Austria: Selected Issues

This Selected Issues paper estimates a dynamic model of foreign currency loans to households in Austria to analyze their behavior and assess the effectiveness of measures intended to stem their rise. This paper also studies the developments in Austria’s economic linkages with Germany and the Central and Eastern European countries (CEECs). It finds that there has been delinking from Germany, albeit measured, while economic relationships with key CEEC trading partners have become stronger. The paper also discusses the dynamics of Austria’s economic linkages with Germany, and examines these linkages with the CEECs.

Abstract

This Selected Issues paper estimates a dynamic model of foreign currency loans to households in Austria to analyze their behavior and assess the effectiveness of measures intended to stem their rise. This paper also studies the developments in Austria’s economic linkages with Germany and the Central and Eastern European countries (CEECs). It finds that there has been delinking from Germany, albeit measured, while economic relationships with key CEEC trading partners have become stronger. The paper also discusses the dynamics of Austria’s economic linkages with Germany, and examines these linkages with the CEECs.

I. What Explains the Surge of Foreign Currency Loans to Austrian Households?1

Abstract: This study estimates a dynamic model of foreign currency loans to households in Austria to analyze their behavior and assess the effectiveness of measures intended to stem their rise. The results suggest that the very rapid growth of loans in foreign currency since 1995 represents a deviation from fundamentals: economic (i.e., credit demand and currency substitution) factors alone cannot explain recent developments. However, once the model has been augmented to account for the formation of herd behavior, a relatively stable equilibrium relationship for foreign currency loans emerges. Supply factors appear to have played a role in the rapid spread of these loans. Statistical tests suggest that the public information campaign launched by the Austrian National Bank (OeNB) as well as strengthened prudential oversight have yet to succeed in stemming the rise of these loans.

A. Introduction

1. Following years of relative stability, foreign currency loans to Austrian households entered a phase of explosive growth around 1995 (Figure 1). Even though their growth has moderated since 2001, foreign currency loans still account for about half the growth of total credit to households. By end-March 2005, 30 percent of outstanding loans to households were in foreign currency, compared with about 5 percent for the euro zone. Nearly all these loans are in Swiss francs and, to a lesser extent, Japanese yen. The vast majority of these loans finance domestic transactions.2 Loans in U.S. dollars and other currencies—usually associated with external transactions—have not changed much in recent years. The popularity of foreign currency loans among households is a uniquely Austrian phenomenon in the euro zone.

Figure 1.
Figure 1.

Austria: Loans in Foreign Currency, 1987-2005

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Source: OeNB; WiFO; and IMF staff calculations.

2. Structural changes in Austria’s banking system that began in the mid-1990s led to intense competition, creating the conditions for the spread of these loans. A new Bank Act in 1994 removed the ability of banks to set common interest rates. Austria’s entry into the European Union in 1995 was accompanied by free bank entry and the removal of the remaining financial restrictions. During 1995–2000, virtually all state-owned banks were privatized. Moreover, the onset of fiscal consolidation in 1996—in an effort to meet the Maastricht criteria—and the achievement of a small budget surplus in 2001 reduced bank financing by 30 percent, prompting banks to expand their consumer loan portfolio. These developments sparked a merger activity and intense competition for market share, with banks using foreign currency loans as an instrument of competition (Braumann, 2004).

3. The boom in these loans appears to be reflecting currency substitution rather than a lending boom. Growth of credit to households (in all currencies) has been relatively subdued (second panel of Figure 1), suggesting that a considerable part of the growth of foreign currency loans has been the result of refinancing. Indeed, because of refinancing of previous schilling or euro loans, the contribution of foreign currency loans to the annual growth of total loans to households approached or exceeded 100 percent in six consecutive quarters beginning in late 1998.3 The spread of this phenomenon has puzzled many. Heightened concerns about risks prompted policymakers to strengthen prudential oversight and launch a public information campaign in 2003 (OeNB, 2003).

4. This study addresses the following questions:

  • Is this phenomenon reflecting an equilibrium rebalancing of portfolios or the buildup of financial imbalances? Is the stock of loans in foreign currency in Austria currently above the steady state equilibrium? If yes, how smooth is the return to equilibrium expected to be?

  • In what respects has Austria been different from other euro zone countries? Can Austria-specific factors be uncovered in a statistical way? Has herd behavior played a role?

  • Have banks contributed to this development? What can the statistical evidence tell us about the role of the banks?

  • Has the housing market been a driving force? How strong is the interaction between these loans and the house and mortgage markets? Could falling prices in the housing market cause serious balance sheet problems for consumers and banks?

  • How large are the measurable risks to the economy stemming from these loans? What could the repercussions of large shocks to interest and exchange rates be?

  • Can the impact of recent policy measures designed to arrest the growth of these loans be identified? Aside from addressing risk management of banks, have policy measures been effective in stemming the rise of these loans? What role can policy play in this area in a free market economy with an open capital account?

5. The literature on models of lending in foreign currency in advanced economies is thin. Moreover, to the author’s knowledge, there has been no empirical study on this topic for Austria. The voluminous literature on dollarization and currency substitution provides some insights on the choice of currency by households, but focuses mainly on developing and transition countries and on the asset—rather than the liability—side of household balance sheets.4 A possible explanation for the paucity of research on credit dollarization in advanced economies is that the need for understanding the driving forces of the demand for foreign currency loans becomes pressing only when the growth of these loans has already reached sizeable proportions. By then, herd behavior and bandwagon effects—which weaken the influence of fundamentals—have likely come into play, and the challenge in modeling herd behavior makes it difficult to investigate the demand for these loans. The problem of identifying credit demand and supply may also have contributed to the modeling problems.5

6. This study aims to contribute to the understanding of the reasons behind the boom in foreign currency loans in Austria in the following respects. First, it allows for a deviation of the equilibrium foreign currency loans from economic fundamentals by modeling the formation and development of herd behavior explicitly. Second, by recognizing that standard credit demand and currency substitution factors alone may not be sufficient to explain the boom in these loans, it investigates the role of the housing market in a formal way. Finally, it aims to assess in a statistical way the effectiveness of recent policy measures in this area.

7. This study formulates a model of the long-run equilibrium demand for loans that is enriched with short-term dynamics to describe the behavior of foreign currency loans to Austrian households. To put the problem in perspective, Section B discusses the main issues related to these loans. Section C provides a brief overview of the theoretical foundations of the model, as well as an interpretation of the data. Section D uncovers an equilibrium long-run relationship between loans to households in foreign currency and their fundamental determinants, and in the process evaluates the influence of individual factors on the behavior of these loans over the long run. Section E describes a foreign currency loan equation that captures both short- and long-run dynamics, and performs experiments to assess the impact on these loans of large shocks to the variables in the model. Section F summarizes the results. Finally, Section G contains concluding remarks. Box 1 summarizes the results presented in the following sections.

Summary of the Main Results

The model is intuitively appealing and fits the data well. The main results are as follows:

  • The strong uptrend in foreign currency loans since 1995 represents a significant deviation from fundamentals: economic (i.e., credit demand and currency substitution) factors alone cannot explain developments since 1995, but once the model has been augmented to account for the formation of herd behavior, a relatively stable long-run relationship for foreign currency loans emerges. These loans currently appear to be significantly above the steady state equilibrium level implied by economic fundamentals, but the return to equilibrium is likely to be slow.

  • The principal factor that distinguishes Austria from other euro zone countries seems to be the formation of herd behavior in the demand for foreign currency loans. The results suggest that the perception of exchange rate stability shaped by two decades of hard currency policy may also have played a role, albeit modest: the influence of the exchange rate on foreign currency loans is large in the long run but disappears in the short run.

  • Banks’ pricing policies may have been a factor behind developments in foreign currency loans. First, demand for these loans appears to have been met by falling interest rates as banks have likely made an effort to gain market share. Second, the transaction costs involved in refinancing these loans may have contributed to the estimated slow return to equilibrium.

  • It appears that housing market developments have not played a significant role.

  • The response of foreign currency loans to modest shocks (stemming, for example, from a depreciation of the exchange rate) should be relatively modest and gradual.

  • Policy measures addressing the demand and supply of foreign currency loans have not had a statistically significant impact on either the levels or growth rates of these loans.

