Constant included, no time trend.

Using largest statistically significant lag.

Using bandwidth chosen by Newey-West criterion.

The null hypothesis of a unit root cannot be rejected for the levels of the individual variables. However, when run on first differences, the null hypothesis is rejected at the 1 percent level. This suggests that all the variables are integrated of the first order.

Using the Johansen (1995) procedure, we first test whether a linear combination of the parsimonious set of I(1) variables is stationary; that is, whether they are cointegrated. The preferred specification (in terms of achieving the greatest significance of the test results) includes neither a constant nor trend in the cointegrating relations or the VAR. Table 5.8, which reports the results of the cointegration tests, indicates the presence of a single cointegrating relation based on the trace and maximum eigenvalue statistics, significant at the 1 percent level.

Table 5.8.Johansen Cointegration Tests
Number of Cointegrating Relations Under Null HypothesisTrace Statistic5 Percent Critical Value1 Percent Critical ValueMax-Eigen-Value Statistic5 Percent Critical Value1 Percent Critical Value
None30.27**24.3129.7523.37**17.8922.99
At most 16.9012.5316.315.6211.4415.69
At most 21.283.846.511.283.846.51

Indicates rejection of the null hypothesis at the 1 percent level.

Indicates rejection of the null hypothesis at the 1 percent level.

The normalized cointegrating relationship has the form (standard errors in parentheses):

The estimated long-run relationship indicates that the credit-to-GDP ratio depends positively on per capita income and negatively on the real rate of interest. The coefficient on the income term can be interpreted as a semi-elasticity. Its estimated magnitude implies that a 10 percent increase in per capita income raises the credit-to-GDP ratio by about 3 percentage points in the long run. A rise in the real interest rate by 1 percentage point is found to lower the equilibrium credit ratio by nearly 2 percentage points. Individual and joint likelihood-ratio tests of zero restrictions on the parameters in the cointegrating relationship were strongly rejected at the 1 percent level.

In the presence of cointegration, the short-run dynamics of the system can be represented by a restricted VAR consisting of two components: the lagged deviation of the individual variables from their equilibrium levels (the error correction term), and lagged first differences of the endogenous variables. A negative coefficient on the error correction term ensures convergence of the variables to their long-run levels, with the absolute size of the coefficient indicating the speed of convergence. As shown in Figure 5.22, the error correction term (calculated as deviations of the actual credit ratio from its equilibrium levels) was generally quite small (averaging –4.4 percentage points), with two-thirds of the observations in the interval [–6; –2] percentage points. The largest deviation occurred in 1993 Q1 (0 percent) and the smallest in 1999 Q1 (–9 percent).

Figure 5.22.Error Correction Term

(Credit/GDP; in percent)

The preferred specification of the VECM includes two lags and no exogenous terms. This specification generates residuals that are multivariate normal (based on the Jarque-Bera joint normality test) and serially uncorrelated (based on the Lagrange multiplier test for serial correlation of residuals). Using the results from the long-run relationship (the first stage of the Johansen procedure), the estimated VECM output is reported in Table 5.9.

Table 5.9.Vector Error Correction Estimates1
Dependent Variable
RegressorD(creditratio)D(lngdp)D(rlti)
EC term–0.089–0.002–0.026
(–1.86)(–3.10)(–0.57)
D(creditratio(–1))0.439–0.002–0.298
(2.67)(0.87)(–1.91)
D(creditratio(–2))–0.0160.004–0.089
(–0.10)(1.58)(–0.58)
D(lngdp(–1))3.623–0.195–12.752
(0.28)(–1.02)(–1.05)
D(lngdp(–2))–14.5020.04813.532
(–1.21)(0.27)(1.19)
D(rlti(–1))–0.1870.0060.165
(–1.11)(2.45)(1.03)
D(rlti(–2))0.2180.000–0.104
(1.24)(0.00)(–0.62)
R20.4000.2090.205
Adjusted R20.3050.0840.080
Standard error of equation0.4780.0070.453
F-statistic4.2161.6731.633

t-Statistics in parentheses.

t-Statistics in parentheses.

A variety of parameter stability tests suggest that the estimated parameters of the VECM vary little over the sample period. This conclusion is strongest for the credit ratio difference equation, where the Chow breakpoint test of parameter stability is not rejected at the 1 percent level at any point in the period. This is confirmed by the cumulative sum of recursive residuals (CUSUM) squares test at the 5 percent level. However, Chow forecast tests and the CUSUM test find some evidence of structural change in 1995 and 1997 (at the 5 percent level). The two other difference equations show somewhat more parameter variability.

We also test for weak exogeneity of the variables that, if present, implies that the short-run dynamics of the weakly exogenous variable are not affected by the error correction term and can be determined outside the VAR system. This would permit a reduction in the number of equations in the VECM. However, the weakly exogenous variable would still remain in the cointegrating relation. Individual and joint likelihood ratio tests for weak exogeneity yielded results as shown in Table 5.10.

Table 5.10.Weak Exogeneity Tests
Null Hypothesis1
D(creditratio)D(lngdp)D(rlti)D(lngdp), D(rlti)
Likelihood ratio statistic3.66*9.37**0.30710.14**
*(**) Rejection of the null hypothesis of weak exogeneity at the 5 percent (1 percent) significance level.

