The boom in household lending in transition countries: a Croatian case study and a cross-country analysis of determinants.

Kraft, Evan

INTRODUCTION

Since the mid-1990s, the transition countries of Central and South East Europe have made major strides in restructuring their banking systems. (1) A wave of foreign investment in banking has occurred, resulting in foreign bank ownership of a majority of bank assets in almost all of the countries. This has also been accompanied by strengthening of banking supervision, and improvements in the legal environment. A much more competitive environment has resulted, and the quality of the banking system seems vastly improved. (2)

Many of the countries have subsequently seen a sharp acceleration of bank lending. This has been associated with high levels of capital inflows in general, and, in many cases, with substantial current account deficits. (For an overview of the lending situation, see Backe and Zumer, 2005.) A striking feature of the acceleration in bank lending has been very rapid growth in lending to households, albeit from a low base.

In this paper, I will examine the determinants of this strong increase in household credit in the transition countries. Supply-side effects are clearly a major cause of the growth in household lending, as banking reform and privatisation have produced well-capitalised banking systems that are ready and able to provide a growing range of banking products. Additionally, interest-rate differentials have been large enough to allow for profit margins that are more than sufficient to compensate for any additional risk.

At the same time, demand-side conditions have also stimulated household lending. Pent-up demand for consumer durables and especially for housing has been a major factor. Robust economic growth has certainly played a role, and the expectation of higher future income, in part aided by the prospect of EU membership, has also boosted loan demand.

The macroeconomic implications of increased household lending have been readily visible. Theory tells us that household lending affects consumption if households are liquidity constrained. Cross-country research by Baccheta and Gerlach (1997) shows that liquidity constraints are common, and that credit aggregates have substantial impacts on aggregate consumption in advanced countries. Estimates by Corricelli et al. (2006) suggest that substantial numbers of transition country borrowers are also liquidity constrained, so that the provision of credit to households has significant effects on aggregate consumption and also on imports in transition countries as well.

Additionally, Herrmann and Jochem (2005) find that part of the current account deficit of Central and Eastern (CEE) and Southeastern (SEE) European countries can be accounted for by the level of GDP per capita. That is, when income grows, the estimated 'normal' current account deficit falls. Although the cross-country regressions cannot directly explain this, it would seem that demand for capital inflows fall as income increases, while previous investment raises productivity and exports, decreasing the merchandise trade deficit.

This suggests that to some extent, the current account deficits and also the rapid growth of credit seen in CEE and SEE countries can be attributed to a normal catching-up process. However, even if such a conclusion is correct in general, it does not imply that catching-up is without its dangers and pitfalls, nor that catching-up countries cannot experience harmful lending booms and subsequent crisis. And, of course, very rapid household credit growth may be problematic in this context.

At the same time, rapid credit growth has been identified as a key factor in banking and currency crises around the world. This raises the question of whether transition countries can have 'too much of a good thing' (Duenwald et al. (2005); see also Caprio and Klingebiel, 1996; Eichengreen and Rose, 1998; Eichengreen and Arteta, 2000; Borio and Lowe, 2002; Cottarelli et al., 2003). Much dispute rages about whether credit growth thresholds really exist, and whether they are relevant everywhere. For example, Gourinchas et al. (2001) suggest that problematic lending booms are mainly a Latin American phenomenon.

Using a case study of Croatia, I find that the expected negative effects of rapid lending growth on loan quality have not (yet?) materialised in the household sector. Furthermore, the rapid growth of housing credit, probably the best-collateralised and least default-prone subcategory of household loans, suggests that the relatively benign prudential status of household lending may continue in the future.

After the Croatian case study, I turn to a cross-country analysis. Using a sample of 90 countries, I study the main determinants of the ratio of household lending to GDP. I include a 'transition effect', which encapsulates the legacy of non-market banking systems offering little household credit. The transition effect, however, seems to be fading away over time as CEE and SEE banking systems converge towards EU standards.

Importantly, the rate and level of convergence varies from country to country. To analyse the diverse experiences of transition countries, I use the European Bank for Reconstruction and Development (EBRD) transition indicators, which provide quantitative indicators of reform progress in several particular areas of transition. The results of a second-step regression analysis of the residuals from the cross-country growth equations for transition countries suggest that stronger banking sector reform increases the level of household credit to GDP, evidently by increasing the supply of credit and in particular household credit. At the same time, stronger real sector reform decreases the supply of household credit by increasing enterprise sector credit demand and creditworthiness. This suggests that there may be a danger of unbalanced reforms, in which banking sector reform proceeds more rapidly than real sector reform, leading to strong household lending growth and weak enterprise sector lending growth, contributing to rapid consumption growth, high current account deficits and foreign debt problems.

The paper is structured as follows: the next section is a case study of Croatia, discussing the evolution of credit quality, and examining possible explanations for continued high repayment rates. It is followed by a brief overview of the household lending phenomenon in the transition countries as a group. The next part of the paper provides a cross-country regression analysis of lending to households, which is followed by an analysis of residuals for transition countries aimed at identifying the specific determinants of household lending in transition countries, and a concluding section.

MOTIVATING THE CROSS-COUNTRY ANALYSIS: THE CASE OF CROATIA

Like most of the transition countries, Croatia emerged from early transition with a minimal stock of household loans, roughly 5% of GDP in 1995. Since then, household lending has grown rapidly, rather faster than lending to enterprises. Household lending was a bit more than 34% of GDP at the end of 2005 (Figure 1).

