Information about Asia and the Pacific Asia y el Pacífico
Journal Issue
Share
Article

Thailand: Selected Issues

Author(s):
International Monetary Fund
Published Date:
January 2006
Share
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details

IV. Credit Booms: The Good, The Bad, and The Ugly1

While financial deepening has been shown to be both a cause and an effect of economic development, the rapid growth of credit aggregates has often been associated with episodes of bank distress, leading to the widespread belief that credit booms are a recipe for financial disaster. However, historically, only a minority of boom episodes has ended in a crash. This paper examines the characteristics of a panel of credit booms and identifies factors that can help the early detection of dangerous bubbles from episodes of healthy financial deepening.

A. Introduction

1. The past 20 years have witnessed a global trend towards increasing financial deepening. Financial intermediation has grown and in that context bank credit has increased dramatically in relation to GDP. In most Asian countries the ratio of bank credit to the private sector to GDP (BCPS ratio) more than doubled between 1980 and 2002. In Thailand, it about tripled over the same period, increasing from 27.5 percent to 77.5 percent

BCPS Ratio in Selected Economies(in percent)
19802002
United States31.442.0
United Kingdom26.4138.1
India20.530.8
Indonesia8.320.8
Korea37.389.4
Malaysia36.795.3
Philippines29.331.6
Singapore65.7106.6
Thailand27.577.5
Source: IFS, staff’s calculations
Source: IFS, staff’s calculations

2. However, this process has not been always a smooth one. While in some countries financial deepening has followed an even path, in others it has been a bumpy process with sharp accelerations in aggregate credit, or credit booms, sometimes followed by episodes of financial distress and banking crises. This has contributed to the widespread belief that credit booms are a recipe for financial disaster.2 However, the evidence shows that several episodes of fast credit growth have soft-landed without causing any disruption.

3. Recent theoretical contributions suggest that different types of credit booms exist, supported by different underlying economic forces.3 Some booms—the “good” ones—reflect healthy financial deepening as credit grows faster than output as an economy develops. Others are associated with the financing needs of the corporate sector during the upside phase of the cycle. These booms, while not intrinsically “bad”, can lead to increased fragility if banks’ risk attitude changes with the cycle, if the rapid credit growth puts the resources of bank supervisors under strain, or if a financial accelerator mechanism leads to excessive bank exposure.4 Finally, credit booms can be the “ugly” reflection of financial intermediation bubbles and destined to lead to financial crises and often associated with macroeconomic imbalances, such as large current account and public deficits and asset price bubbles.

4. The question is, then, whether we can tell healthy from dangerous credit growth in advance and possibly intervene early, hence, avoiding major financial problems. This paper sheds some light on this issue by identifying episodes of faster-than-normal credit growth and examining if, based on the information available to policy makers and market participants at the time of the episode, it is possible to distinguish between benign financial deepening and dangerous credit bubbles. The paper presents a new empirical framework to tackle this issue. The results suggest that it is not possible to fully discriminate between “good” and “bad” (or “ugly”) credit booms, but several macroeconomic variables help to predict whether a boom is heading for some form of financial distress. In particular, booms that are longer lasting and greater in size and those associated with high inflation rates and large current account deficits are more likely to lead to crises. There is also some evidence that bad booms are associated with large increases in asset prices and with rapid investment growth. Finally, bad booms often occur in the context of real exchange rate appreciations and asset price bubbles, possibly associated with capital inflows. Further research is needed to refine these estimates in particular with regard to the role of sectoral imbalances and asset prices for which limited data have been available so far.

5. The proactive use of prudential regulation and bank supervision may complement monetary policy to preserve financial stability during booms. Monetary policy alone may not be sufficient to manage the trade-off between fast financial deepening and financial stability. First, especially in inflation-targeting regimes, monetary policy may find itself facing conflicting objectives if a boom develops during a period of low inflation. Second, monetary policy may be ineffective to the extent that a boom is fueled by capital inflows under an open capital account. Finally, monetary policy is a blunt and expensive instrument to tackle credit “bubblets” which may develop in specific sectors of the economy, while aggregate credit numbers are stagnant. In that context, regulatory curbs and prudential measures have been used with some success to limit sectoral credit growth.5