Coefficient on the error correction term in the corresponding VECM equation(s) is (are jointly) zero.

*(**) Rejection of the null hypothesis of weak exogeneity at the 5 percent (1 percent) significance level.

Coefficient on the error correction term in the corresponding VECM equation(s) is (are jointly) zero.

These indicate that changes in real long-term interest rates are not driven by convergence of credit, per capita income, and real interest rates to their long-run equilibria and therefore do not contribute to the restoration of equilibrium. However, the assumption of joint weak exogeneity of real interest rates and per capita income is rejected, so the full three-variable VECM is maintained.

If—in contrast to the assumption in this section—the central rate in ERM2 were substantially below the market rate, the associated interest rate differential vis-à-vis the euro area and/or market expectation of future revaluation could attract capital inflows and increase the vulnerability to foreign exchange market turmoil inconsistent with the exchange rate stability criterion. After euro adoption, this vulnerability disappears, although large inflows could still feed a credit boom.

Most portfolio investment inflows to the CECs come from “real money” institutional investors in Western Europe that have a long-term view on the region and take unhedged positions in medium-term local currency government bonds. Leveraged investors, such as hedge funds, have a much smaller presence that tends to increase in periods of high volatility.

While 10-year bonds are used in the assessment of the interest rate criterion, the 5-year bond market is a more telling indicator of convergence because of its higher liquidity.

Empirical work is scarce on the macroeconomic determinants of bank credit in the euro area or its members. Calza, Gartner, and Sousa (2003) estimate the level of bank credit to the private sector using a VECM for the euro area as a whole, but their specification of the cointegrating relationship precludes applying the parameter estimates to individual countries.

During 1991–2000, the euro area is defined as consisting of the 11 founding members of the currency union; Greece is included from 2001.

Values for individual countries are obtained from the estimated cointegrating relationship evaluated at actual real interest rates and PPP GDP per capita as of end-2002.

Out-of-sample values for interest rates and GDP growth are determined outside the fully endogenous VECM, recognizing that the three-variable system is insufficiently complex to capture the true determinants of all the variables. Thus, for each country, the real long-term interest rate is assumed to converge to 1½ percent (near the prevailing CEC average) in 12 quarters and to remain at that level. Per capita GDP in PPP terms is assumed to grow by 3 percent annually.

See Bell and Pain (2000) for a summary of the literature.

Gourinchas, Valdes, and Landerretche (1999, 2001) define a credit boom as a period when the ratio of credit to GDP differs from its long-run trend (proxied by a HP filter) by more than a specified threshold, and Tornell and Westermann define it as a cumulative increase in real credit over two years that exceeds a specified threshold.

Padoa-Schioppa (1999) notes that, within the euro area, the lender-of-last-resort function will remain at the national level. Only if the liquidity injection is on a scale sufficient to impact euro-area monetary conditions would the Eurosystem be actively involved.

During the 1990s, the Hong Kong Monetary Authority adjusted loan-to-value ratios for property lending over a range of 60–70 percent in a way that, ex post, was countercyclical.

The Bank of Spain introduced a system of “statistical provisions” in 2000 to complement existing specific and general provisioning. Statistical provisions are built up when problem loans are low and drawn down when problem loans are high, thereby smoothing the cyclicality of bank profits, limiting excessive lending in the upswing, and contributing to greater medium-term bank solvency (De Lis, Pages, and Saurina, 2000).

The model is a hybrid of two IMF models, GEM and MULTIMOD. Production is given by a constant elasticity of substitution technology, labor supply is fixed, and firms change their investment gradually due to adjustment costs but can borrow an unlimited amount at prevailing real interest rates. Two types of consumers exist: those able to borrow abroad at a rising cost and those who are liquidity constrained and do not have access to capital markets. The latter are only able to consume their labor income or accumulated savings.

Using disaggregated trade data (three-digit SITC), the EC (2002c) estimates that the quality gap between the accession countries and the EU, proxied by a price gap, narrowed on average by over 1 percent a year during 1995–2001 in most industries and by several percentage points in “technology driven industries.”

This measure is close to wage shares used by Lipschitz and McDonald (1991). Here, however, productivity is calculated per person employed (including self-employed), but wage costs are calculated per employee (excluding self-employed). This avoids a bias due to—sometimes tax-system-related—differences in the importance of self-employment across countries.

In the Czech Republic, a growing divergence has been observed between highly profitable foreign-owned firms and chronically loss-making domestic firms.

For a sample of developing countries, they estimate the log of the PPP exchange rate ratio as a linear function of the log of PPP GDP per capita in dollars and find that an increase in PPP per capita GDP of 1 percent increases the exchange rate ratio by 0.41 percent.

This approach is discussed by Isard and Faruqee (1998) and Isard and others (2001).

A related approach is the estimation of the natural real exchange rate (NATREX). Based on more rigorous modeling of stock-flow interaction in a macroeconomic growth model, it makes a distinction between medium-term equilibrium (with external and internal balance) and long-run equilibrium (with net foreign debt constant and the capital stock at a steady-state level). Its application to transition countries is particularly difficult.

They define the equilibrium external balance in terms of a target for external debt at the end of a simulation period, rather than in terms of a sustainable level of the current account.

Slovenia was not included in this study.

    Other Resources Citing This Publication