[FIGURE 1 OMITTED]

The restructuring of Croatia's banking system was accelerated by a wave of bank failures starting in 1998-1999. The failures were also one of the triggers of a recession that lasted from the fourth-quarter of 1998 through the third-quarter of 1999. During the recovery, the government privatised three of the four largest banks. By the end of 2000, banks with majority foreign owners controlled 84.3% of total assets in the Croatian banking system. This combination of economic recovery and ownership transformation resulted in sharp improvements in bank performance, leading to substantial increases in credit availability.

On the funding side, the euro conversion process in the late 2001 brought a substantial inflow of deposits, as euro-legacy currency held 'in mattresses' flowed into bank deposits. Foreign exchange deposits grew by 2.8 billion euros (ECB, 2002). But after this extraordinary inflow, deposit growth simply was not adequate to fund banks' credit expansion plans. Banks closed the gap with extensive foreign borrowing.

The persistent interest-rate gap between Croatia and the Eurozone is a key explanatory factor in the persistent growth in lending in general, and in household lending in particular. Since 2000, interest earned on lending in Croatia has been far higher than on lending in the Eurozone (6%-10% versus 3%-5%), so that earnings (unadjusted for risk) are higher.

At the same time, with the ECB main reverse repo rate at 2% after June 2003, and exchange-rate pressures on the kuna more often on the appreciation side, borrowing on the European market at less than 4% was a useful complement to deposit funding. However, it must be kept in mind that one of the key developments in this period has been the increasing importance of long-term lending, most notably mortgage (housing) lending. This type of lending creates funding issues, since long-term sources of Croatian kuna are difficult to come by. Thus maturity, and not only price, is one of the important drivers of bank foreign borrowing.

In the light of the rapid growth of household lending, it is important to look at how the quality of household loans has held up. Somewhat surprisingly, it seems that quality has actually improved over time (see Figure 2). The usual 'seasoning effect' has not kicked in, at least yet.

[FIGURE 2 OMITTED]

A more detailed breakdown is available starting in the early 2004. The most rapidly growing subgroup is housing loans (37.4% of total loans as of end of 2005), which also has the lowest past-due rate (under 0.5%). The increasing share of housing loans provides further reason to suspect that loan quality will hold up in the future, since such loans are the last ones that households will fail to repay. By contrast, the category with the highest default rate is 'other', which includes loans for white goods as well as overdrafts on current accounts. The consequences of default on such loans are far less cataclysmic than those of default on loans for one's house or car, and a high and growing share of this kind of loans would be more of a cause for concern.

What can account for the improved quality of household loans in Croatia? Obviously, cyclical factors (sustained growth since 2000) should come first. In addition, it is important to know that loans to households in Croatia have traditionally come with very stringent conditions. Either a co-debtor or two guarantors (or both) were often required on many loans. In addition, collateral levels have been very high, with banks sometimes taking real estate worth substantially more than the loan amount, or requiring the holding of compensating balances deposits at the bank. (3) Although these requirements are now being eroded by an ever-growing competition, they remain common.

Two other considerations deserve mention here. First, due to the large unofficial economy, and strong family ties that bind Croatians in Croatia with relatives in more prosperous countries such as Germany, Switzerland and Australia, both the income and wealth of Croatian borrowers is probably underestimated by official statistics.

Second, the distribution of credit is biased towards wealthier households. Using data from the Household Budget Survey of the Central Statistical Office, Croatian National Bank (2006) shows that the debt burden is highest in the seventh to tenth (ie highest) income deciles (see Figure 3). These deciles account for some 65% of total disposable income, so it is clear that they are the ones who also have received the lion's share of household credit. Furthermore, the absolute amount of lending to the first decile is lower than lending to the seventh decile by a factor of about 7, so that the high indebtedness of the lowest decile need not present a problem for the banking system as a whole.

[FIGURE 3 OMITTED]

The Croatian National Bank became seriously concerned about rapid lending growth in the late 2002. It imposed a 'tax' on lending growth above a 4% quarterly growth rate in 2003. This measure did slow lending growth, but was accompanied by a surge in foreign borrowing by banks and enterprises. In the mid-2004, the central bank responded with a marginal reserve requirement on increments in banks' foreign liabilities. This marginal reserve requirement was raised in several steps, reaching some 55% in early 2006. While the rate of growth of foreign borrowing did slow, the country's foreign debt-to-GDP ratio remained above 80%, and the current account deficits in 2005 and 2006 were uncomfortably high. (4)

To summarise, in the Croatian experience, predictions of prudential problems resulting from the household lending boom have not materialised, but macroeconomic problems have.

HOUSEHOLD LENDING IN TRANSITION COUNTRIES

While cross-country data on loan quality are difficult to obtain and generally too methodologically heterogeneous to compare rigorously, it does seem that other countries seem to be sharing Croatia's experience of rapid household lending growth without significant worsening of aggregate loan quality. For example, Duenwald et al. (2005) do not see significant prudential effects of rapid loan growth in Bulgaria and Romania. Although they do express concern about financial stability in Ukraine, it is not clear how large the role of household lending is in this respect. In any case, Duenwald et al. spend much more time worrying about macroeconomic effects.