B. Issues Related to Foreign Currency Loans to Households

8. The appeal of these loans is mainly in the lower interest rates on the borrowed currencies. Interest rates on these loans are linked to the London interbank offer rate (LIBOR) of the currency of the loan (Box 2). Money market rates for the main currencies of choice have been lower than those for the domestic currency virtually throughout the past 20 years (Figure 2). Despite the appreciation of the Swiss franc and Japanese yen over the same period, it appears that borrowers have been attracted by the lower interest payments and count on a stable exchange rate. Even though the degree of substitution between foreign currency and euro loans does not differ from that between loans in different foreign currencies, exchange rate movements have led to switches to other foreign currencies. The appreciation of the Japanese yen between late 1998 and late 2000 prompted a switch—albeit with a considerable lag—away from the yen to the more stable Swiss franc (bottom panel of Figure 2).

Features of Foreign Currency Loans

A typical foreign currency loan to an Austrian household

  • is used predominantly for purchasing a house and has a size of about EUR 100,000;

  • is secured by real estate collateral;

  • has a relatively low loan-to-value ratio;

  • has a maturity of up to 25 years;

  • has an adjustable interest rate linked to the LIBOR rate of the currency of the loan and is repriced every three to six months;

  • is a balloon loan (involving monthly payments of interest only, with full principal paid at maturity);

  • has a narrower interest margin and carries higher fees than a comparable euro loan;

  • offers the option of switching currencies—typically at repricing dates—for a 1-2 percent conversion fee;

  • has forced conversion clauses, allowing conversion into a euro loan without the borrower’s consent; and

  • requires the establishment of a repayment vehicle (usually a life insurance contract or a mutual fund) through which monthly payments are made and which is to be used to repay the principal at maturity.

Sources: Financial Market Authority; OeNB; IMF (2004); and Waschiczek (2002).
Figure 2.
Figure 2.

Austria: Interest Rate Differentials and Exchange Rates, 1987-2004

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Sources: OeNB; WiFO; IMF, IFS: Information Notice System; and IMF staff calculations.1/ Domestic less foreign interest rate. The foreign rate is a weighted average of the Swiss franc, Japanese yen, and U.S. dollar rate, with weights equal to the share of these currencies in total loans to households in foreign currency. For the period before 1997, the 1997 weights were used.2/ An increase denotes an appreciation of the euro (Austrian schilling up to end-1998).3/ An increase denotes an appreciation of the Swiss franc.

9. With common interest and exchange rates in the other euro zone countries, why have these loans taken off only in Austria? The practice of borrowing in foreign currency (mainly Swiss francs) began in the western part of the country, where tens of thousands of Austrians commute to work in Switzerland and Liechtenstein. This partly explains why the share of these loans was higher in Austria, even during the 1980s. Word of mouth and aggressive promotion by financial advisors helped spread the popularity of these loans to the rest of the country. By the mid-1990s, newspaper ads placed by banks began to appear, fueling public interest (text figure). Another factor facilitating the spread of these loans could be the belief in the stability of the exchange rate deriving from Austria’s hard currency policy (a peg of the schilling to the deutsche mark) since 1980.6 The success of this policy may have created a psychology of an exchange rate immune from risks, notwithstanding the appreciation of the Japanese yen and, to a lesser extent, the Swiss franc since the mid-1980s.

uA01fig01

Austrian Press Agency News Stories on Foreign Currency Loans, 1987-2004 1/

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Source: Factiva’s NewsPlus.1/ Number of news reports (in German) by the Austrian Press Agency that included the word “fremdwaehrungskredite” (foreign currency loans). Waschczek (2002) has presented the results of a similar search of news reports.

10. At first glance, Austria’s house and mortgage markets do not appear to explain the boom in these loans. The experience of other countries suggests that house prices influence mortgage borrowing (the predominant liability of households) through two channels. First, a rise in house prices typically requires a larger mortgage to acquire a given property. Second, fluctuating house prices affect the capacity of households to borrow through the change in the value of their collateral (mortgage equity withdrawal). This relationship has become stronger since the 1980s, as financial reforms have increased households’ access to mortgage borrowing. However, some mitigating factors appear to be at work in the house and mortgage markets in Austria. The boom in house prices that began in the mid-1980s halted in the first half of the 1990s as the supply of housing increased (text figure). House prices have edged up somewhat recently, but remain well below their long-run average.7 Moreover, Austria’s owner occupation rates are among the highest and outstanding mortgage debt one of the lowest in Europe. This could imply a weak transmission mechanism from the house and mortgage markets to the rest of the economy. However, house prices could still have played a role in helping foreign currency loans take off (a hypothesis that can be tested empirically), even though they do not appear to have played a role in recent years.8

uA01fig02

House Prices, 1987-2003

(1995=100)

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

11. Foreign currency loans expose households to a host of risks (Box 3). The experience of other countries—for example, Australia, and, to some extent, the United Kingdom in the 1980s—suggest that severe balance sheet problems, mainly for households, could develop in the event of a large and rapid exchange rate movement, with repercussions felt throughout the economy.9 Concerned about these risks, the OeNB and the Financial Market Authority (FMA) implemented a number of measures: the OeNB launched a public information campaign in mid-2003 and, together with the FMA, intensified prudential oversight and took steps to strengthen banks’ risk management.

Risks Associated with Loans in Foreign Currency

Households

Loans in foreign currency expose households to the following risks:

  • Exchange rate risk, from households’ large, unhedged exposure. The switch away from the Japanese yen to the more stable Swiss franc has mitigated this risk. Nevertheless, this risk is still not trivial and can be aggravated by the fact that the full principal is not due until maturity, which may be many years away. While, in principle, the repayment vehicle could be used to mitigate this risk, in practice, multicurrency exposure often occurs when households invest in a third currency.

  • Interest rate risk from the adjustable-rate loans with frequent repricing dates.

  • Additional market risk from the uncertain returns of the investment in repayment vehicles.

  • Maturity mismatch risk of assets and liabilities. Household assets (repayment vehicles) offer long-term returns, while liabilities (foreign currency loans) are exposed to short-term rates.

  • Double-exposure risk from borrowing in foreign currency and investing in real estate. The assets usually funded by foreign currency loans are real estate, a nontraded good, which tends to underperform when the terms of trade are depressed. With a correlation coefficient between Austria’s nominal effective exchange rate (NEER) and the housing price index for 1987–2003 of 0.72, a household could be hit adversely by the same shock (namely, depreciation of the domestic currency) on both the asset and liability sides of its balance sheet.

  • Litigation risk. Forced conversion clauses in loan contracts have their drawbacks: in the event of a depreciation and subsequent strengthening of the euro, the households would have locked-in their losses while missing the opportunity to get loans with the potential for capital gains. This could create frictions with the banks, undermine confidence, and lead to litigation.

Banks

Even though banks have limits on open foreign exchange positions, are largely hedged, and apply conservative loan-to-value ratios, they are exposed to indirect risk. Risks to the banks include:

  • Credit risk from the default of households, including from borrowers’ exposure to market risk.

  • Reputation and litigation risk from large losses suffered by clients and from the liquidation of their collateral.

  • Concentration risk from the homogeneity of collateral and the possible bank losses resulting from having to sell large amounts of real estate in small markets. Low loan-to-value ratios in Austria mitigate this risk but could still be considerable if many households default in the same market.

Government

There is a political risk that households might hold policymakers responsible for their having suffered large losses, arguing that they were not warned or protected adequately.

C. The Theoretical Model and the Data

The model

12. The literature on currency substitution provides the theoretical foundation of the estimated model.10 Under this theory, the choice of the currency mix of the assets of a household is a portfolio-balance decision guided by the relative returns of the assets, and subject to an intertemporal wealth constraint. Liabilities can be modeled in this framework by treating them as negative assets. With two types of liabilities (in domestic and foreign currency), the planned size of household liabilities in foreign currency, LP, is a positive function of the interest rate on domestic currency loans, Rd, and net wealth, W, and a negative function of the interest rate on foreign currency loans less the expected appreciation of the domestic currency, RfEe, all expressed in real terms:

Lp=f(R(+)d,Rf()Ee,w)(+)(1)

13. Subtracting the cost of the liabilities in foreign currency from the cost in domestic currency, equation (1) can be reformulated in terms of relative interest costs:

Lp=f(RdRf,Ee,W)(+)(+)(+)

14. Given that planned household liabilities in foreign currency are unobservable, actual liabilities at time t, Lt, and planned liabilities are related as follows:

Lt=ft(RdRf,Ee,W)+ϵt

where εt is the error term at time t.

