The real estate and credit bubble: Evidence from Spain

Akin, Ozlem; García Montalvo, José; García Villar, Jaume; Peydró, José-Luis; Raya, Josep Maria Article The real estate and credit bubble: Evidence from Spain SERIEs - Journal of the Spanish Economic Association Provided in Cooperation with: Spanish Economic Association Suggested Citation: Akin, Ozlem; García Montalvo, José; García Villar, Jaume; Peydró, José-Luis; Raya, Josep Maria (2014) : The real estate and credit bubble: Evidence from Spain, SERIEs - Journal of the Spanish Economic Association, ISSN 1869-4195, Springer, Heidelberg, Vol. 5, Iss. 2/3, pp. 223-243, https://doi.org/10.1007/s13209-014-0115-9 This Version is available at: https://hdl.handle.net/10419/158524 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/2.0/ SERIEs (2014) 5:223–243 DOI 10.1007/s13209-014-0115-9 ORIGINAL ARTICLE The real estate and credit bubble: evidence from Spain Ozlem Akin ·José García Montalvo · Jaume García Villar ·José-Luis Peydró · Josep Maria Raya Received: 31 May 2013 / Accepted: 7 July 2014 / Published online: 14 August 2014 © The Author(s) 2014. This article is published with open access at SpringerLink.com Abstract We analyze the determinants of real estate and credit bubbles using a unique borrower-lender matched dataset on mortgage loans in Spain. The dataset contain real estate credit and price conditions (loan principal and spread, and the appraisal and market price) at the mortgage level, matched with borrower characteristics (such as income, labor status and contract) and the lender identity, over the last credit boom and bust. We find that lending standards are softer in the boom than in the bust. Moreover, despite some adjustment in lending conditions in the good times depending on bor- We wish to thank the suggestions of two referees and the editor in charge of the paper (Nezih Guner). Montalvo would like to acknowledge the financial support of Project ECO2011-25272 (Ministerio de Ciencia e Innovacion) and the Fellowship ICREA-Academia for Excellence in Research funded by the Generalitat de Catalunya. J. García Villar would like to acknowledge the financial support of Project ECO2012-39553-C04-01 (Ministerio de Economía y Competitividad). The usual disclaimer applies. O. Akin ·J. G. Montalvo ·J. García Villar ·J.-L. Peydró ·J. M. Raya Department of Economics and Business, Universitat Pompeu Fabra, Barcelona, Spain O. Akin Faculty of Economics and Administrative Sciences, Ozyegin University, Istanbul, Turkey J. García Villar ·J.-L. Peydró ·J. G. Montalvo Barcelona Graduate School of Economics, Barcelona, Spain J.-L. Peydró ·J. G. Montalvo ICREA, Barcelona, Spain J.-L. Peydró Centre for Economic Policy Research, London, UK J. M. Raya (B) Escola Universitaria del Maresme (Tecnocampus, Universitat Pompeu Fabra), Ernest Lluch 32, 08302 Mataro, Barcelona, Spain e-mail: [email protected] 123 224 SERIEs (2014) 5:223–243 rower risk, the results suggest too soft lending standards and excessive risk-taking in the boom. For example, mortgage spreads for non-employed are identical to employed borrowers during the boom. Banks with worse corporate governance problems soften even more the standards. Finally, we analyze the mechanism by which banks could increase the supply of mortgage loans despite of regulatory restrictions on LTVs. The evidence is consistent with banks encouraging real estate appraisal firms to introduce an upward bias in appraisal prices (29 %), to meet loan-to-value regulatory thresholds (40 % of mortgages are just bunched on these limits), thus building-up the credit and the real estate bubble. Keywords Lending standards ·Credit supply ·Excessive risk-taking ·Bank incentives ·Conflicts of interest ·Moral hazard ·Prudential policy ·Financial crises · Asset price bubble JEL Classification G01 ·G21 ·G28 1 Introduction In the summer of 2007 the economies of the United States and Western Europe starting suffering bank liquidity and real estate problems, which were followed by a severe banking crisis, with a strong credit reduction and economic recession that still last in Spain. This sequence of events was not unique: Banking crises are recurrent phenomena, often-triggering deep and long-lasting recessions (Reinhart and Rogoff 2009; Schularick and Taylor 2012). A weakening in banks’ balance-sheets may lead to a contraction in the supply of credit and to a slowdown in real activity (Bernanke 1983). Moreover, highly leverage non-financial borrowers may face tightened lending conditions in crises due to a debt overhang problem (Myers 1977). Importantly, banking crises are not random exogenous phenomena, but regularly come on the heels of periods of strong credit growth and asset-price bubbles (Kindleberger 1978;Schularick and Taylor 2012;Gourinchas and Obstfeld 2012). As in other important banking crises in history, the boom and the bust in the housing market, and the associated credit cycle, seem the main drivers of the crises that hit USA, Ireland, UK and Spain, among other countries, in 2007–2008. It is therefore crucial to answer the following questions: Were the lending conditions and standards to housing loans too soft in the boom? Did all banks behave similarly or were there