Of course, the situation in the transition countries is quite heterogeneous. Figure 4 shows the household credit to GDP ratio for the transition countries as of 2005. It is immediately apparent that quite a few transition countries still have extremely low stocks of household credit. In particular, Albania, Armenia, Azerbaijan, Georgia, the Kyrgyz Republic and Romania have levels below 5% of GDP, and Belarus is just above. These are all countries whose transitions have been slow and, in some cases, fraught with setbacks (eg the Albanian pyramid schemes, and armed conflict in Armenia, Azerbaijan, and Georgia, among others).

[FIGURE 4 OMITTED]

At the other extreme, Croatia and Estonia have ratios above 30%. This is above Italy's 28%, and not so far away from such EU-15 members as Greece (36%), Belgium (41%) and Austria (44%).

However, household credit in transition countries has grown extremely rapidly. This can be seen by plotting the size of household loan growth relative to GDP (Figure 5). Here, several of the countries with larger stocks stand out: Estonia, with growth of over 7 percentage points of GDP per year in 2004-2005, and Bulgaria and Lithuania, with growth of more than 4 percentage points of GDP per years in 2004-2005.

[FIGURE 5 OMITTED]

CROSS-COUNTRY DETERMINANTS OF HOUSEHOLD LENDING

This brief data survey shows clearly that household credit is growing rapidly in most of the transition countries, and that, while some transition countries have very low stocks of household credit, in some other transition countries stocks have caught up to some of the highly developed European economies. These observations lead me to pose two questions:

(1) To what extent are the stocks of household credit in transition countries still below 'normal'?

(2) What policy variables, if any, can explain the extent to which different transition countries stocks vary from the 'normal' level?

To answer the first question, in this section I will use a broad cross-country data to analyse the cross-country determinants of lending to households. This analysis will produce a cross-country curve that can serve as a benchmark against which transition country stocks of lending can be judged.

To answer the second question, in the next section I will analyse the residuals for transition countries from the cross-country regressions.

Turning to the cross-country analysis, I choose to model the stock variable household lending to GDP, as an indicator of the degree of development of this aspect of the financial system. This choice is inspired by the finance-growth literature finding that overall bank credit to GDP is a powerful explanatory variable for long-term growth (King and Levine, 1993; Levine et al., 2000; Wachtel, 2001; Rousseau, 2002). Household credit to GDP is of course a subcategory of the broader credit to GDP ratio.

The econometric strategy followed here is to use a broad sample of countries to study the main determinants of household lending. The use of a broad sample also allows us to pin down the degree to which transition countries lag behind in providing household lending, controlling for the main cross-country determinants of such lending.

Intuitively, it is clear that the provision of household credit should be a function of income, measured as GDP per capita. (5) In countries where much of the population is close to subsistence, households would have very limited ability to repay credit. Although the microcredit movement has developed a technology for lending to the very poorest, the aggregate amount of such lending remains very small.

While income levels will turn out to be powerful explanatory variables in explaining credit to households, we need to examine factors that can explain the development of the financial system in general. To begin with, macroeconomic stability should be relevant to the level of household credit. High inflation, in particular, often decreases the deposit base, and banks' inclination to lend. To take this into account, I include the cumulative increase in consumer prices over the five prior years. That is, I relate the ratio household credit to GDP in 2003 to the ratio of the price level in 2002 to the price level in 1998. For scaling purposes, I use the logarithm of this variable in the regressions.

Another important factor is that the structure of enterprise lending might affect household lending. In advanced countries, enterprises rely more heavily on direct finance, although of course the degree of reliance on direct finance varies dramatically between the capital markets-based financial systems such as the US and UK and the bank-based systems of Germany and Japan. Still, there might be a positive correlation between stock market development and credit to households if high levels of stock market development are associated with fewer corporate lending opportunities for banks that are then more eager to lend to households. Conversely, heavy corporate demand due to low stock market capitalisation might compete with lending to households.

However, it is reasonable to question the strength of the affect of enterprise credit demand on the supply of loans to households. That is, can we argue that there is a given 'lump of credit' to allocate in the short-run? I suggest that credit is difficult to expand beyond certain limits in the short-run, whether due to funding constraints, limited bank managerial and loan processing capacity, and costs of opening new branches or distribution outlets. (6)

Much recent research has emphasised the role of strong institutions in fostering development in general, and financial development in particular. For this reason, I use the Transparency International Corruption Perceptions Index, which is available for a very broad sample of countries, as an indicator of institutional quality. In addition, I use the World Bank's estimate of the number of days required to start a business as an alternative index of business climate and institutional quality. A highly bureaucratised system, with many legal barriers to entry and therefore a long start-up time, would presumably be a harder system for banks to operate in, and have lower levels of household credit to GDP.

Additionally, I draw on the law and finance literature pioneered by La Porta et al. (1997, 1998). They suggest a strong role for legal factors in explaining the degree to which creditor rights are protected, which in turn helps explain financial development. La Porta et al study in detail creditor protection provisions in legislation, and suggest that broad differences in the degree of creditor protection can be explained by the origin of countries' legal systems. Countries modelled on English common law generally provide relatively strong protection, while countries modelled on French civil law provide rather weak protection. The German and Scandinavian legal families fall somewhere in between.