15. Furthermore, net wealth can be broken down to human wealth—proxied by some function of contemporaneous labor income or consumption, CONSt—and net housing wealth (the collateral), proxied by an index of residential house prices, HOUSE_Pt.11 Making an obvious switch in notation, the following semi-log-linear long-run equation for loans in foreign currency to households can be derived:12

In (LOANSt)=β0+β1(INT(+)Dt)+β2In(EXC(+)Rt)+β3In(CONSt(+))+β4In(HOUSE(+)Pt)+θt

Data issues

16. The following variable definitions were employed. A short-term interest rate (12-month money market rate) was used to capture the cost of credit because foreign currency loans are predominantly adjustable rate.13 In line with a common practice in the literature, Austria’s NEER was used as a proxy for exchange rate expectations. Foreign currency loans rely on expectations about exchange rates that involve long horizons. Longer-term expectations of the exchange rate tend to be mean reverting, that is, agents expect rate movements to be reversed, which implies that the present level of the exchange rate carries all the information.14 Private consumption rather than personal disposable income was used because it is more closely correlated with borrowing, owing to consumption smoothing and mortgage equity withdrawal.

Interpretation of the data

17. The evolution of foreign currency loans since 1995 does not track economic fundamentals. Successful modeling begins with a good understanding of the data (Figure 3 shows the variables in the model). Loans in foreign currency exhibit a strong upward trend, which is matched only partly by the trend in consumption. Until about 1995, loans and their main determinants were positively correlated, in line with the theoretical model. However, after 1995, foreign currency loans grew at an explosive pace (which is much more visible in the original units of the variable shown in Figure 1) despite broad declines in the main determinants, except for consumption. Evidently, the series underwent a structural shift around 1995 that was not caused by the economic variables in the model; moreover, these variables cannot fully account for the developments in foreign currency loans since then. These observations are confirmed by the estimation results, which found that loans and their fundamental determinants cointegrate (i.e., they form a stable long-run relationship) in the period 1987–95 but do not cointegrate thereafter (see the next section).

Figure 3.
Figure 3.

Austria: The Data, 1987-2004

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Sources: OeNB; WiFO; IMF, IFS: Information Notice System; and IMF staff calculations.1/ A constant was added to the natural logarithm to facilitate graphing.2/ Euro rate (Austrian schilling up to 1998) less a weighted average of the rates for the Swiss franc, the Japanese yen, and the U.S. dollar. Weights are equal to the share of these currencies in total foreign currency loans to households in each quarter.

18. A possible explanation for the structural break is the emergence of herd behavior in the mid-1990s. Indeed, the switch to Swiss franc—but not euro—loans following the sharp movements of the yen over the past six years (see Figure 1), and the associated capital losses incurred by a number of borrowers suggest that entrenched herd behavior might have been at play (Box 4). While it is unclear what exactly triggered the herd behavior, the factors mentioned in paragraph 9 might have played a role. Consequently, a variable proxying herd behavior was added to the model to help explain developments since 1995.

Herd Behavior

Herd behavior occurs when people do what others do rather than rely on their own (incomplete) information, which might be suggesting something different (Banerjee, 1992). The suppression of private information could lead to “information cascades” when decisions are made sequentially and a large enough number of people choose identical actions. In such settings, the decisions of a critical few people early on are enough to tilt group behavior toward a certain direction. Mimicking the behavior of others might be rational because of uncertainty about one’s own information as well as the need to economize on information-gathering costs. Rational herd behavior is the subject of a recent strand of behavioral finance (see Montier, 2002, for an introduction).

Herd behavior can arise in a variety of environments, including in financial markets. However, it is difficult to disentangle empirically the effects of macroeconomic or other fundamental determinants from those caused by herd behavior. Herd behavior often results in volatility because it is susceptible to abrupt shifts or reversals, and thus has the potential to destabilize markets.

Empirical studies have shown that the dynamics of herd behavior often resemble an S curve: initially only a few adopt a certain behavior, but, past a certain critical mass, a take-off state takes hold where a rapidly growing number of people adopt this behavior. Toward the end of this process, a moderation of the dynamics takes place as the potential pool of adoptees is exhausted. These dynamics are often modeled using a logistic growth model, because of the ability of the logistic function to capture the dynamics of herd behavior: see Tsoularis (2001) for a discussion of logistic growth models in the context of several applications; and Yoshifuji and Demizu (1998), Levi-Faur (2002), and Maggioni (2004) for formal models of herd behavior in various markets, using variations of the logistic function.

19. Herd behavior was modeled as a logistic function. The logistic function 1/(1+α/eβx) was fitted with the share of foreign currency loans in total loans to households (argument x). It was parameterized so that it exhibits the following properties: the function takes off at just under x = 5 percent (roughly the share around 1995 when these loans took off—see Figure 1) and slows down around x = 30 percent (just above the present share, which has been accompanied by moderating growth).15 Moreover, the function was scaled up by a constant (one) so that its natural logarithm takes only positive values. The bottom panels of Figure 3 show the resulting proxy variable for herd behavior and the underlying logistic function.

20. The logistic function employed here has a number of appealing properties that track closely the evolution of foreign currency loans in Austria. First, it is an exponentially increasing function of the share of foreign currency loans in total loans: the larger the number of households who have a foreign currency loan, the faster the spread of the word of mouth. Second, it requires a critical mass to take off: unlike other euro zone countries, Austria was near this critical mass because of the popularity of foreign currency loans in the western part of the country, and a small perturbation was sufficient to set the herd process in motion. Third, it accelerates as the herd behavior gathers momentum, which describes loan developments since 1995. Finally, its pace of expansion eventually slows, as most of those who wanted to jump on the bandwagon have already done so, as currently appears to be the case (see Figure 1).

21. Tests for the nonstationarity of the variables in the model suggest that loans in foreign currency and the variable for herd behavior could be I(2) processes, that is, double differencing is needed to make the series stationary (see the Appendix). This requires that the search for a stationary linear combination of all variables (called the cointegrating vector) be done using Johansen’s I(2) cointegration analysis.16

22. Cointegration analysis of I(2) processes is complicated, and the long-run equilibrium may involve not only the levels, but also the differences of the I(2) variables.17 This involves: (i) searching for a linear combination of the two I(2) variables (namely, loans in foreign currency and the proxy for herd behavior) that is an I(1) process; and (ii) using this combination to transform the two variables into a single one, and focusing on the I(1) system. This has the added appeal that, if foreign currency loans and the variable for herd behavior cointegrate (and thus can be transformed to a single variable), then the influence of true “economic” factors can be separated from that of herd behavior in the demand for loans in foreign currency. This study follows this modeling approach, and the following section presents the model for the transformed variable. Inference from the transformed system involves no loss of information, provided that the two I(2) variables share one common I(2) trend that enters the cointegrating relation with the coefficients found in the unrestricted model.

D. Estimation of the Long-Run Demand for Loans in Foreign Currency

23. The search for a stable and meaningful long-run relationship between foreign currency loans and their determinants was unsuccessful for the full sample(1987:Q1–2004:Q3). Because of the structural break around 1995, the estimation could not establish a long-run cointegrating relationship (interpreted as steady state equilibrium) between real loans to households in foreign currency and the following: (i) their fundamental determinants (the nominal interest rate differential, the nominal exchange rate, real consumption, and real house prices); (ii) the proxy for herd behavior; and (iii) a dummy variable capturing the effects of the policy measures since mid-2003 mentioned in paragraph 11. The search was unsuccessful even after house prices and the dummy were dropped from the model.

24. A likely explanation is that the full sample captures two distinct behaviors: one describing the pre-1995 and another the post-1995 period. To test this hypothesis, distinct models were estimated for each period. The herd variable and the dummy were added only in the latter period. This modeling approach allowed a long-run cointegrating relationship to be uncovered in each subsample. Box 5 presents the results for the pre-1995 period.18

Estimation Results for 1987–95

The following long-run relationship was estimated for 1987:Q1–1995:Q4 (t-ratios in parentheses):

loanst=0.0094(2.49)*INTDt+19.001(6.07)*excrt+1.074(0.68)*const
article image

The model is robust to alternative specifications and small changes in the sample size. The statistical evidence suggests the following for the period 1987–95:

  • By far, the principal determinant of the long-run demand for foreign currency loans was the exchange rate; the interest rate differential played a secondary role. This evidence is in line with the fact that the demand for foreign currency loans during this period was mainly for foreign transactions.

  • House prices did not play a role in the spread of these loans.

  • The return of foreign currency loans to equilibrium from a disequilibrium state is reasonably quick: the feedback coefficient for foreign currency loans (–0.124) suggests that it takes roughly eight quarters for these loans to fully respond to a shock and reach their new equilibrium.