6. In recent years Thailand has experienced sluggish credit growth, but some segments of the credit market have been developing fast. Since 1998, aggregate bank credit to the private sector has been anemic, reflecting bank restructuring and corporate sector deleveraging, and the BCPS ratio has decreased consistently until 2004. Since 2002, however, banks have been actively extending credit to the household sector, resulting in fast rates of growth in the mortgage and credit-card markets. While, between 2001 and 2004, total loans grew by only about 19 percent on a nominal basis, housing loans grew by about 42 percent, lending support to a recovering real estate market, and overall loans to households grew by 66 percent. Over the same period, credit-card debt outstanding almost doubled. The Bank of Thailand has acted early to avoid the accumulation of potentially dangerous imbalances in the banking system. For example, prudential curbs were introduced to limit the growth of credit-card debt, establishing ceilings on outstanding balances and minimum income requirements for card holders. These measures have encountered some success with the rate of growth of credit-card debt slowing to an annualized 9 percent in the first four months of 2005. It should be noted that the analysis below does not apply to the current situation in Thailand. Thailand is not experiencing a credit boom, but only fast credit growth in some sectors of the economy.

Thailand: Nominal Credit Growth 1991-2004

(in percentage change over previous year)

Source: Bank of Thailand.

B. Stylized Facts

7. A positive relationship between financial development and growth has been long established. Furthermore, a more recent literature based on industry-level data has shown that financial development is not just the result but also a determinant of economic growth.6 Fast credit growth is, then, a positive development to the extent that it reflects fast financial deepening. For example, Ireland’s strong macroeconomic performance of recent years has been accompanied and supported by an extremely fast growth of the BCPS ratio, without causing any major financial problems.

Financial Development and Growth

(1970-2002 Average)

Selected Booms and Crises

8. However, excessively fast credit has also been associated with increased financial fragility and banking crises. Most major banking crises in the past 25 years have been in the wake of periods of extremely fast credit growth. This regularity is not limited to emerging markets, but extends to advance economies as well: the Scandinavian banking crisis of the early 1990s followed a period of extreme credit growth. Other notable examples include Argentina in 1980, Chile in 1982, Mexico in 1994, and the Asian crisis of 1997. In the run-up to these crises the annual growth rate of the credit-to-GDP ratio often exceeded 5 percent, and in some countries was as high as 10 percent. All these crises involved heavy macroeconomic losses and were followed by prolonged periods of sluggish credit growth.

9. That said, only a minority of credit booms has led to episodes of financial distress. Out of 150 credit booms identified in this paper, only about a fifth (31) precedes systemic banking crises, with that proportion rising to about a third (47) if minor episodes of financial distress are included. Banking crises can also occur without lending booms. In our sample, for example booms precede less than half of the 70 systemic crises and the 110 episodes of financial distress.

C. Data and Methodology

10. This paper identifies credit booms by examining whether the actual rate of growth of credit in an economy—as measured by BCPS ratio—appears abnormally high relative to its previous trend. Then, based on a panel of developing and industrialized countries, it constructs a dummy variable that takes the value zero in the case of “good” booms and the value one in the case of “bad” booms defined as those that are followed by an episode of financial distress within two years from their end. Finally, it collapses the panel to a cross-section of credit boom episodes and regresses the dummy variable on a set of macroeconomic variables contemporaneous to the boom.

11. The sample used in our regressions consists of 73 countries over the period 1980–2002. It was selected along criteria similar to those in Gourinchas, and others (2001): for all the countries in the sample at least 15 years of data are available and the private credit to GDP ratio is at least 15 percent. In addition, several countries for which data appeared unreliable or that were affected by exogenous shocks such as persistent political turmoil and civil wars were excluded from the sample. Bank credit data are from line 22d of the International Financial Statistics (IFS) and GDP data are from line 99b. These series were corrected for structural breaks which otherwise would have likely been identified as booms. The series on inflation and the current account are from the WEO. Data on banking crises and episodes of financial distress are from Caprio and Klingebiel (2003).

12. Since credit is a stock variable measured at year-end, the BCPS ratio is constructed with the geometric average of GDP in years t and t+1. This measure has two main advantages. First, it can be built with readily available data with widespread country and time-series coverage. Second, it does not consider the financial sector in isolation, but relates it to the size of the economy, while at the same time correcting for the procyclicality of bank lending. That said, because of the positive relationship between financial development and growth, bank lending follows a positive trend, even when measured in relation to GDP. Therefore, credit booms need to be isolated as definite events separate from normal increments in the volume of credit.