While the classification used by La Porta et al. has been criticised by Berkowitz et al. (2003), who emphasise the difference between 'receptive' and 'unreceptive' legal transplants, their classification is useful here because of its simplicity. I therefore include legal origin variables, but only for non-transition countries. Transition countries have drastically rewritten their legal codes since 1990, so that the question of legal origin seems less relevant for them. Also, as Pistor et al. (2000) show, enforcement is a key issue in transition countries, with the degree of enforcement perhaps being a more important indicator of legal system quality than laws on the books.

What about the overall level of household lending in transition countries? Large portions of the pre-existing stock of bank assets in transition countries, mainly claims on large socialist enterprises, were written off during transition. This, as well as high inflation in many of the transition countries, led to very low levels of credit to GDP in transition countries. Furthermore, transition countries faced the challenges of banking sector reforms, which required time to put in place.

Thus, the transition countries may not fit the cross-country pattern for the rest of the world. For this reason, I include a 0-1 transition dummy in the cross-country regression. In the next section, I examine the transition country residuals more closely.

The data comprise observations on 90 countries in the years 2003-2005. The time series is limited because relatively few countries provide longer time series for this category of lending. The major international source for financial data, the IMF's International Financial Statistics, does not break lending down by sector, so there is no single source for the main time series required for this analysis. The ECB began to publish data on consumption lending in 2003, and this seemed to be a convenient starting point. (7) Furthermore, some 23 transition countries have provided series on household lending in their publications for the years 2003-2005, allowing a relatively full coverage of the transition sample.

An important statistical caveat is that, while some countries classify loans purely by borrower (ie loans to households), others classify by use (ie consumption). The latter is what I would like to measure and both the European Central Bank data for all the Eurozone countries and data from the United States are defined in this way. However, data for the transition countries are on a borrower basis. This imparts an upward bias to the transition country data, since some loans to households are actually loans to single proprietors or small businesses that are used for productive purposes. As we will see, despite this bias, transition countries clearly lag the overall curve.

In addition, among transition countries, while there is some variation in the proportion of sole proprietors, there is no reason to believe that this variation is greater than the variations among the large sample of developing countries that also report data on a sectoral rather than a user basis.

Another caveat is that I am unable to include credit to households granted via leasing or cross-border borrowing. While these channels clearly are important for enterprise lending, and leasing is clearly of some importance to households, it is not clear how important cross-border lending to households is. Because of this data limitation, this study should be understood to be confined to bank lending to households (in some cases including finance companies and various types of savings cooperatives, housing banks, etc), and not to the broader issue of credit available to households in general.

A final caveat is that it would probably be useful to control for asset prices and in particular for the extent of home ownership. However, I was unable to find a reasonable cross-country data series to represent this.

The regressions use panel OLS, with White robust standard errors to correct for panel heteroskedasticity. Although the issue of causality between financial variables and growth has been an important issue in the finance-growth literature, in this context there is no strong case for arguing that household lending causes long-term growth. Clearly, household lending is tied to consumption, if households would otherwise be liquidity constrained. This implies a possible correlation between household credit and short-term fluctuations in GDP or GDP per capita, but these variables are also determined by a longer-term growth. Still, as a precaution to limit the chances of endogeneity, all explanatory variables are lagged.

In addition, the short data series makes it impossible to use the technique widely employed in empirical growth analysis of regressing 5-year averages of economic growth on initial condition variables for the year prior to the 5-year growth period. Also, because the dynamics of transition country catch-up are very rapid, S-year averages would actually make it more difficult to see the catch-up process than year-by-year variables.

In initial regressions, I noticed that multicollinearity between the Transparency International Corruption Perceptions Index and GDP per capita made it difficult to identify their effects. To avoid this problem, I regressed the Corruptions Perceptions Index on GDP per capita and then used the residuals from this regression as on explanatory variable. It is an estimate of the extra corruption holding the influence of GDP per capita on the index constant.

As a robustness check, I also constructed a longer series for the years 2000-2005. Unfortunately, to gain these three additional years, I lose about one-third of the sample: I was only able to gather data for 54 countries for 2000-2005.

The results of the cross-country regressions are shown in Table 1. Columns (1)-(4) provide results for the 2003-2005 sample, while column (5) shows results for the 2000-2005 sample.

As expected, GDP per capita proves a very powerful variable. The coefficient on GDP per capita does not vary substantially with the inclusion of other variables. Past inflation inhibits household lending, as expected. Latin American countries show lower levels of household credit, and English and German legal origin contribute positively. The adjusted corruption index proves highly significant. However, a longer time to start a business, as a sign of a weaker legal environment and also lower business credit demand, turns out to be insignificant. The development of securities markets, measured by the market capitalisation of listed companies as a share of GDP, is positively associated with household lending. This could be either a substitution effect (less business demand for loans), or a supply-side effect (stronger financial markets).

The results for the longer sample from 2000 to 2005 confirm the shorter sample results. However, it should be noted that the longer time series includes only 15 transition countries, as compared to the 23 included in the 2003-2005 sample.

In short, a high level of income, a history of macroeconomic stability in the recent past, and strong institutions (low corruption and English or German law) all contribute to higher household lending.

