قیمت خانه، بی ثباتی های بانکی، و رشد اقتصادی: شواهد از مدل آستانه

This paper examines the effects of house prices on bank instability when gauged at various levels of income growth. Bank stability may respond differently to house price changes or deviations from fundamental values in an economic boom environment than in a bust circumstance. A threshold estimation technique developed by Hansen (1999) is applied to a panel of 286 U.S. Metropolitan Statistical Areas (MSAs) over the period 1990Q1–2010Q4. We consider two house price indicators: the house price changes and the house price deviations from long-run equilibrium. The results suggest the existence of income growth threshold effects in the relationship between house prices and bank instability. Specifically, there are two income growth thresholds when using the house price changes and one income growth threshold when the house price deviations are applied. Robustness results using the non-MSAs sample from 1995Q1 to 2010Q4 provide further evidence of income growth threshold effects.

The role of the real estate market in the U.S. economy is undoubtedly important and conditions in the housing market signal the state of the economy as a whole. The U.S. economy has been sluggish for 4 years since the subprime mortgage crisis developed in 2007 and 2008, which was triggered by the 2005 housing bubble burst. The banking system, which functions as mortgage lenders and frequently uses real estate as collateral, is a link between the housing market and the macro-economy. Recent bank failures have been associated with the housing bubble burst. For example, over four hundred U.S. banks failed in 2008–2011.1 This paper addresses the questions of whether and how house prices affect the stability of banks under different income growth levels using the house prices and bank variables of 286 U.S. Metropolitan Statistical Areas (MSAs) from 1990Q1 to 2010Q4. What distinguishes our work from the previous studies on the relationship between house prices and bank stability is that our work (1) takes into account the various levels of income growth and (2) uses disaggregated data2 which better reveals the heterogeneity of regional real estate markets and commercial banks in the US.3 We define housing markets as MSAs, which correspond to labor market areas within which workers are willing to commute. MSAs can vary considerably from the national average of house prices. Despite the sizable boom-bust pattern in house prices at the national level, regional housing markets in the U.S. experience considerable heterogeneity in the amplitudes of their cycles or in the house price dynamics. Sinai (2012) documents different magnitudes in the booms and busts across MSAs in the US: the 75th percentile MSA experienced 111% trough-to-peak growth in real house prices in the 1990s and 2000s, whereas the 25th percentile MSA had only 32% trough-to-peak real house price growth. We thus take into account the heterogeneity of immobile real estate assets and the regional variation of house prices when measuring their dynamics and deviations from their fundamental values, using quarterly information on real estate prices in 286 U.S. MSAs. Housing markets vary across U.S. regions due to disparities in economic development and population growth, and they are likely to have different impacts on bank stability. MSA-level data have been used to conduct various empirical studies related to the U.S. housing markets. For instance, Bhattacharya and Kim (2011) use a panel of 20 MSAs in the U.S. over the period 1990Q2–2009Q2 to study the impact of underlying economic factors on real house prices. In another study, Zabel (2012) investigates how the housing market affects the flow of workers across cities using house price indexes for 277 U.S. MSAs from 1990 to 2006. In this study, we focus on residential property (one-to-four family residential property) markets in the US, which on average account for over 76% of mortgage debt outstanding for all holders (including major financial institutions, federal and related agencies, mortgage pools or trusts, individuals, and others). Commercial property (nonfarm, non residential) markets on average account for 17% of mortgage debt outstanding for all holders; multifamily residential property markets account for 6%; and farm land accounts for 1%.4 Thus, residential property was and remains a key element in fueling the turmoil in financial markets in terms of their share as collateral in asset-backed securitization, as opposed to other real estate segments.5 Following the previous empirical work, we consider two measures of the house price indicator when assessing the impact of house prices on bank instability: percentage changes in house price index and house price deviations from fundamental values. The first of these is commonly used in the literature. However, researchers have recently argued that house price deviations from the long-run equilibrium should also be considered to study the relationship between house prices and bank stability (Koetter and Poghosyan, 2010). We apply the pooled mean-group (PMG) and mean-group (MG) estimators to estimate house price dynamics and deviations from fundamental values in 286 U.S. MSAs. Our results confirm a common long-run positive relationship among house prices, personal income, and labor force growth in the MSAs, and provide evidence of a house price adjustment to the long-run equilibrium. To assess the impact of house prices on bank instability, we need to determine the state of the banking system. Non-performing loans6 (NPLs) have been a popular indicator used in the literature (Nkusu, 2011 and Kauko, 2012, among others). However, empirical studies using disaggregated bank-specific data for MSAs in the US remain scarce. In this paper, we use NPLs to gauge bank instability at the MSA level7: larger amounts of NPLs relative to total loans in banks indicate increasing bank instability. Other measures for bank instability in previous studies include bank failure rates (Cebula et al., 2011) and the probability of distress events (Koetter and Poghosyan, 2010). To our knowledge, NPLs are the best available measure of bank instability for commercial banks in U.S. MSAs. In the literature, there are two competing theories about the effects of house prices on bank stability: the collateral value hypothesis (Daglish, 2009 and Niinimaki, 2009) and the deviation hypothesis (Von Peter, 2009 and Gerlach and Peng, 2005). The collateral value hypothesis argues that rising house prices promote bank stability by increasing the value of the houses owned by the bank and the value of the collateral pledged by borrowers; thus, it suggests a negative relationship between nominal house price changes and the bank’s NPLs. In contrast, the deviation hypothesis contends that persistently rising house prices enhance larger exposure and the accumulation of risky assets in banks due to (1) a bank’s excessive lending to risky borrowers at cheap rates and (2) risky borrowers’ higher credit demand from banks who bet on further rises in house prices; consequently, it predicts a positive relationship between house price deviations from the fundamental values and the bank’s NPLs. Koetter and Poghosyan (2010) find evidence using data on housing markets and banks in Germany during 1995–2004 to support the deviation hypothesis where bank instability is attributed to house price deviations instead of to changes in nominal house prices. We conjecture that the responses of NPLs to house price changes or deviations could be different when gauged at various levels of income growth, and then apply the threshold model proposed by Hansen (1999) to test the above two hypotheses under different income growth levels. Banks’ asymmetric responses to house price changes or deviations during booms and busts might be attributed to the bounded rationality of investors as described in Gennaioli and Shleifer, 2010 and Gennaioli et al., 2012, and Dieci and Westerhoff (2012). Ample empirical evidence shows that human agents generally act in a boundedly rational manner, and are subject to limited ability and the use of simple heuristics to predict prices or returns (Kahneman et al., 1986 and Smith, 1991). Not all contingencies are represented in the investors’ thought processes, and only the most likely events are retrieved (Gennaioli and Shleifer, 2010). This local thinking, or neglect of low probability risks, results in over-issuance of new securities and financial fragility (Gennaioli et al., 2012). A sharp decline in prices due to fire sales after a substantial surprise to the market can have especially adverse welfare consequences (Shleifer and Vishny, 2010 and Stein, 2012). On the other hand, appreciation in prices would have a less severe impact. Other influential models include Day and Huang, 1990, Chiarella, 1992, De Grauwe et al., 1993 and Chiarella et al., 2002, and De Grauwe and Grimaldi (2006), and Dieci and Westerhoff (2012). To test our conjecture, we use the threshold model to examine the impact of house prices on bank instability under different income growth levels and estimate the income growth threshold endogenously, instead of imposing an exogenous criterion for splitting the sample by income growth levels. Personal income growth rate is the threshold variable which interacts with one of the house price indicators in the threshold model. We consider two model specifications depending on which house price indicator interacts with the threshold variable (personal income growth rate). Empirical results suggest the existence of income growth threshold effects in the relationship between house price and bank instability. In particular, two income growth thresholds are found when changes in house prices index are used, and one income growth threshold is reported when house price deviations from the fundamental values are applied. First, there exist two income growth thresholds of −5.342% and 3.972% when changes in house price index interact with income growth. Changes in house price index have a significant negative effect on NPLs and the size of the impact depends on the thresholds. When income growth is below −5.342%, NPLs decrease by 0.466% if the changes in house price index increase by 1%, holding other things equal. When income growth is between −5.342% and 3.972%, the negative impact is smaller with a coefficient of −0.181. When income growth is greater than 3.972%, NPLs decrease by 0.097% if the changes in house price index increase by 1%, holding other factors equal. Overall, our results using percentage changes in house price index are consistent with the collateral value hypothesis: the larger the changes in house price index, the lower the bank instability. Second, we find single income growth threshold of −4.285% when house price deviations from fundamental values are used. In particular, house price deviations have a significantly positive impact on NPLs with a coefficient of 0.015 when income growth is below −4.285%. This finding is consistent with the deviation hypothesis: the larger the house price deviations, the higher the bank instability. The effect of house price deviations on NPLs is significantly negative but with a smaller coefficient of −0.001 when income growth is above −4.285%. House price appreciations above the long-run equilibrium can slightly lower NPLs when income growth is above the threshold. Finally, we conduct the same analysis using the non-MSAs sample from 1995Q1 to 2010Q4 for robustness checks. In contrast with the MSAs, we find no common long-run relationship among house prices, personal income, and population growth in the non-MSAs real estate markets. This could be due to the even larger heterogeneity in the housing markets in the non-MSAs. Therefore, house price deviations are not well estimated by the PMG estimator, and we thus proceed to estimate the threshold model using only the house price changes. Single income growth threshold is suggested, and the point estimate is −6.19%. NPLs decrease in response to a positive change in house price index, and the size of the impact depends on income growth threshold. Thus, the collateral value hypothesis is supported, which is consistent with the results for the MSAs. The remainder of this paper is organized as follows. Section 2 describes the empirical methodologies and data (1) to measure the deviations of house prices from fundamental values and (2) to examine the effects of house price indicators on bank stability under different income growth levels. The empirical results are presented in Section 3. Section 4 reports the robustness results using the non-MSAs sample. Finally, Section 5 concludes and sheds light on some policy implications.