  • Excess demand for foreign currency loans did not have an effect on the exchange rate—an external variable—and on the interest rate differential (both variables have insignificant feedback coefficients). Interpreting the interest rate differential as the reaction function of banks, it appears that banks did not use these loans as an instrument of competition during this period.

25. For the period since the mid-1990s (the period on which this study will focus), the results are (t-ratios in parentheses):19

Long-Run Equilibrium, 1994:Q1–2004:Q31

(loanst3.244*herdt)=1.097(7.29)*INTDt+7.106(3.96)*excrt+7.907(3.66)*const1.198(0.55)*Δloanst(2)
article image

The start of the sample was set to 1994 rather than 1995 to add degrees of freedom and increase the robustness of the results. Only the selected equation is reported.

26. Equation (2) has been rearranged so that the left-hand side represents the part of the demand for foreign currency loans that is linked entirely to fundamentals. It has all the features of a demand relation. The variable Δloanst (the first difference of loanst) is the common I(2) trend shared by loans t and herd t, and is included in the long-run relationship to eliminate any remaining I(2) trends in the system, a standard practice in I(2) analysis. The intuition behind the inclusion of Δloanst is that the change in the level of foreign currency loans is directly related to herd behavior and thus to deviations of these loans from the optimal holdings determined by fundamentals.

27. Figure 4 (top panel) presents a decomposition of the long-run equilibrium relationship, revealing the relative contributions of the determinants of loans to households in foreign currency. The statistical evidence for the period 1994–2004 suggests the following:

Figure 4.
Figure 4.

Austria: Long-Run Equilibrium, and Actual and Fitted Values of Loans in Foreign Currency, 1994-2004

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Sources: OeNB; WiFO; IMF, IFS: Information Notice System; and IMF staff calculations.1/ Variables have been scaled so that they have a zero mean over the estimation period.2/ Excluding the impact of the variable proxying herd behavior.
  • A stable long-run relationship between foreign currency loans in Austria and their fundamental determinants emerges only when the herd behavior variable is added to the model. However, the model is not very robust to alternative specifications and small changes in the sample size. The relationship has the signs prescribed by theory. The negative coefficient of Δloans t is intuitive: a positive change in these loans contributes to the strengthening of herd psychology and thus to the departure of the demand for these loans from fundamentals.

  • The estimation could not establish a link between foreign currency loans and more recent house price developments.20

  • Recent policy measures do not appear to have influenced the long-run demand for foreign currency loans. The dummy variable proxying these measures either did not cointegrate with the other variables or was not statistically significant.

  • Loans in foreign currency appear to be highly sensitive over the long run to changes in the exchange rate and consumption, with long-run elasticities over 7. The large coefficients for these two variables most likely capture financial liberalization and the ongoing internationalization of the Austrian economy, as well as deviations from fundamentals not fully captured by the variable for herd behavior.21

  • The large, positive, and highly significant feedback coefficient for INT_Dt, which captures the response of the interest rate differential to excess demand for foreign currency loans, lends support to the hypothesis that banks have competed for market share since the mid-1990s by lowering interest rates on these loans, while probably recouping the lost revenue through higher fees on these loans.22 This positive response feeds back into the demand for loans, slowing its return to equilibrium.

  • Foreign currency loans adjust to equilibrium from a disequilibrium state very slowly (the feedback coefficient of the transformed variable for loans is -0.015). This is consistent with households’ reluctance to switch currencies in response to a shock, partly because of the transaction fees charged by banks.

  • Much of the strength of loans in foreign currency in recent years seems to be due to herd behavior, with the perceived lower cost of these loans (interest rate differential adjusted for exchange rate movements) still playing an important, though diminishing, role (see Figure 4, top panel). The influence of herd behavior appears to have increased in 2004, largely offsetting the effect from the decline in the interest rate differential and accounting for about half the current level of these loans.

28. The next step in the estimation of the model is to combine the long-run relationship with short-run influences in a single dynamic equation of foreign currency loans.

E. A Dynamic Model of Loans in Foreign Currency

Error-correction model

29. Deviations of foreign currency loans from their long-run equilibrium levels do not always reflect disequilibria; they also reflect temporary factors and adjustment lags. To capture the short-run dynamics, foreign currency loans were modeled in an error-correction representation, using—in addition to the variables mentioned above—a dummy variable attempting to capture the effect of recent policy measures on the growth rate of these loans.

30. Starting from a general specification that includes several lags and variables and reducing it to a parsimonious representation with the aid of appropriate tests, the estimation yielded the following error-correction model of foreign currency loans for 1994:Q1–2004:Q3:23

Δ(loans3.2*herd)t=2.044(3.33)+0.916(14.07)*Δ(loans3.2*herd)t1+0.061(2.92)*ΔINTDt*ECM(loans3.2*herd)t1(3)
RBAR2=0.839σ=0.040

31. All estimated coefficients have the anticipated signs, and the equation passes a series of diagnostic tests. Tests for parameter stability and for the presence of structural breaks confirmed that the results are robust (Figures 5 and 6).24 Furthermore, the estimated model fits the data well, suggesting that the recent strength of foreign currency loans can be explained adequately by the variables included in the analysis (see lower panel of Figure 4).

Figure 5.
Figure 5.

Austria: Parameter Stability, 1996-2004

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Source: IMF staff calculations.
Figure 6.
Figure 6.

Austria: Tests for Structural Breaks, 1996-2004 1/

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Source: IMF staff calculations.1/ Test values in panels two through four are scaled by one-off critical values from the F -distribution at the 1 percent probability level as adjustment for changing degrees of freedom. Test values below 1.0 cannot reject the null hypothesis.2/ Points outside the two standard-error region are either outliers or associated with coefficient changes.3/ One-step forecast test of the null hypothesis of constant parameters.4/ Chow tests of the null hypothesis of no difference between the N -periods-ahead and one-period ahead forecasts.5/ Chow tests of the null hypothesis of no difference between the one-period-ahead and N -periods-ahead forecasts.

32. The quarter-on-quarter growth rate of foreign currency loans to households is highly autoregressive (as evidenced by the large short-run coefficient of Δ(loans - 3.2*herd)t-1 in (3), signaling considerable momentum in the growth rates of these loans and providing evidence of the influence of herd behavior, even in the short run. Among economic variables, only the interest rate differential appears to influence short-term fluctuations of the demand for loans in foreign currency. However, rolling regression estimates suggest that the relation between the growth of foreign currency loans and changes in the interest rate differential has weakened over the past several years (see bottom-left panel of Figure 5), suggesting that cost savings are becoming less of a driving force behind the growth of these loans.

Dynamic simulations

33. How much did the strength of loan growth in 2004 depend on the contribution of herd behavior? A dynamic simulation of the error-correction model in (3)—which involved holding herd behavior constant at its end-2003 level during 2004—indicates that the year-on-year growth of foreign currency loans would have slowed to around 7 percent (Figure 7). This simulation confirms the result from decomposing the cointegrating relationship that much of the recent strength of these loans derives from the steady influence of herd behavior (see Figure 4). It also provides a lower bound estimate of the impact of the policy measures outlined in paragraph 11 on the prospective developments in foreign currency loans during 2004, should the measures had been successful in mitigating herd behavior.

Figure 7.
Figure 7.

Austria: Real Loans to Households in Foreign Currency, 2003-04

Citation: IMF Staff Country Reports 2005, 249; 10.5089/9781451802375.002.A001

Sources: OeNB; and IMF staff estimates.1/ Assumes unchanged herd behavior in 2004 from the end-2003 level. All other variables assume their actual values.

34. Dynamic simulations were also performed to test the response of foreign currency loans to large, counterfactual shocks to their economic determinants. Each determinant was subjected to a structured shock, and the depth and duration of the impact on foreign currency loans were observed.25 The experiments suggest that foreign currency loans have become quite resilient to relatively large adverse shocks to their economic determinants, though the extent and duration of the impact on loan growth do vary among the controlled variables.

35. The counterfactual simulations should be interpreted with caution because they are based on partial analysis. With given paths for the explanatory variables, the simulations preclude feedbacks between variables. For instance, it is likely that a decline in the exchange rate of a larger magnitude than that assumed in the experiment would be accompanied by other developments (such as erosion of confidence) that could have a larger impact on foreign currency loans (and through those to economic activity variables, such as consumption) than the one suggested by the experiments.