Finally, the transition dummy is consistently significant and negative, but shrinking from 2003 on. This shows that transition countries indeed lag behind established market economies in the provision of household credit, once other major determinants are controlled for, but that the lag is decreasing for the group of transition countries as a whole. Using the coefficients in equation 1, the equation implies that a non-transition country would have had 62 % higher household loans than a transition country with the same characteristics in 2003. This gap then fell to 31% and only 18% in 2004 and 2005, respectively, a remarkably rapid convergence.

An examination of the residuals for individual transition countries shows that Croatia and neighbouring Bosnia-Herzegovina have high positive residuals that are outside the normal confidence bounds. In other words, Croatia and Bosnia-Herzegovina seem to be outliers at this stage. Albania (2003 and 2004}, the Kyrgyz Republic (all years) and Romania (2003 and 2004) are outliers at the lower end of the spectrum, indicating exceptionally slow development of household lending even for transition countries.

REFORM AND HOUSEHOLD LENDING: ANALYSIS OF RESIDUALS

In this section, I try to explain the variation in the residuals for transition countries in the cross-country analysis above. I subject the residuals from the cross-country regression for 2003-2005 for transition countries to regression analysis, using reform progress indicators as explanatory variables. Since these residuals represent deviations from a cross-country regression that takes structural factors such as level of development and legal system into account, and also incorporates an across-the-board transition effect, variation in the residuals may to a large extent be explained by differences in the progress made in reform among the transition countries. By narrowing the sample down to the transition countries only, I am able to use the EBRD's transition reform indices, one of the few quantitative measures of reform progress in existence.

The analysis of residuals looks at three sets of factors: (1) the relative strength of the banking system, indicative of loan supply, (2) the degree of privatisation and enterprise reform, as an indicator of enterprise credit demand and enterprise credit worthiness and (3) the development of nonbank financial intermediation as a substitute to bank lending. The variables are taken from the EBRD's Transition Report, and are available for 23 transition countries.

The basic hypotheses of the transition residuals analysis are that greater progress in banking reform will lead to higher residuals (higher levels of household lending than predicted by the cross-country analysis), while greater progress in enterprise reform and privatisation will lead to greater enterprise credit demand and lower residuals. Similarly, greater progress in development of non-bank financial intermediaries is expected to result in less household lending, as non-bank financial intermediaries provide a larger role in the household credit market.

Note that the overall cross-country analysis included two policy-related variables: cumulative inflation and corruption. More precisely, these variables represent outcomes of policies and therefore reflect both the policies adopted and the economic and social conditions in the countries. In any case, since these variables are universal ones that can be expected to affect all the countries in the sample, they were included in the large data set. The transition country variables reflect specific reform challenges facing transition countries. Finally, to minimise possible endogeneity, residuals in 2005 are explained by reform indicators for 2004.

Table 2 shows the regression estimates where the dependent variable is the residuals from equation 2 in Table 1 (the cross-country equation with the highest [R.sup.2]). However, the results from other specifications are similar. Only two explanatory variables are entered in each regression because of strong multicollinearity among the explanatory variables.

Expectedly, banking reform, representing the supply side of lending, is highly significant. However, so are all four of the indicators of the enterprise demand side: percent of GDP produced by the private sector, EBRD scores for large-scale and small-scale privatisation and enterprise reform. The development of non-bank financial institutions turns out to have a positive coefficient, something of a surprise. This could either reflect complementarities between the development of banks and non-banks, or may simply be the result of multicollinearity.

Furthermore, examination of the residuals from equation 1, which has the highest adjusted [R.sup.2], shows that that the country with the largest positive residuals is Bosnia-Herzegovina. Bosnia-Herzegovina is clearly a case of unbalanced reforms. Although its banking reform score was only 2.3 in 2002, which was better than Belarus and Russia, and equal to the scores for four other countries (Albania, Armenia, Serbia and Montenegro and Ukraine), Bosnia-Herzegovina scored next to last on small-scale privatisation, only better than Belarus. Bosnia-Herzegovina's share of the private sector was also tied for next to last, suggesting very slow privatisation progress indeed.

The Kyrgyz Republic's residuals are strong negative outliers in 2003-2004, but in 2005 move to roughly zero. Romania is also a negative outlier in 2003-2004, and, intriguingly, so is Slovenia. The Kyrgyz Republic and Romania could perhaps be categorised as slow starters in household lending: once household lending began to grow rapidly, they quickly moved into the mainstream of transition countries. However, Slovenia is a more difficult case to understand. It seems that household lending is well below what would be predicted from the country's high GDP per capita and strong reform record. One might speculate that the strong role of the Slovene government in ownership of the largest banks might have bolstered lending to enterprises as opposed to lending to households.

Finally, it is interesting that the two countries with the highest levels of household credit, Croatia and Estonia, have residuals within the confidence bands. What this tells us is that the high levels of household credit in Croatia and Estonia can be explained by reforms and policy outcomes. Both countries have high scores on banking reform, and reasonably good inflation records. Croatia's performance on privatisation and enterprise reform has been weaker than Estonia's, suggesting lower enterprise credit demand.