differences? Were pervasive bank incentives present in the boom? Did they contribute to the real estate and credit bubble? If so, how did banks circumvent the regulatory constraints? Did the lending standards tighten in the crisis? Spain offers an excellent setting to analyze these questions. Spain, a bank dominated economy, suffered one of the highest boom and bust in the housing and credit market over the last 10 years. Household mortgages were at the peak 65 % of the GDP and loans to real estate developers and construction firms accounted for another 45 % of the GDP. Therefore, the size of the loans’ pool related directly with real estate activities (production and transactions) amounted to more than 100 % of GDP. Moreover, household debt in Spain (loans to households for mortgages and consumer credit) was 123 SERIEs (2014) 5:223–243 225 91 % of GDP in 2010, just below 106 % in UK and 95 % in USA, but substantially higher than France and Germany, with 69 and 64 %, respectively.1 Despite the importance of this concentration of banking risk on the real estate sector, in particular on household mortgage credit, there is scant evidence identifying the channels that explain the real estate and credit booms. The main reason is the lack of individual (loan) borrower-lender matched level data.2We have access to a unique dataset obtained from a housing market intermediary on mortgage loans from 2005 to 2010 matched with borrower and lender identity for 30,262 mortgage loans concentrated somewhat in the middle-low distribution part of the price distribution. The loan-level dataset contain the loan price (spread), principal and appraisal value of the house. Moreover, we know the identity of the lender and, therefore, we can classify the lender as a commercial bank, a savings bank (caja de ahorros) and a non-bank financial institution (financiera), and also exploit the variation across banks in the ex-post (revealed) risks that they took (whether banks were rescued or intervened). In addition, we know crucial characteristics of the borrower such as labor contract (temporal vs. permanent), status (employed in the private or public sector or not employed), income, education, age, location, etc. Finally, we have been able to match part of this sample (10.92 %) with a dataset also from the housing market intermediary containing the market price and some characteristics of the dwelling. For this subsample, we have the appraisal, the transaction (market), and the officially registered price for each house. We first analyze—at the loan level—credit conditions (both loan to values and loan spreads) in both good (2005:Q1–2007:Q2) and bad (2007:Q3–2010:Q4) times, where the turning point in bank liquidity, credit and real estate dynamics starts in the summer of 2007.3We find robust evidence that lending conditions and standards were softer in the boom than in the bust. For example, household income and labor contract/status matter more (statistically and economically speaking) for LTV (loan to value) and loan pricing in the bust than in the boom. Despite some adjustment in lending conditions in the good times depending on household risk, the results suggest too soft lending standards (excessive bank procyclically/risk-taking) in some loan margins in the boom. Controlling for other key borrower variables such as income, workers with temporal contracts get in the boom the same LTVs (both statistically and economically speaking) as workers with permanent contracts in the crisis, instead, temporal workers obtain less LTVs and pay substantially more spreads. As temporal workers are the ones that mainly went unemployed in the crisis period, these results suggest not only ex-ante high risk, but also ex-post. Even more important for credit supply and excessive bank risk-taking, we 1See Cecchetti et al. (2011), in particular Table A2.1. 2Note that even the Spanish house price indices are based on appraisals or administrative prices. None of them take market prices as their basic source. In fact, the dataset used in our paper is the first one that contains market prices of Spanish properties for a large sample of dwellings, and also the first paper using loan-level data with loan principal and prices. Bover et al. (2014) uses a European survey on borrower debt, while in our case we are using actual data (not coming from a survey) and a matched lender-borrower data which is crucial to make any inference on credit supply and risk-taking. 3The tightening of lending conditions came in Spain, as in the whole Euro Area, in the summer of 2007 (see Maddaloni and Peydró 2011). Some indicators of real estate prices also started going down in the summer of 2007. See Sect. 2. 