F. Summary of Results

36. The main results of the model can be summarized as follows:

  • The demand for foreign currency loans in Austria began to exhibit a departure from fundamentals around 1995. A variable capturing herd behavior appears to be successful in isolating the effects of the fundamental economic variables in household demand for foreign currency loans. These loans currently appear to be significantly above the steady state equilibrium level implied by economic fundamentals, but the return to equilibrium is likely to be slow.

    • The exchange rate plays an important role in the long run but does not appear to have a statistically significant influence on foreign currency loans in the short run. This finding is consistent with the observed lag in switching to the Swiss franc following the protracted rise of the Japanese yen. It suggests that exchange rate movements need to be persistent in order to be taken into account and overcome the psychology of a stable exchange rate.

    • The interest rate differential seems to play an important—albeit diminishing—role in explaining the short-run dynamics of foreign currency loans: the estimated coefficient in the error-correction model implies that, if the differential were to rise by 10 basis points, the immediate impact on quarter-on-quarter loan growth would be an increase of roughly 0.6 percentage point.26

  • The principal factor that distinguishes Austria from other euro zone countries seems to be the formation of herd behavior in the demand for foreign currency loans. The roots of this behavior can be traced to the western part of the country. But unlike other euro zone countries that may have seen pockets of household borrowing in foreign currency in some regions, Austria was near the critical mass required for the onset of herd behavior because of the large popularity of these loans in the western part of the country even since the 1980s. It is likely that some Austria-specific factors, including structural changes in the Austrian banking system, helped trigger the herd behavior.

  • Supply factors (notably banks’ pricing policies) appear to have played a role:

    • The reaction of banks to excess demand for foreign currency loans appears to have changed since the mid-1990s. The feedback coefficient for the interest rate differential is statistically equal to zero in the pre-1995 period but large, positive, and highly significant thereafter. This suggests that banks have responded to the growing demand for these loans by lowering interest rates to maintain or gain market share.

    • Foreign currency loans adjust very slowly to deviations from their long-run equilibrium, pointing once again to a possible reluctance of households to switch currencies in the face of shocks owing to transaction fees.

  • It appears that housing market developments have not played a significant role.

  • The response of foreign currency loans to modest shocks (stemming, for example, from a depreciation of the exchange rate) should be relatively modest and gradual. Nonetheless, major and protracted declines could have a substantial impact, especially when accompanied by erosion of confidence, and spill over to the rest of the economy. These loans are more responsive to changes in herd behavior.

  • Policy measures and the public information campaign have yet to cause a decline in either the level or the growth rate of these loans. Dummy variables in both the long-run relationship and the short-run error-correction equation were statistically insignificant. The observed decline in 2004 in the transformed demand for loans (after adjusting for the influence of herd behavior) is fully accounted for by economic determinants (see top panel of Figure 4).

G. Concluding Remarks

37. The rapid growth of foreign currency loans to Austrian households has raised questions about the forces driving this increase and heightened concerns about risks. This study aimed to uncover both the short- and long-run influences on foreign currency loans and to assess their prospects. The results are generally robust, the estimated model has a dynamic structure and appealing economic interpretation, and performs well in explaining the data.

38. Certain themes emerged consistently throughout various aspects of the empirical investigation. The rapid spread of the popularity of foreign currency loans among households seems to reflect to a significant extent herd behavior. Among economic determinants, although the exchange rate appears to have a larger influence than the interest rate differential in the long run, it could not explain the short-term dynamics. Pricing policies of banks appear to have played a central role. Policy measures do not appear to have been successful in lowering either the level or the growth rate of these loans.

39. The results raise the question of the role of policy in containing the spread of these loans. Establishment of a strong supervisory framework in the financial sector and high prudential standards can help guard against risks. Indeed, the proactive stance of supervisors has improved the risk management of banks and helped reduce risks by facilitating the switch to the less volatile Swiss franc. However, the broader effort to address the issue of foreign currency loans has yet to have a tangible impact on the demand for these loans. This does not mean that policy can be ineffective. Herd behavior can be tempered when new information points clearly to the need to change behavior, including through a forceful public information campaign to raise consumer awareness about the risks involved.

40. Additional research on this issue may help improve our understanding. First, a better way to model herd behavior might increase the stability of the results and improve inference. Second, the fact that house prices were not found to be related to foreign currency loans in the sample does not necessarily mean that they are not part of the broader issue. The subsidization of housing construction in Austria may have weakened the signal house prices carry, thereby making it difficult to integrate them into the model. Nevertheless, successful integration of the housing market into a model of foreign currency loans to Austrian households would help us understand an important propagation channel into the rest of the economy. Finally, in light of the evidence on the banks’ role in this issue, modeling the pricing (and fee-setting) behavior in a reaction function of banks may help shed more light on the role of the supply of these loans in recent developments.

APPENDIX A Model of Foreign Currency Loans to Austrian Households

41. This appendix describes a model of foreign currency loans to households in Austria based on quarterly data for 1987:Q1–2004:Q3.

A. Data

42. The following series were used in this study (notation in parentheses):

Foreign currency loans (LOANS): Quarterly average stock of foreign currency loans to households by Austrian banks, divided by the implicit GDP deflator (1995=100; seasonally adjusted).

Interest differential (INT_D): 12-month LIBOR rate (in percent) for the euro (Austrian schilling up to 1998) less a weighted average of the 12-month LIBOR rates for the Swiss franc, the Japanese yen, and the U.S. dollar. Weights are equal to the share of these currencies in total foreign currency loans to households in each quarter. For the period before 1997, the 1997 weights were used.

Exchange rate (EXC_R): For 1987:Q1–1997:Q3, the nominal effective exchange rate for Austria (currencies of all major trade partners, 1999=100) with trade weights from the IMF’s Information Notice System. For 1997:Q4–2003:Q4, the nominal effective exchange rate for Austria (Swiss franc, the Japanese yen, and the U.S. dollar), with weights set equal to the share of these currencies in total foreign currency loans to households in each quarter.

Consumption (CONS): Total final consumption expenditure of households and nonprofit institutions serving households in 1995 prices; seasonally adjusted.

Herd behavior (HERD): The logistic function 1/(1+α/eβ x) was fitted with x = the share of foreign currency loans in total loans to households by Austrian banks (quarterly averages). It was parameterized α = 1000, β = 35 so that it exhibits the following properties: the function takes off at just under x = 5 percent (the share of foreign currency loans around 1995) and slows down around x = 30 percent (just above the present share). Moreover, the function was scaled up by a constant (one) so that its natural logarithm takes only positive values.

House prices (HOUSE_P): For 1987:Q1–2001:Q2, residential property price index compiled by the Technical University (Vienna). For 2001:Q3–2004:Q3, residential property price index as reported by the OeNB. Both series are available only with semiannual frequency. Consequently, they were interpolated (1990=100) and deflated by the consumer price index (1996=100).

Dummy variable for policy measures (DUMMY): D = 0 for 1987:Q1–2003:Q2; D = 1 for 2003:Q3–2004:Q3.

B. Integration

43. To determine the appropriate estimation procedure, augmented Dickey-Fuller (ADF) tests for nonstationarity of the above variables were carried out, which suggested that INT_D, exc_r, and cons might be integrated of order one, and loans and herd (and possibly house_p) might be integrated of order two (Table 1). ADF tests (not shown) of the null hypotheses of integration of third order for each variable rejected the null. All variables are thus stationary in first or second differences, and Johansen’s I(2) cointegration analysis among the level variables (and possibly the changes of the I(2) variables) is required.27

Table 1.

ADF Statistics Testing for a Unit Root

article image
Note: Critical values are: -2.91 (5 percent level, *), and -3.54 (1 percent level, **). Lag length was determined by the choice of the lag with the highest AIC (Akaike information criterion) statistic in absolute value.

C. Cointegration

Estimation for 1987–95

44. Using data for 1987:Q1–1995:Q4, the Johansen procedure found evidence of the following long-run relationship between foreign currency loans to households in Austria and their fundamental determinants:28

loanst=0.088(2.40)*INTDt+18.599(6.10)*excrt+1.234(0.81)*const(A.I)

45. All estimated coefficients of the selected vector have the anticipated signs. However, the coefficient of consumption is insignificant at the 1 percent confidence level. Nevertheless, consumption was retained in the long-run relationship because its feedback coefficient was statistically significant (see Table 2 below). This argues in favor of retaining const in the model because of a one-way interaction between the levels of foreign currency loans and consumption during 1987–95: from loans to consumption, but not the other way. House prices did not cointegrate with the remaining variables and thus do not belong in the long-run cointegrating relationship for the period up to 1995. Table 2 reports the tests for the presence of cointegrating vectors in the Johansen procedure, as well as the feedback coefficients and their standard errors corresponding to the cointegrating vector in (A.1).