CONCLUDING DISCUSSION

This paper studies the implications of the rapid increase in household lending in transition economies in the last several years. The Croatian case study suggests that rapid household loan growth does not necessarily lead to major prudential problems. The bulk of lending is skewed towards wealthier households, which, so far, seem to be able to service their obligations. Furthermore, much of lending growth is in the area of mortgage lending, which tends to be less problematic both because of the high collateral values involved and because of the strong commitment of households to protecting their homes from foreclosure. Of course, all of this comes with major caveats, since it will only be after the next recession that we will be able to fully see the adequacy of banks' provisioning and capitalisation policies. The presence of reputable foreign banks, mainly from the EU-15, gives some comfort here, but is no guarantee of sound risk-management policies by the banks. And, of course, the situation may vary somewhat in other transition countries.

If prudential concerns are muted, it seems that the main threat to financial stability posed by the household lending boom comes through macroeconomic effects. By increasing consumption, the household lending boom can exacerbate current account problems. Over a period of years, lopsided allocation of credit to households, as opposed to firms, could lead to lagging productivity growth and slower output growth than what would be available with a more balanced allocation of credit. (8)

The cross-country regressions provided above suggest that successful macroeconomic policy stimulates consumer lending to the extent that it achieves low rates of inflation. Similarly, successful banking reforms, and the achievement of lower levels of corruption, also stimulate household lending. In other words, one of the fruits of successful reforms is higher household lending.

At the same time, countries can become victims of their own success to an extent. For while all of these things are positive in and of themselves, if progress along these dimensions listed above outstrips progress in enterprise privatisation and restructuring, the stage can be set for lopsided, consumer-oriented credit allocation.

The challenge for policymakers in transition countries, then, is to ensure reform progress across the board. Privatisation and real sector restructuring often involve severe political obstacles and constraints. But failure to make progress in these areas will ultimately result in failure to improve living standards, slower growth and problems with external debt. Strong banking reform can raise living standards temporarily, but if more wealth is not produced, increased consumption levels and external imbalances may become unsustainable.

While it would certainly be wise to work towards broad reform progress, and not just progress in macroeconomic stability and banking reform, it is realistic to expect that substantial progress in macrostability and banking reform is likely to lead to stronger consumption growth. Accelerations in imports and foreign borrowing also become very likely. Prudent macroeconomic policymakers should therefore adjust their projections accordingly, and consider proactive measures.

For some of the transition countries, the process of negotiating accession to the European Union, and the process of gradual adjustment to EU norms via the Stabilisation and Association Process, should provide a useful framework for planning and implementing the needed reforms. The countries of the former Soviet Union studied here (Ukraine, Russia, Georgia, Armenia, Azerbaijan, Kazakhstan and the Kyrgyz Republic) however, have much smaller chances, if any, of EU membership in the foreseeable future, and will need to find a different set of motivations for their reform processes. Nonetheless, they face the same danger if they are more successful in banking reform than in real sector reform.

In either case, the authorities are likely to face unpleasant choices between allowing household lending, and current account deficits and foreign debt ratios, to expand at uncomfortable speed, and implementing restrictive policies, including administrative measures such as those used by Croatia. The use of administrative measures may improve macroeconomic stability in the short-run. But there may well be trade-offs between greater short-term stability and the distortions created by administrative measures. Such distortions are likely to be inimical to long-term market development, and thus the stability gains must be weighed carefully against the development losses. Unfortunately, there seems to be little chance of avoiding such unappealing dilemmas in practice.

DATA APPENDIX

Table Al: Descriptive statistics for variables used in
the cross-country analysis

 Mean Maximum Minimum

Household loans/GDP 0.266 1.087 0.004
GDP per capita 13961 77784 113
Inflation (cumulative 1.71 4.20 0.86
price change in
preceding 5 years)
Corruption Score 5.12 9.70 1.80

Legal origin
 English 0.22 1.00 0.00
 German 0.04 1.00 0.00
Stock market 65.53 528.59 0.13
capitalisation
Time to start a 40.33 152.00 2.00
business

 Observations Source

Household loans/GDP 255 Central banks, IFS
GDP per capita 257 Central banks, IFS
Inflation (cumulative 266 Central banks, IFS
price change in
preceding 5 years)
Corruption Score 246 Transparency international
 (www.tranparency.org)

Legal origin
 English 270 LaPorta et al. (1997)
 German 270 LaPorta et al. (1997)
Stock market 150 World Bank Development
capitalisation Indicators
Time to start a 233 World Bank Development
business Indicators

Table A2: Transition variable descriptive statistics

 Mean Maximum Minimum

Cross-country equation residual 0.003 1.190 -1.688
EBRO banking reform 2.89 4.0 1.7
% of GDP produced by 66.38 80 25
private sector
EBRD large-scale privatisation 3.13 4.0 1.0
EBRD small-scale privatisation 3.90 4.3 2.0
EBRD enterprise reform 2.46 3.3 1.0
EBRD non-bank reform 2.42 3.7 1.7

 Observations Source

Cross-country equation residual 69 Author's calculation
EBRO banking reform 69 EBRO transition report
% of GDP produced by 69 EBRO transition report
private sector
EBRD large-scale privatisation 69 EBRO transition report
EBRD small-scale privatisation 69 EBRD transition report
EBRD enterprise reform 69 EBRD transition report
EBRD non-bank reform 69 EBRD transition report

REFERENCES

Baccheta, P and Gerlach, S. 1997: Consumption and credit constraints: International evidence. Journal of Monetary Economics 40: 207-238.