123 226 SERIEs (2014) 5:223–243 find that borrowers who are not employed pay identical loan spreads than employed ones in the boom. However, in the crisis period, the difference in spreads is substantially different. Furthermore, we also find that higher LTV did not impact higher loan spreads (and the same for spreads on LTVs) in the boom.4All in all, the results in Spain suggest too soft lending standards and excessive risk-taking in the boom. Finally, controlling for borrowers’ fundamentals and other characteristics, we find that rescued banks increase even more the LTVs in the boom than other banks, where banks were rescued either individually at the beginning of the crisis or later in the crisis with the Spanish banking rescue fund for bank restructuring called FROB (Fondo de Reestructuración Ordenada Bancaria). As all these rescued banks were savings banks (cajas), the results suggest that banks with worse corporate governance problems soften even more the lending standards, thereby taking excessive risks in the boom.5 Our previous results suggest that high LTVs were used for risky borrowers (e.g. temporal workers) and by risky banks (the rescued ones). However, there were regulations and restrictions on LTVs in Spain. How could banks overpass the tough regulation in terms of LTVs? How did banks contribute to the real estate bubble, both in terms of real estate pricing and mortgage loan principals? What were key moral hazard (conflicts of interest) problems that explain the build-up of the credit and real estate bubble in Spain? The specific agency mechanism that inflated the bubble in the US was quite different from the forces at work in the Spanish case. In both cases, lax standards and excessive credit were the ultimate causes of the house price inflation. However, in the case of the US, those lax standards for mortgage granting were the result of perverse 4Besley et al. (2013) argue that a regression between loan principal (LTV) and spreads can explain whether lending standards are soft or not. Of course, the two variables are endogenous and thus they have to be interpreted without causality implications. However, despite on this limitation, in Besley et al. (2013), who analyze the UK market, there is always a positive significant relationship between loan principal (LTVs) and spreads (in the Spanish case it is only in the crisis period). 5Only one commercial bank failed, Banco de Valencia, but it was controlled by a caja, Bancaja. Nationalized banks include what is called Group 1 banks, which needed a large capital injection and are controlled by the FROB: Catalunya Banc, BFA-Bankia, and Nova Caixa Galicia (sold to Banesco in 2013) and Banco de Valencia (sold to CaixaBank in 2012). Banks in Group 2 include financial institutions that needed public funds to fill their capital shortfall but the public sector did not get the majority of the capital. That group that received the injection of public funds includes Liberbank, Banco Mare Nostrum, CEISS and Caja3. For all the information regarding the restructuring banking process in Spain during the crisis, including FROB and savings banks, see Bank of Spain’s website, in particular: http://www.bde.es/bde/en/secciones/prensa/ infointeres/reestructuracion/. In 2005 there were 46 savings banks (cajas) representing 42 % of total bank assets in Spain. Commercial banks are traditional banks (including foreign banks) that have shareholders as owners of the bank. Cajas on the other hand rely on a general assembly for governance, consisting of representatives of regional and municipal government (politicians), depositors’ representatives, and non-governmental organizations such as the Catholic Church, for instance. The general assembly elects a board of directors who look for a professional manager to run the banking business. However, in many cases these bank managers and members of the Board did not have adequate human capital to run these banks (Cunat and Garicano 2010). Commercial banks’ profits can either be retained as reserves or pay out as dividends. For the Cajas, the profits are either retained or paid out as social dividend (i.e. to build and run educational facilities, libraries, sport facilities, pensioners’ clubs and so on where the Cajas operate). However, despite their differences in governance structures, both commercial banks and Cajas operate under the same regulatory framework and compete against each other in common markets. After the banking restructuring process during the recent crisis, only two tiny cajas remain, all the others partly merged and became commercial banks. 123 SERIEs (2014) 5:223–243 227 incentives in the housing finance sector related with the securitization process, and the possibility of taking out of the banks’ balance sheets the securitized mortgages.6 We show evidence consistent with an alternative channel for Spain: real estate appraisal firms were encouraged to introduce an upward bias in appraisal prices to satisfy their owners or most important clients (banks). The Bank of Spain recommends not to grant credits over 80 % of the value of the property and covered bonds, a crucial source of financing for banks, was possible only if collateralized with residential mortgages with LTVs lower than 80 %. Moreover, it was considered a bad practice in lending if LTVs were higher than 100 %. In fact, the requirements of information of the Bank of Spain ask banks to report the mortgages classified in three categories: 80 % or below; between 80 and 100 %; and above 100 %. Appraisal firms’ incentives in Spain were distorted, as financial institutions own most of the appraisal firms (or are the crucial clients for these firms).7During the boom period, appraisers had the incentive to approve even