Table 2.

Johansen Test of Existence of Long-Run Relationships, and Feedback Coefficients, 1987:Q1-1995:Q41

article image

Standard t-ratios are reported in parentheses. Feedback coefficients represent the response (adjustment) of the respective variable to a deviation of foreign currency loans from the long-run equilibrium estimated in (A.1).

A single asterisk denotes a significant Trace statistic at the 5 percent level; two asterisks denote significance at the 1 percent level. Based on the Trace statistic, the Johansen procedure uncovered a single cointegrating vector, the one in (A.1).

46. The feedback coefficient is statistically different from zero in the loans and consumption equations. This means that the single cointegrating vector (A.1) enters both equations in the vector autoregression, and that the adjustments to deviations from equilibrium (excess demand for foreign currency loans) tend to bring these loans back to equilibrium (negative feedback coefficient of loans) even though they raise consumption (positive feedback coefficient).

47. The estimated relationship in (A.1) was used to test the hypotheses that both the feedback coefficient for exc_r and the one for INT_D are equal to zero. The test of the hypothesis yielded the following statistic (p value in square brackets):

x2 (1) = 0.028 [0.986]

Based on the large (well in excess of 0.05) p value of the statistic, the null hypothesis cannot be rejected, and the restricted long-run relationship becomes:

loanst=0.094(2.49)*INTDt+19.001(6.07)*excrt+1.074(0.68)*const(A.2)

The restricted equation in A.2 was reported in Box 5, together with the unrestricted feedback coefficients for the exchange rate and interest rate differential.

48. The result of the test for weak exogeneity (i.e., that the feedback coefficients of INT_D, exc_r, and cons in Table 2 are jointly equal to zero) is the following:

LR test of restrictions: x2 (4) = 1.10 [0.011]*

The null hypothesis of weak exogeneity is barely rejected at the 1 percent level. Therefore, at the 5 percent level, a more parsimonious single-equation estimation could be used without any loss of information about the cointegrating vector. Proper I(2) analysis relying on the eigenvalues of the long-run matrix showed no evidence of any I(2) trends in the system.

1994—2004 estimation

49. Using data for 1994:Q1–2004:Q3, the Johansen procedure found evidence of the following long-run relationship between foreign currency loans to households in Austria and their fundamental determinants:29

(loanst3.244*herdt)=1.097(7.29)*INTDt+7.106(3.96)*excrt+7.907(3.66)*const1.198(0.55)*constΔloanst(A.3)
article image

50. The estimation established only one cointegrating vector, and its estimated coefficients have the anticipated signs. The coefficient of Δloanst is insignificant at the 1 percent confidence level. Nevertheless, Δloanst was retained in the long-run relationship because it helped remove any remaining I(2) trends in the system. Table 3 reports the tests for the presence of cointegrating vectors in the Johansen procedure, as well as the feedback coefficients and their standard errors corresponding to the cointegrating vector in (A.3).

Table 3.

Johansen Test of Existence of Long-Run Relationships, and Feedback Coefficients, 1994:Q1-2004:Q31

article image

Standard t-ratios are reported in parentheses. Feedback coefficients represent the response (adjustment) of the respective variable to a deviation of the “transformed” loans from the long-run equilibrium estimated in (A.3).

A single asterisk denotes a significant Trace statistic at the 5 percent level; two asterisks denote significance at the 1 percent level. Based on the Trace statistic, the Johansen procedure uncovered a single cointegrating vector, the one in (A.3).

51. The feedback coefficient is statistically different from zero in the loans and interest rate differential equations. This implies that the single cointegrating vector (A.3) enters both equations in the vector autoregression, and that the adjustments to deviations from equilibrium tend to bring foreign currency loans back to equilibrium (negative feedback coefficient). The large positive feedback coefficient for the interest rate differential equation has an appealing economic interpretation. It can be seen as evidence of competition by banks: banks respond to excess demand for foreign currency loans by lowering rates to maintain or gain market share.30 Growing competition among lenders has forced them to make these loans attractive. The feedback coefficient for exc_r (a foreign variable) is not different from zero, in line with the a priori expectation that exchange rates are determined outside this system.

52. The results of the test for weak exogeneity (i.e., that the feedback coefficients of INT_D, exc_r, cons, and Δloans in Table 3 are jointly equal to zero) are the following:

LR test of restrictions: χ2 (4) = 31.304 [0.000]**

The hypothesis is strongly rejected. Therefore, a more parsimonious single-equation estimation could be used without any loss of information about the cointegrating vector only if the estimates of the vector of coefficients from equation (A.3) are used in the single-equation estimation.

53. The transformed theoretical model shows no evidence of a remaining I(2) trend, confirmed by the eigenvalues of the long-run matrix. Foreign currency loans and the herd variable (both I(2) processes sharing a common I(2) trend) cointegrate to I(1), which in turn cointegrates to I(0) with the other variables in the model. The results confirm that Δloanst is the common I(2) trend, and its inclusion in the model removes any remaining I(2) components. Since there is no theoretical requirement that the common trend is embedded in equal proportions, the estimation in the unrestricted model found that the common trend was embodied in foreign currency loans and the herd variable with coefficients (1,–3.244).

D. Single-Equation Modeling (Error Correction)

54. This section introduces the long-run dynamics established in (A.3) into a single-equation, conditional, and parsimonious model for loans to households in foreign currency in Austria that encompasses both short- and long-run dynamics. The short-run dynamics are derived from the following vector of changes in the log-levels of the variables (levels for INT_D) used so far in the analysis:

Is={Δvtj};j = 0,1…,

where vt-j is the vector of all variables (contemporaneous and lagged) used so far, with the addition of two new stationary variables, orthogonal to the existing repressors. These are:

Δhouse_p = Quarter-on-quarter change of house prices, and

DUMMY = Dummy variable for policy measures: D = 0 for 1994:Q1–2003:Q2; and D = 1 for 2003:Q3–2004:Q3.

55. An error-correction model for loans to Austrian households in foreign currency was estimated using quarterly data over 1994:Q1–2004:Q3. Estimation began from a general specification with several variables and lags, which also included the vector Is and the deviations of foreign currency loans from the long-run relationship (A.3) established above. Subsequently, parameter tests were performed (mostly zero restrictions) to reduce the model to a more manageable form. Below is the resulting specification for loans to households in foreign currency and the relevant statistics (t-ratios are shown in parentheses below the estimated coefficients) and diagnostic tests (p values are shown in brackets next to the estimated statistics):

Δ(loans3.2*herd)t=2.044(3.33)+0.916(14.07)*Δ(loans3.2*herd)t1+0.061(2.92)*ΔINTDt0.032(3.34)*ECM(loans3.2*herd)t-1

R-BAR2 = 0.839, DW = 2.36, σ^ = 0.040

article image

56. The estimated equation fits the data well (see lower panel of Figure 4), and all coefficients appear with the anticipated signs. Moreover, the model passes a battery of diagnostic tests. The hypotheses: of no serial autocorrelation up to fourth order; normality of the residuals; and homoskedasticity could not be rejected. Finally, the coefficient of determination of 0.84 is high, considering that variables are seasonally adjusted and expressed in changes.

57. In addition, the model passes a series of diagnostic tests for misspecification and structural breaks. First, tests for parameter constancy showed that the estimated coefficients are generally robust in recent years (though there is some instability around 2000) and that their standard errors diminish over time, suggesting correct specification of the estimated model (see Figure 5). The tests suggest that the sensitivity of foreign currency loans to the interest rate differential has declined in recent years, though the hypothesis of a decline barely passes a 5 percent significance test. Second, a series of diagnostic tests confirmed the absence of structural breaks (see Figure 6). The upper-left panel of Figure 6 shows that the errors of one-period-ahead forecasts are white noise with constant variance (except for two outliers). The remaining panels show that, at 1 percent level of confidence, the model does not exhibit any structural breaks at any subset of the period under study, estimated in either ascending or descending chronological order.

58. Tests for the exclusion of relevant variables showed that neither house prices nor the dummy proxying the recent policy measures belong in the dynamic equation for a reasonable number of lags and for the period under study.

References

  • Alevy, J. E., M. S. Haigh, and J. A. List, 2003, “Information Cascades: Evidence from a Field Experiment with Financial Market Professionals,Department of Agricultural and Resource, Economics Working Paper 03-15 (College Park: The University of Maryland).