This paper applies the threshold model to examine the effect of income growth on the relationship between house prices and bank stability using the house prices and bank variables for 286 U.S. MSAs from 1990Q1 to 2010Q4. In the empirical study, we choose non-performing loans to measure bank instability and consider two house price indicators: the percentage changes in house price index and house price deviations from their long-run fundamental values. Following Koetter and Poghosyan (2010), we apply the pooled mean-group and mean-group estimators to estimate house price determinants in the US. Our results confirm that the equilibrium house prices increase with rising demand due to income and labor force growth and provide evidence of a house price adjustment to the long-run equilibrium. Personal income growth rate is considered as the threshold variable which interacts with the house price indicators in the threshold models. Empirical results show the existence of the income growth thresholds in the relationship between house prices and bank instability. In particular, two income growth thresholds (−5.342% and 3.972%) are found in the models where the percentage changes in house price index (ΔHPI) interact with the personal income growth rate; additionally, one income growth threshold (−4.285%) is suggested in the models where the house price deviations (HPD) are used. On one hand, our results suggest that ΔHPI has a significantly negative effect on NPLs, which is consistent with the collateral value hypothesis. Moreover, the negative impact is stronger as the economy experiences a recession with severe negative income growth. On the other hand, house price deviations have a significant positive effect on NPLs when income growth is below −4.285%, which is consistent with the deviation hypothesis. When the economy experiences a recession with income growth below the threshold, house price deviations could deteriorate bank stability. However, the impact is opposite when income growth is above the threshold. For comparison purposes, we conduct the same analysis using the non-MSAs sample from 1995Q1 to 2010Q4 for robustness checks. We find no common long-run relationship among house prices, personal income, and population growth in the real estate markets. This finding is not surprising due to the even larger heterogeneity in the housing markets for non-MSAs. Therefore, house price deviations for non-MSAs are not well estimated by the PMG estimator. We thus apply ΔHPI in the threshold model and find a single income growth threshold of −6.19%. Consistent with the results for MSAs, the collateral value hypothesis is supported where NPLs decrease in response to a positive change in ΔHPI. In summary, this paper finds evidence to support our conjecture that the responses of NPLs to house prices vary under different levels of income growth using disaggregated housing and bank data which better reveal the heterogeneity of regional real estate markets and commercial banks in the US. Moreover, this paper verifies the collateral value hypothesis in the literature when using house price changes in the threshold model for both MSAs and non-MSAs, suggesting that the larger are the changes in house prices, the lower the bank instability. Our study provides new evidence on the relationship between house prices and bank stability and can shed light on some policy implications. First, heterogeneity across the regional real estate markets in the US should be taken into account when making policies. We find that a common positive long-run relationship among house prices, personal income, and labor force growth does exist in MSAs but not in non-MSAs. Policies that promote personal income and labor force growth should help appreciate house prices in the long run in MSAs, but may not have a substantial impact in non-MSAs. Based on these findings, we propose that different policies should be applied to MSAs and non-MSAs, respectively. Second, banks’ responses to house prices are asymmetric during economic booms and busts. In an economic boom, with favorable housing market conditions, the impact of house price deviations on banks’ performance or stability will be less severe because house price changes or deviations are supported by strong economic growth. However, when economic growth is sluggish, undesirable housing market conditions could significantly deteriorate bank stability. Thus, policies for promoting housing market recovery would be a priority in an economic recession. The recent financial crisis in the US has been associated with a sharp decline in house prices, widespread bank failures, and sluggish economic growth. Our results are based on the US data for the period 1990Q1–2010Q4. Therefore, further study about the real estate-financial fragility-economic growth nexus applying longer time series in the future to fully capture the effects of financial crisis would be desirable. Moreover, empirical evidence from other countries at various levels of economic and financial developments would be of great interest.