an inflated transaction price in order to satisfy its client (financial institutions) by allowing the borrower to obtain a high mortgage principal by adjusting the actual loan to value of the mortgage when the borrower did not have enough resources for the down payment or did have the resources but preferred to borrow more, thereby circumventing regulatory restrictions. As the appraisal price was the price used by financial institutions to determine the loan to value ratio, this artificial increase in appraisal prices permit to draw larger mortgages. All in all, our results suggest that banks encourage real estate appraisal firms to introduce an upward bias in appraisal prices, to meet loan-to-value regulatory thresholds (40 % of loan mortgages are just bunched in these regulatory limits), thus building-up the credit and the real estate bubble.8 Our paper contributes to the literature in two ways. There are several papers analyzing the lending conditions and standards in the Spanish market (Jiménez et al. 2012, 2013a,b), but these papers only analyze business loans and do not analyze loan prices. Moreover, there are papers analyzing mortgage loans in other countries; in particular, Besley et al. (2013) analyze UK mortgages, but do not compare lending standards in boom versus bust nor analyze key borrowers’ risk characteristics such as temporal or inactive workers (see Peydró 2012), which is our first main contribution to the literature. Finally, we uncover a key mechanism by which banks drive the housing and credit bubble, which is our second main contribution. We identify the specific mechanism that led to the creation of a large credit and housing bubble in Spain. Opposite to the US case, the Spanish banking regulation has a much larger regulatory perimeter for banks than the US, and therefore it did not foster the creation of shadow banking activities.9 In addition, Spain was pioneer on the use of countercyclical buffers. Despite these 6For the US case, see Mian and Sufi (2009)andKeys et al. (2010). For the European (including Spain) versus the US case, see Maddaloni and Peydró (2011). 7An appraiser is supposed to value homes as if they were purchased for cash, without any financial or other incentives to the buyer (Mae 2005,2007). Lacour-Little and Malpezzi (2003) find a significant negative correlation between the quality of appraisals and mortgage defaults. Lang and Nakamura (1993) already pointed that, in this case, the bank would require a larger down-payment. 8See also Garcia-Montalvo (2009). 9While in the US the business model during that period was “originate to distribute”, in Spain the model was “originate to hold”. Spain (and Portugal) differed in their regulation of capital requirements from US 123 228 SERIEs (2014) 5:223–243 regulatory constraints, Spain could not avoid a credit and housing bubble. The mechanism did not rely on the lack of incentive for monitoring the quality of mortgages but the ability of financial institutions to influence the valuations of properties by the appraisal companies. The paper proceeds as follows. Section 2discusses the Spain’s housing and credit boom and bust. Section 3introduces the characteristics of the unique dataset we use. Section 4presents and discusses the results. Section 5concludes. 2 The housing and credit markets in Spain During the first decade of this century, Spain has experienced one of the most important housing booms among developed economies. This housing boom was one of the main engines for economic growth in Spain. In fact, during the period 2002–2006 the growth of the construction sector explained around 20 % of GDP growth. During many years, the production of new dwellings in Spain was higher than the sum of the new dwellings in Germany, France and Italy. The boom was the strongest during the period 2005–07. For example, based on the official statistics of the Department of Public Works, housing initiations reached 860.000 dwellings in 2006. Between 1998 and 2007, the housing prices in Spain tripled in nominal terms as reflected by the housing price statistics of the Department of Public Works. The average number of conceded mortgages was of more than 1 million per year.10 These amounts are quite remarkable if we consider that in Spain the annual average number of households in that period was of 15.5 million. Likewise, these amounts were also possible by a very strong competition in the mortgage origination business. Spanish financial institutions offered the lowest mortgage rates of the Euro area. In fact during the period 2003–06 the average mortgage rate in the Euro area was 4.51 while the average in Spain was 3.71. The average Euro area mortgage rate was 21 % higher than the Spanish counterpart. Given this small loan spread in Spain, the competition took place through massive origination of mortgage loan principals. This competitive pressure implied that managers of financial institutions could only increase profits drastically by originating a large number of new mortgages. The economies of different countries have been affected with different degrees of intensity according to their exposure to some of the main drivers of the financial crisis. The excessive dependence of