    • Search Google Scholar
    • Export Citation
  • Ball, M., 2004, “Austria,RICS European Housing Market Review 2004, Chapter 4, pp. 17 -27 (London: RICS Europe).

  • Banerjee, A. V., 1992, “A Simple Model of Herd Behavior,The Quarterly Journal of Economics, Vol. CVII(3), pp. 797 -817.

  • Braumann, B., 2004, “Tu Felix Austria: Evidence for a De-celerator in Financial Reform,International Economics and Economic Policy, Vol. 1(1), pp. 53 -72.

    • Search Google Scholar
    • Export Citation
  • Drehmann, M., J. Oechssler, and A. Roider, 2003, “Herding and Contrarian Behavior in Financial Markets—An Internet Experiment,” Discussion Paper (London: Bank of England).

    • Search Google Scholar
    • Export Citation
  • Fiess, N., and R. MacDonald, 2001, “The Instability of the Money Demand Function: an I(2) Interpretation,Oxford Bulletin of Economics and Statistics, Vol. 63(4), pp. 475 -95.

    • Search Google Scholar
    • Export Citation
  • Giovannini, A., and B. Turtelboom, 1992, “Currency Substitution,NBER Working Paper No. 4232 (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Gnan, E., 1994, “Austria’s Hard Currency Policy and European Monetary Integration,De Pecunia, Vol. 6(3), pp. 27 -72.

  • Hendry, D. F., and J. A. Doornik, 1999, Empirical Econometric Modeling Using PcGive, (West Wickham: Timberland Consultants Ltd.).

  • Hofmann, B., 2001, “The Determinants of Private Sector Credit in Industrialized Countries: Do Property Prices Matter?BIS Working Paper 108 (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation
  • IMF, 2004, “Austria: Financial System Stability Assessment,IMF Country Report No. 04/238 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Johansen, S., 1988, “Statistical Analysis of Cointegration Vectors,Journal of Economic Dynamics and Control, Vol. 12, pp. 231254.

    • Search Google Scholar
    • Export Citation
  • Johansen, S., 1992, “An I(2) Cointegration Analysis of the Purchasing Power Parity Between Australia and the United States,Article 9 in C. P. Hargreaves (editor) Macroeconomic Modelling of the Long Run, p.p. 229 -48 (Aldershot: Edward Elgar Publishing Limited).

    • Search Google Scholar
    • Export Citation
  • Johansen, S., 1995, Likelihood-based Inference in Cointegrated Vector Autoregressive Models, (Oxford: Oxford University Press).

  • Kaufmann, S., and M. T. Valderrama, 2004, “Modeling Credit Aggregates,OeNB Working Paper 90 (Vienna: Oesterreichische Nationalbank)

    • Search Google Scholar
    • Export Citation
  • Kingston, G., 1995, “The Foreign Currency Loans Affair: An Economist’s Perspective,School of Economics Discussion Paper 95/10 (Sydney: The University of New South Wales).

    • Search Google Scholar
    • Export Citation
  • Komárek, L., and M. Melecký M., 2003, “Currency Substitution in a Transitional Economy with an Application to the Czech Republic,Eastern European Economics, Vol. 41 (4), pp. 7299.

    • Search Google Scholar
    • Export Citation
  • Levi-Faur, D., 2002, “Herding Towards a New Convention: On Herds, Shepherds, and Lost Sheep in the Liberalization of the Telecommunications and Electricity Industries,Nuffield College Working Paper in Politics 6 (Oxford: University of Oxford).

    • Search Google Scholar
    • Export Citation
  • Luca, A., and I. Petrova, 2003, “Is Credit Dollarization Driven by Firms or Banks?unpublished (Lancing: Michigan State University).

    • Search Google Scholar
    • Export Citation
  • Maggioni, M. A., 2004, “The rise and Fall of Industrial clusters: Technology and the Life Cycle of Regions,Instut d’Economia de Barcelona Working Paper 2004/6 (Barcelona: University of Barcelona).

    • Search Google Scholar
    • Export Citation
  • Montier, J., 2002, Behavioural Finance: Insights into Irrational Minds and Markets(Chichester: John Wiley & Sons Ltd.)

  • OeNB, 2003, Financial Stability Report 5, pp. 26 -28 (Vienna: Oesterreichisch Nationalbank).

  • Tsoularis, A., 2001, “Analysis of Logistic Growth Models,Research Letters in Informatics and Mathematical Sciences, Vol. 2, pp. 2346.

    • Search Google Scholar
    • Export Citation
  • Tzanninis, D., 2002, “Why Has U.K. Household Consumption Been so Strong?in United Kingdom: Selected Issues, IMF Country Report No. 02/46 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Waschiczek, W., 2002, “Foreign Currency Loans in Austria—Efficiency and Risk Considerations,in Financial Stability Report 4, OeNB, pp. 8399 (Vienna: Oesterreichische Nationalbank).

    • Search Google Scholar
    • Export Citation
  • Yoshifuji, S., and H. Demizu, 1998, “A Study of the Mechanism of Stress/Shock Movements in Financial Markets,unpublished (Tokyo: The Bank of Tokyo—Mitsubishi, Ltd.).

    • Search Google Scholar
    • Export Citation
1

Prepared by Dimitri Tzanninis.

2

These loans are secured by real estate and finance predominantly house purchases.

3

Waschiczek (2002) provides a comprehensive discussion of foreign currency loans in Austria since 1995, their main features and the principal risks involved, and suggests—but does not model—the possibility of herd behavior.

4

See, for example, on the asset side, Komárek and Melecký (2003) for an application of the currency substitution theory to a transition economy, and, on the liability (i.e., credit) side, Luca and Petrova (2003) for a study of 22 transition economies.

5

Using a model of credit aggregates (in all currencies) for four European countries, including Austria, Kaufmann and Valderrama (2004) find credit in Austria to be mostly demand driven.

6

Gnan (1994) studies the hard currency policy in Austria since its adoption in 1980.

7

Ball (2004) reviews the structure of Austria’s house and mortgage markets and discusses recent developments.

8

Hofmann (2001) provides empirical evidence of a relationship between credit and property prices in 16 industrialized countries.

9

Kingston (1995) discusses the economic and legal issues related to the fall of the Australian dollar and the surge of foreign currency liabilities in Australia in the 1980s.

10

See the survey of literature provided in Giovannini and Turtelboom (1992).

11

With housing being the principal asset of a household, other net wealth is not modeled.

12

All variables are expressed in natural logarithms (except for the interest rate differential) and in real terms (except for the interest rate differential and the exchange rate). See the Appendix for a discussion of the series used, the statistical tests, and the estimation results.

13

According to Kaufmann and Valderrama (2004), lending rates in Austria track the short-term rates more closely than long-term rates.

14

This study follows Komárek and Melecký (2003) and Giovannini and Turtelboom (1992). These authors show how the real interest rate differential is equivalent to the nominal one plus the expected change in the nominal exchange rate, and apply the logarithmic transformation separately to the interest rate differential and to the exchange rate series.

15

Parameterization of the logistic function based on the properties of the data is not uncommon in the literature of herd behavior (see, for example, Alevy et al, 2003, and Drehmann et al, 2003). Nevertheless, the results of this study were largely invariant to different parameterizations.

16

See Johansen (1992) for an application of the I(2) analysis to Australian and U.S. data.

17

Fiess and MacDonald (2001) explain the steps involved in estimating I(2) models with an application to Danish data.

18

Lowercase letters denote logarithms of the variables introduced above, and their estimated coefficients can be interpreted as long-run elasticities, with the exception of the interest rate differential, whose coefficient is a semielasticity.

19

The estimation involved two steps, in line with the procedure described in paragraph 22. First, the unrestricted system was used to derive an estimate of the coefficient of the herd variable. Second, using this coefficient, a new, transformed variable (loanst–3.244*herd t) was created and the system reestimated.

20

The absence of an empirical link—though consistent with currency substitution rather than higher demand for mortgages and houses—does not allow the use of the model for the study of counterfactual scenarios involving shocks to house prices and their impact on consumption.

21

The high elasticities—especially that for consumption—are not very appealing from an economic perspective, but the other features of the estimated model offer useful insights.

22

Waschiczek (2002) argues that increased bank competition narrowed the spreads on foreign currency loans.