Backe, P and Zumer, T. 2005: Developments in credit to the private sector in Central and Eastern European EU member states: Emerging from financial repression--A comparative overview. Focus on Transition 2:83 109.

Berkowitz, D, Pistor, K and Richard, J-F. 2003: Economic development, legality and the transplant effect. European Economic Review 47(1): 165-195.

Borio, C and Lowe, P. 2002: Asset prices, financial and monetary stability: Exploring the nexus, BIS Working papers no. 114, July 2002.

Caprio, G and Klingebiel, D. 1996: Bank insolvency: Bad luck, bad policy or bad banking. Annual Bank Conference on Developing Economies: 79-114.

Cottarelli, C, Del'Ariccia, G and Vladkova Hollar, I. 2003: Early birds, late risers, and sleeping beauties: Bank credit growth to the private sector in Central and Eastern Europe and the Balkans, IMF Working paper WP/03/213, November 2003.

Corricelli, F, Mucci, F and Revoltella, D. 2006: Household credit in the New Europe: Lending boom or sustainable growth, CEPR Working paper no. 5520, March 2006.

Croatian National Bank. 2006: Macroprudential Analysis, Number 3, September, p. 8.

Duenwald, C, Gueorguiev, N and Schaechter, A. 2005: Too much of a good thing? Credit booms in transition economies: The cases of Bulgaria, Romania and Ukraine, IMF Working papers WP/05/ 128, June 2005.

Eichengreen, B and Rose, A. 1998: Staying afloat when the wind changes: External factors and emerging-market banking crises, NBER Working paper 6730, January 1998.

Eichengreen, B and Arteta, C. 2000: Banking crisis in emerging markets: Risks and red herrings. In: Blejer, M and Skreb, M (eds). Financial Policies in Emerging Markets. MIT Press: Cambridge, pp. 47-94.

European Bank for Reconstruction and Development. 1998: Transition Report: London, various issues.

European Central Bank. 2002: Review of the International Role of the Euro. European Central Bank: Frankfurt.

Galac, T. 2005: Results of the third CNB bank survey: Croatian banks in the consolidation and market positioning stage, 2000-2002, Croatian National Bank Survey 8-14, December 2005.

Galac, T and Dukic, L. 2005: Rezultati cevurtoga HNB-ova anketiranje banaka, Pregledi Hrvatske Narodne Banke P 20, August 2005 (in Croatian only).

Gourinchas, P-O, Valdes, R and Landerretche, O. 2001 : Lending booms: Latin America and the world, NBER Working paper 8249, April 2001.

Herrmann, S and Jochem, A. 2005: Determinants of current account deficits in the Central and East European EU member states--consequences for the enlargement of the euro area. Deutsche Bundesbank Discussion paper no. 32/2005.

King, R and Levine, R. 1993: Finance and growth. Quarterly Journal of Economics 108: 717-737.

Kraft, E. 1998: Credit policies of Croatian Banks, Croatian National Bank Survey #8 April 1998.

Kraft, E. 2000: The lending policies of Croatian Banks: Results of the Second CNB Bank interview project, Croatian National Bank Survey S-3, December 2000.

Kraft, E. 2006: How competitive is Croatia's Banking system? Croatian National Bank Working Papers W-14, March 2006.

Kraft, E and Jaukov, L. 2005: Does speed kill? Lending booms and their consequences in croatia. Journal of Banking and Finance 29 (l): 105-121.

La Porta, R, Lopez-de-Silanes, F and Schleifer, A. 1997: Legal determinants of external finance. Journal of Finance 52: 1131-1150.

La Porta, R, Lopez-de-Silanes, F and Shleifer, A. 1998: Law and finance. Journal of Political Economy 106(6): 1113-1155.

Levine, R, Loayza, N and Beck, T. 2000: Financial intermediation and growth: Causality and causes. Journal of Monetary Economics 46:31-77.

Pistor, K, Raiser, M and Gelfer, S. 2000: Law and finance in transition economies. Economics of Transition 8(2): 325-368.

Rousseau, P. 2002: Historical perspectives on financial development and economic growth, NBER Working paper 9333, November 2002.

Wachtel, P. 2001: Growth and finance--What do we know and how do we know it? International Finance 4: 335-362.

World Bank Development data (www.devdata.worldbank.org/data-query/).

(1) The views expressed in this paper are the author's, and do not necessarily represent those of the Croatian National Bank. The author would like to thank Maxwell Watson, Paul Wachtel and participants of the 12th Dubrovnik Conference on Transition Economies, as well as an anonymous referee, for helpful comments. All remaining errors are the author's sole responsibility.

(2) For an introduction to banking reform in transition countries, see EBRD Transition Report (1998), and the issues thereafter. For recent findings on competition in Croatia, see Kraft (2006).

(3) For details on these practices, see the Croatian National Bank surveys: Kraft (1998, 2000), Galac (2005), Galac and Dukic (2005).

(4) For further discussion of the macropolicy issues, see Kraft and Jankov (2005).

(5) I use GDP at market prices, because it is more readily available titan GDP at PPP. Experiments using GDP at PPP for 1 year showed no major differences in regression results compared to GDP at market prices.

(6) I would like to thank Maxwell Watson for suggesting this point to me.

(7) I would like to thank Adalbert Winkler from the ECB for providing me with the ECB data and kindly explaining its background.