13. We apply the methodology developed in Gourinchas et al. (2001) and define a lending boom as an episode where the BCPS ratio deviates from a rolling, backward-looking, country-specific trend (estimated by a non-linear trend).7 This means, that credit growth in each year x will be compared with a trend estimated over the period 1980-x. The idea is that such trend represents the historically “normal” pace of credit growth for each particular country. Furthermore, the estimated trend summarizes the information about past credit growth available to policy makers and market participants at the time of the boom. Alternatively, a trend could be estimated over the entire sample period, as in Mendoza and Terrones (2004). However, this approach would have two drawbacks. First, it would tend to overestimate bad credit booms because of the bias introduced by the subsequent crisis. Second, it would make use of information not available at the time of the boom, and hence, would make the estimates difficult to apply operationally.

14. Based on this approach, an episode of fast credit growth becomes a boom if its deviation from the trend exceeds a certain threshold. As in Mendoza and Terrones (2004), this paper employs country- and path-dependent thresholds, based on the standard deviation of the historical deviations of the BCPS ratio from its estimated trend. More specifically, an episode becomes a boom if the BCPS ratio exceeds or meets either of the two following conditions:

  • i) The deviation from trend is greater than 1.5 times its historical country-specific standard deviation and the annual growth rate of the BCPS ratio exceeds 10 percent.
  • ii) The annual growth rate of the BCPS ratio exceeds 20 percent.

This definition takes into account country-specific conditions and reflects both the relative level and the speed of the BCPS ratio. A country-specific threshold is needed since what may seem like a large deviation in countries with a historically smooth credit growth may be the norm in a country with an experience of uneven growth. The growth rate of the BCPS ratio is included to control for those cases where because of a relatively smooth acceleration in credit, extremely fast credit growth may occur while the actual BCPS ratio falls close to its trend.

15. Once a credit boom is identified, its starting point is defined according to a similar criterion, that is the earliest year in which: (i) the BCPS ratio exceeds its trend by more than three-fourths of its historical standard deviation and its annual growth rate exceeds 5 percent; or (ii) its annual growth rate exceeds 10 percent. A boom ends as soon as either of the two following conditions is met: (i) the growth of the BCPS ratio turns negative; (ii) the BCPS ratio falls within three-fourths of one standard deviation from its trend and its annual growth rate is lower than 20 percent.

16. The panel is then collapsed to a cross-section of booms. Two versions of a dummy variable, BAD, are constructed taking value one for booms followed within two years from their end by episodes of financial distress and by full-fledged banking crises, respectively. The country-specific mean value over each boom period of several macroeconomic and structural variables, such as inflation, the current account balance, GDP growth, the capital account balance, the change in stock market capitalization, boom duration, and deviation of the BCPS ratio from its trend are associated with each observation. Finally, the following model is estimated with a probit regression:

where α is a constant and X is a vector of country-specific macroeconomic and structural variables averaged over the boom period.

D. Results

17. The descriptive statistics of macroeconomic variables during credit boom episodes suggest that good and bad booms differ along several dimensions:

  • i) Bad booms appear to occur more often in countries with relatively higher per-capita income and BCPS ratios. This suggests that episodes of fast credit growth are likely to reflect a catching up in financial deepening when they occur in poorer and less financially intermediated countries.
  • ii) The size and duration of the boom matter. Bad booms last on average almost a year longer and have deviations from trend about 25 percent larger than good ones.
  • iii) Bad booms are associated with higher economic growth than good booms, but are accompanied by higher inflation and larger current account deficits. This is consistent with the idea that bad booms are bubble-like phenomena often accompanied by domestic and external macroeconomic imbalances.
  • iv) Bad booms are associated with large capital account surpluses and real exchange rate appreciations, consistent with the notion that domestic banking systems may run into problems when they need to intermediate large capital inflows.
  • v) Finally, there is evidence that asset price bubbles and investment booms may play an important role in determining whether a credit boom is bad or good. During bad booms stock market capitalization increases on average by 113 percent compared to a much lower 47 percent in good booms. The investment-to-GDP ratio increases by 1.3 percentage points compared to 0.7 percentage points.
Selected Macroeconomic Indicators during Credit Booms 1/(in percent, unless otherwise indicated)
AllGoodBad
Number of Episodes795029
GDP per-capita (in US$)$5,050$4,840$5,390
BCPS ratio34.530.940.1
BCPS ratio deviation from trend (in percent of GDP)4.33.94.8
Duration (years)3.02.73.5
Annual GDP Growth3.73.44.2
Annual Inflation13.210.318.2
Current Account Balance (in percent of GDP)-1.8-0.8-3.5
Capital Account Balance (in percent of GDP)1.30.62.8
Change in REER 2/8.06.710.6
Change in Stock Market Capitalization 2/81.847.2113.9
Change in Investment (in percent of GDP) 2/0.90.71.3
Sources: WEO, IFS, Staff’s calculations