23

First differences in the variables (prefixed by Δ) are interpreted as the quarter-on-quarter growth rates of those variables expressed in logarithms. Standard t-ratios are reported in parentheses. Only significant coefficients are reported. ECM(loans–3.2*herd)t-1 is the lagged deviation of the actual “transformed” loans from the long-run relationship derived in (2) above; together with its coefficient, it represents the error-correction term capturing the adjustment of foreign currency loans to deviations from their long-run equilibrium.

24

See Appendix, Section D for an explanation of the tests. The estimation found evidence of instability of the demand for loans in late 2000 even after correcting for herd behavior. Perhaps the proxy for herd behavior was not entirely successful at capturing the prevailing herd behavior late in that year. However, the demand for loans has stabilized more recently.

25

Each explanatory variable was subjected to a negative shock in 2002:Q4, equivalent to two standard deviations below the average year-on-year historical rates of change. The variable was interpolated linearly between 2001:Q4 (the last actual value used) and 2002:Q4 (when the full force of the shock was felt), and remained unchanged thereafter. All other explanatory variables assumed their actual values. The starting point of the simulation was selected to allow the dynamics to play out fully by the end of the period (2004:Q3).

26

This result is based on a partial analysis, which does not allow for the impact of a rise in the interest rate differential on the other variables. Moreover, it applies only to small changes.

27

In contrast to ADF tests, which test the null hypothesis of nonstationarity, Johansen’s procedure tests the null of stationarity.

28

See Johansen (1988 and 1995). Estimation and testing were carried out in PcGive Professional v10.3; see Hendry and Doornik (1999).

29

The sample beginning in 1994:Q1 produced the most appealing results.

30

It is unlikely that banks responded to excess demand for foreign currency loans by raising rates on euro loans, which would have been consistent with rising spreads.

APPENDIX Definitions of Variables Used in the Cross-Section Regression

The variables included in Σz=1nθzX are defined as follows:

Contiguous border–a binary dummy variable that is unity if countries i and j share a common border and zero otherwise.

Common language a binary dummy variable that is unity if i and j share a common language and zero otherwise.

Landlocked–the number of landlocked countries in the country pair (i.e., 0, 1, or 2).

Island–the number of island countries in the country pair (0, 1, or 2).

Area–the area size of the country.

Common colonizer–a binary dummy variable that is unity if i and j were ever colonies after 1945 under the same colonizer.

Ever colony–a binary dummy variable that is unity if i ever colonized j, or vice versa.

Currently colonized–a binary dummy variable that is unity if i were a colony of j at time t, or vice versa.

Common country a binary dummy variable that is unity if i and j were part of the same country during the sample period.

EU, NAFTA, MERCOSUR, ASEAN–binary dummy variables that are unity if i and j both belong to the same regional free trade arrangement.

GSP–a binary dummy variable that is unity if i were a Generalized System of Preferences (GSP) beneficiary of j, or vice versa.

References

  • Arora, Vivek, and Athanasios Vamvakidis, 2005, “How Much Do Trading Partners Matter for Economic Growth?IMF Staff Papers, Vol. 52 (No. 2), pp. 2440.

    • Search Google Scholar
    • Export Citation
  • Bussière, Matthieu, Jarko Fidrmuc and Bernd Schnatz, 2005, “Trade Integration of the New EU Member States and Selected South-East European countries: Lessons from a Gravity Model,” in European Economic Integration and South-East Europe: Challenges and Prospects, ed. by K. Liebscher and others (Cheltenham, England and Northhampton, Massachusetts; forthcoming).

    • Search Google Scholar
    • Export Citation
  • Cheung, Yin-Wong, and Frank Westermann, 1999, “An Analysis of German Effects on the Austrian Business Cycle,Weltwirtschaftliches Archiv, Vol. 135 (No. 3), pp. 52231.

    • Search Google Scholar
    • Export Citation
  • Cheung, Yin-Wong, and Frank Westermann, 2000, “Does Austria Respond to the German or the US Business Cycle?International Journal of Finance and Economics, Vol. 5 (No. 1), pp. 3342.

    • Search Google Scholar
    • Export Citation
  • Demekas, Dimitri G. and others, 2005, “Foreign Direct Investment in Southeastern Europe: How (and How Much) Can Policies Help?IMF Working Paper 05/110 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and Douglas A. Irwin, 1998, “The Role of History in Bilateral Trade Flows,” in The Regionalization of the World Economy, ed. by Jeffrey A. Frankel (Chicago: University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Fidrmuc, Jarko, and Iikka Korhonen, 2003, “Similarity of Supply and Demand Shocks Between the Euro Area and the CEECs,Economic Systems, Vol. 27 (No. 3), pp. 31334.

    • Search Google Scholar
    • Export Citation
  • Fidrmuc, Jarko, and others, 2002, “EU Enlargement to the East: Effects on the EU-15 in General and on Austria in Particular,Austrian National Bank , Focus on Transition, 1/2002.

    • Search Google Scholar
    • Export Citation
  • Kose, M. Ayhan, Guy Meredith, and Christopher M. Towe, 2004, “How Has NAFTA Affected the Mexican Economy? Review and Evidence,IMF Working Paper 04/59 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Kose, Ayhan M., Eswar S. Prasad, and Marco E. Terrones, 2003, “Volatility and Comovement in a Globalized World Economy: An Empirical Exploration,” IMF Working Paper 03/246 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Rose, Andrew K., 2004, “Do We Really Know That the WTO Increases Trade?American Economic Review, Vol. 94 (No. 1), pp. 98114.

  • Winckler, Georg, 1993, “The Impact of the Economy of the FRG on the Economy of Austria,” in Unequal Partners: A Comparative Analysis of Relations Between Austria and the Federal Republic of Germany and Between Canada and the United States, ed. by H. von Reikhoff and H. Neuhold (Boulder, Colorado: Westview Press).

    • Search Google Scholar
    • Export Citation
31

Prepared by Natan Epstein.

32

For the purpose of the analysis in this paper, the group of CEECs is defined as Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Romania, the Slovak Republic, and Slovenia. This group represents the bulk of Austria’s trade and foreign direct investment (FDI) links with the region.

33

In 2004, Austria’s GDP per capita was ranked seventh by Eurostat among OECD countries (per purchasing power standards).

34

By comparison, the next three largest destinations for Austria’s exports are Italy, the United States, and Switzerland, which account for 9, 6, and 5 percent of Austrian exports, respectively.

35

Computed over the previous ten-year window; for example, the 0.89 correlation coefficient in 1999 is estimated over the period 1989-99, where the observation for 1991 is omitted in order to abstract from the large jump in the data series associated with the German unification.

36

Defined as private consumption plus gross fixed capital formation.

37

For example, in the Czech Republic, the Slovak Republic, and Croatia, the market shares have reached about 30 percent, 45 percent, and 40 percent, respectively, according to Austrian National Bank figures.

38

Source: Austrian National Bank.

39

Demekas and others (2005) note that the sources of FDI in Southeastern Europe are highly concentrated among five countries, of which Austria ranks second behind Germany.

40

While comovement with Germany was estimated over a ten-year rolling window, limited data for the transition period render the five-year rolling window more appropriate for the CEECs.

41

The economy of the Czech Republic contracted by 0.8 percent and 1.0 percent in 1997 and 1998, respectively, while Austria’s real GDP expanded by 1.6 percent and 3.9 percent during the same period.

42

IMF, World Economic Outlook, April 2005.

43

In the case of Mexico’s increased linkages to a less volatile U.S. economy, Kose, Meredith, and Towe (2004) note that the greater integration with the United States has brought about a decrease in Mexico’s output volatility.

44

In this figure, volatility is measured as the standard deviation of the Hodrick-Prescott-filtered quarterly—year-on-year—real GDP growth rate series and is computed over a five-year rolling window.

45

We use Hungary as a representative country for the CEECs, since it is Austria’s key trading-partner in the region and long-run data for most of the other countries in the region are more limited.

46

In Rose (2004), the gravity model is used to analyze the effects of multilateral trade agreements on trade, while in BFS (2005) the gravity model is adopted to analyze trade linkages between the euro area and Central and Southeastern Europe.

47

For data consistency reasons, we used the Rose panel data through 1997.

48

For definitions of the variables included in Σz=1nθzX see the Appendix.

49

The figures are separated so as to distinguish between the two panel data sets (Figures 6a and 7a reflect the Rose data, while Figures 6b and 7b reflect the BFS data). While the two regressions produce similar estimated coefficients, the panels are not identical and thus the two series are not continuous.

Austria: Selected Issues
Author: International Monetary Fund