(8) I thank Maxwell Watson for emphasising this point to me.

EVAN KRAFT

Croatian National Bank, Trg Hrvatskih Velikana 3, Zagreb 10002, Croatia.

E-mail: [email protected]

Table 1: Cross-country determinants of lending to households dependent
variable log (household loans/GDP)

 (1) (2) (3)

Constant -6.99 ** -6.78 ** -6.97 **
 (36.42) (47.18) (23.33)
Log (GDP per capita-1) 0.60 ** 0.56 ** 0.59 **
 (18.95) (32.21) (21.56)
Log (cumulative price change -0.31 ** -0.29 ** -0.31 **
in previous 5 years) (16.23) (12.72) (15.58)
Transition country dummy -0.48 ** -0.44 ** -0.44 **
 (11.67) (17.81) (8.69)
Transition dummy x 2004 dummy 0.21 ** 0.23 ** 0.18 **
 (42.00) (38.80) (40.37)
Transition dummy x 2005 dummy 0.31 ** 0.32 ** 0.29 **
 (40.19) (37.33) (46.50)
TI corruption index 0.20 ** 0.16 ** 0.20 **
(lagged, adjusted for GDP per capita) (10.37) (7.95) (8.89)
2004 dummy 0.05 0.03 ** 0.07 **
 (18.64) (9.74) (9.70)
2005 dummy 0.16 ** 0.14 ** 0.18 **
 (31.95) (29.89) (15.74)
Latin America -0.22 **
Dummy (8.96)
English legal origin dummy 0.41 **
 (8.64)
German legal origin dummy 0.50 **
 (18.10)
Time required to start-up company -0.00
 (0.60)
Market cap of listed companies
Transition dummy x 2001 dummy
Transition dummy x 2002 dummy
Transition dummy x 2003 dummy
2001 dummy
2002 dummy
2003 dummy

Total observations 239 239 239
Countries included 83 83 83
Adjusted [R.sup.2] 0.750 0.774 0.751
F-Statistic 90.15 75.05 72.56
Probability (F) 0.000 0.000 0.000

 (4) (5)

Constant -7.06 ** -7.71 **
 (94.62) (29.10)
Log (GDP per capita-1) 0.61 ** 0.68 **
 (69.58) (24.48)
Log (cumulative price change -0.58 ** -0.55 **
in previous 5 years) (132.34) (11.53)
Transition country dummy -0.38 ** -0.48 **
 (9.11) (10.02)
Transition dummy x 2004 dummy 0.33 **
 (12.82)
Transition dummy x 2005 dummy 0.45 **
 (16.20)
TI corruption index 0.18 **
(lagged, adjusted for GDP per capita) (8.55)
2004 dummy 0.06 ** 0.15 **
 (11.76) (26.01)
2005 dummy -0.35 ** 0.22 **
 (0.58) (27.79)
Latin America
Dummy
English legal origin dummy

German legal origin dummy

Time required to start-up company

Market cap of listed companies 0.00 **
 (8.33)
Transition dummy x 2001 dummy -0.08 **
 (8.91)
Transition dummy x 2002 dummy -0.05 **
 (4.11)
Transition dummy x 2003 dummy 0.16 **
 (6.56)
2001 dummy -0.03 **
 (10.36)
2002 dummy 0.05 **
 (11.67)
2003 dummy 0.13 **
 (28.28)

Total observations 145 315
Countries included 75 54
Adjusted [R.sup.2] 0.749 0.805
F-Statistic 71.45 93.51
Probability (F) 0.000 0.000

White cross-section standard errors, absolute values of t-statistics
in parentheses.

* Significant at 5%, ** significant at 1%.

Table 2: Analysis of transition country residuals

 (1) (2) (3)

Constant 0.87 -0.60 * 0.62
 (1.08) (8.89) (6.38)
EBRD banking reform score 0.69 ** 0.67 ** 0.66 **
 (11.43) (5.06) (9.87)
Percent of GDP produced by the -0.03 **
 private sector (20.06)
EBRD large-scale privatisation score -0.44 **
 (9.00)
EBRD small-scale privatisation score -0.66 **
 (19.19)
EBRD enterprise reform score

EBRD non-bank financial
 institution reform score
Number of observations 69 69 69
Countries included 23 23 23
Adjusted [R.sup.2]-squared 0.435 0.258 0.422
Durbin-Watson 0.276 0.219 0.336
F-statistic 25.45 12.82 25.80
Probability (F) 0.000 0.000 0.000

(Equation 2, Table 1).

 (4) (5)

Constant -0.88 ** -0.99 **
 (8.57) (11.73)
EBRD banking reform score 0.59 ** 0.24 **
 (14.23) (5.96)
Percent of GDP produced by the
 private sector
EBRD large-scale privatisation score

EBRD small-scale privatisation score

EBRD enterprise reform score -0.35+
 (27.69)
EBRD non-bank financial 0.11 **
 institution reform score (4.09)
Number of observations 69 69
Countries included 23 23
Adjusted [R.sup.2]-squared 0.209 0.110
Durbin-Watson 0.220 0.152
F-statistic 10.00 5.16
Probability (F) 0.000 0.008

(Equation 2, Table 1).

White cross-section standard errors, absolute values of
t-statistics in parentheses.

* Significant at 5%, ** significant at 1%.