Based on 79 boom episodes for which all data is available and inflation < 100 percent.

Based on a subset of episodes for which data is available.

Sources: WEO, IFS, Staff’s calculations

Based on 79 boom episodes for which all data is available and inflation < 100 percent.

Based on a subset of episodes for which data is available.

18. These stylized facts are reflected in the results of our probit regression. Data availability limited our choice of variables for this empirical model. The stylized facts described above suggest that variables such as real exchange rate appreciation, capital account balance, and changes in stock market capitalization should enter our regression. However, this would reduce drastically the number of observations and would make it close to impossible to obtain meaningful estimates. We, then, estimate a very parsimonious model where the dummy variable BAD is regressed over the current account balance (CA), inflation (INFL), boom duration (DUR), and the deviation of the BCPS ratio from its trend (DEV):

19. All coefficients have the expected sign and are statistically significant. Furthermore, their effect is economically relevant. Inflation and the absolute deviation from trend have the largest impact. A 1 percent increase in the inflation rate increases the probability that a boom will result in a crisis by 0.5 percentage points and an equivalent increase in the deviation from trend increases it by 0.6 percentage points. One percent increases in boom duration and the current account balance have smaller impacts of about 0.4 percentage points and 0.1 percentage points, respectively. Notably, the pseudo R-square of the regression is relatively low, indicating that still a lot of what makes a credit boom bad remains to be explained and that further research is needed to refine these estimates. For example, based on a smaller set of countries than in this paper, Borio and Lown (2002) find that sustained rapid credit growth combined with large increases in asset prices appear to increase the probability of financial instability.

20. Finally, a word of caution on the interpretation of these results. While models as that in this paper can be useful to predict whether a boom is bad or good, they have limitations when it comes to policy analysis. The extreme simplicity of the model leaves it open to omitted variable biases and potential simultaneity biases associated with the variables employed as regressors sheds doubts over the results of comparative static exercises based on the model. That said, this model is a first step towards a better understanding of what determines whether a credit boom is bad or good.

Probit Regression(Sample of 79 booms with inflation < 100 percent)
Dependent Variable BADCoefficientStd. ErrorzElasticity
Current Account Balance-0.05*0.03-1.660.11
Duration0.10*0.061.650.36
Deviation from trend0.12**0.062.030.61
Inflation0.03***0.012.620.48
Constant-1.88***0.45-4.15
BAD is based on all episodes of financial distress.*, **, and *** represent statistical significance at the 10 percent, 5 percent, and 1 percent, respectively
BAD is based on all episodes of financial distress.*, **, and *** represent statistical significance at the 10 percent, 5 percent, and 1 percent, respectively

E. Related Literature

21. A large literature on banking crises finds a positive, but often small and not always significant, link between credit growth and financial crises. Demirguc-Kunt and Detragiache (2002) and Kaminsky and Reinhart (1999) find evidence that fast credit growth increases the probability of banking crises. Gourinchas, and others (2001) examine a large number of episodes characterized as lending booms and find that the probability of having a banking crisis increases after such episodes and that the conditional incidence of having a banking crisis depends critically on the size of the boom. However, they find that the increase is not statistically significant, and that, as with this paper, and consistent with Tornell and Westermann (2001), most lending booms are not followed by crises. Mendoza and Terrones (2004) reach, instead, the conclusion that lending booms are typically bad. However, their definition of boom may entail a bias as their trend is estimated over the entire sample period. Hilbers, and others (2005) compare the behavior of several macroeconomic variables around booms and find evidence consistent with the results in this paper, in particular with regard to inflation and the current account balance.

22. A few recent theoretical papers have provided explanations for why lending booms can lead to financial crises, especially in emerging economies. Here we provide a brief and far from exhaustive review of these contributions. “Financial accelerators” (Kyiotaki and Moore, 1997): an increase in value of collateralizable goods releases credit constraints. This leads to an increase in the volume of lending, which in turn fuels further increases in asset values. When a negative shock inverts this cycle, the banking system may find itself overexposed. “Institutional memory” (Berger and Udell, 2004): in periods of fast credit expansion it is difficult for banks to recruit enough experienced loan officers (especially if there has not been a crisis for a while). This leads to a deterioration of loan portfolios, which reduces bank profitability and increases the probability of a crisis. “Adverse Selection and the Business Cycle” (Dell’Ariccia and Marquez, 2005): during the expansionary phase of the cycle, adverse selection is less severe and banks find it optimal to reduce borrower screening and lending standards to trade quality for market share. This leads to deteriorated portfolios, lower profits, and an increased probability of a crisis.8

References

    BergerA. and G.Udell2004The Institutional memory Hypothesis and the Procyclicality of Bank Lending Behavior,Journal of Financial Intermediation Vol. 13 pp. 458495.

    • Search Google Scholar
    • Export Citation

    BorioC. and P.Lowe2002Asset Prices, Financial and Monetary Stability: Exploring the Nexus,BIS Working Paper No. 114.

    CaballeroR. and A.Krishnamurthy2001International and Domestic Collateral Constraints in a Model of Emerging Market Crises,Journal of Monetary Economics Vol. 48 pp. 513548.

    • Search Google Scholar
    • Export Citation

    CaprioG. and D.Klingebiel2003Episodes of Systemic and Borderline Financial Crises,World Bankmimeo.

    Dell’AricciaG. and R.Marquez2005Lending Booms and Lending Standards,Journal of Financeforthcoming.

    Demirguc-KundA. and E.Detragiache2002Does Deposit Insurance Increase Banking System Stability? An Empirical Investigation,Journal of Monetary Economics Vol. 49 pp. 13731406.

    • Search Google Scholar
    • Export Citation

    GourinchasP.O.R.Valdes and O.Landerretche2001Lending Booms: Latina America and the World,Economia Spring Issue pp. 4799.

    • Search Google Scholar
    • Export Citation

    HilbersP.I.Otker-RobeC.Pazarbasioglu and G.Johnsen2005“Assessing and Managing Rapid Credit Growth and the Role of Supervisory and Prudential Policies” IMF mimeo.

    • Search Google Scholar
    • Export Citation

    KaminskyG. and C.Reinhart1999The Twin Crises: The Causes of Banking and Balance-of-Payments Problems,American Economic Review Vol. 89 pp. 473500.

    • Search Google Scholar
    • Export Citation

    KiyotakiN. and J.Moore1997Credit Cycles,Journal of Political Economy Vol. 105 pp. 211248.

    RajanR.1994Why Bank Credit Policies Fluctuate: A Theory and Some Evidence,Quarterly Journal of Economics Vol. 109 pp. 399441.

    • Search Google Scholar
    • Export Citation

    RajanR. and L.Zingales1998Financial Development and Growth,American Economic Review Vol. 88 pp. 559586.

    TerronesM. and E.Mendoza2004“Are Credit Booms in Emerging Markets a Concern?” World Economic Outlook pp. 148166 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    TornellA. and F.Westermann2002Boom-Bust Cycles in Middle Income Countries: Facts and Explanation,IMF Staff Papers Vol. 49Special Issue pp. 111153.

    • Search Google Scholar
    • Export Citation
1Prepared by Giovanni Dell’Ariccia.
4See Rajan (1994), Kiyotaki and Moore (1997), and Dell’Ariccia and Marquez (2004).
5See Hilbers et al. (2005) for a discussion of this issue.
6See for example, Rajan and Zingales (1998).
7Gourinchas et al. (2001) employs a Hodrick-Prescott (HP) filter. This paper uses a cubic trend to avoid problems related to the end-point bias associated with the HP methodology.
8Other related theoretical contributions include Caballero and Krishnamurthy (2001) and Tornell and Westermann (2002).

Other Resources Citing This Publication