the real estate industry, jointly with a softening of the credit standards, caused that economic and financial crisis hit Spain more severely than to other developed economies. In this context of economic recession in Spain, one of the most controversial issues is the Spanish Mortgage Law, which was approved in 1909, and with a small number of posterior modifications, still applies now to the Spanish mortgage market. Footnote 9 continued and other European countries. These countries required sponsors to hold the same amount of regulatory capital for assets on balance sheets and for assets in ABCP conduits. Acharya and Schnabl (2010) find that the regulatory capital arbitrage motive for the creation of the shadow banking system was crucial, and thus they find that Spanish and Portuguese banks did not sponsor ABCP conduits. 10 The statistics of the Bank of Spain report 1.1 million new mortgages for home acquisition in 2005 and 1 million in 2006. The INE (Spanish Statistical Office) reports almost identical numbers once you subtract mortgage modifications. 123 SERIEs (2014) 5:223–243 229 The Spanish mortgages are loans with full-recourse in contrast with the limited recourse mortgages loans in the most states of the US. While in those states the guarantee is the dwelling, this is not so in Spain. In the event of a mortgage foreclosure, the lender seizes mortgagor’s dwelling, which is sold at auction at a price generally quite bellow its market price. After that, borrowers still hold a debt consisting in the outstanding mortgage debt minus the auction prize of the dwelling plus the interests for late payment, which are generally quite high. Mortgage foreclosure usually takes a long period (an even more in crisis times);11 therefore, these interests consist in the accumulation of the monthly payment for the long period priced at an interest rate of 15–20 %. If mortgagors do not hold other real estate properties or businesses that can be included as guarantee in the mortgage contract, lenders will force borrowers to include another person, usually a relative, as a guarantor. The guarantor is responsible of the mortgage monthly payments in the event that borrowers cannot meet these payments. If the guarantor has no earnings or income to take responsibility of these mortgage monthly payments, lenders can seize also guarantor’s properties. To what extent could banks lower credit standards? The regulation of the Bank of Spain establishes the weights assigned to each type of mortgage credit following the international regulation (Basel). Low risk mortgages (those with a loan to value lower than 80 %) have a lower weight that mortgages with loan to value above that threshold. This means that they require less capital. In addition, banks report to the Bank of Spain their mortgages classified in three groups (below 80, 80– 100 and above 100 %). Obviously, having many mortgages above 100 % could trigger some action from the Bank of Spain. In addition, one of the main ways that a financial institution can reduce its minimum amount of capital is to reduce its risk weighted assets. In this sense, financial institutions are encouraged to increase its percentage of mortgages with a loan to value lower than 80 %, and over-appraising is an effective way of doing so. In the summer of 2007, credit conditions and the real estate market started changing. The Bank of Spain survey on lending conditions and standards show that the tightening of lending conditions came in Spain, as in the whole Euro Area, in the summer of 2007 (see Maddaloni and Peydró 2011). Bank liquidity problems massively increased in August 2007: the ECB injected more than 90 billion dollars on the 9th of August 2007, interbank market spreads massively increased, and bank liquidity problems lead to a tightening of lending conditions to households and firms.12 In addition, the reduction in transactions in the housing market and new house constructions began in 2007, although the effect on official prices (based on appraisals) was not clear until 2008 because of the particular dynamics of that type of prices. However, the price index of INE (Spanish Statistical Office) began going down in September of 2007, the prices of the Official Association of Notaries on July of 2007, and the index of Fotocasa-IESE (ask prices) started going down in June of 2007. All in all, the bust in credit and real estate conditions started in the summer of 2007. 11 The length depends on the cyclical situation of the housing market. While in 2005 the process could take eight months, in 2010 the process until repossession extended much longer. 12 For the Spanish data, see http://www.bde.es/webbde/en/estadis/infoest/epb.html and for the Euro area data, see http://www.ecb.europa.eu/stats/money/surveys/lend/html/index.en.html. 123 230 SERIEs (2014) 5:223–243 Table 1 Comparison of the sample with other sources (second semester of 2012) Growth rate comparison Price level comparison Transaction price Appraisal price Our sample INE (registered Our sample Department of price) public works (%) (%) (e/m2)(e/m2) Barcelona −21.64 −17.4 2,569 3,103 L’Hospitalet de Llobregat −31.00 −17.4 1,949 1,647 Madrid −17.77 −18.5 2,326 2,459 Málaga −17.74 −14.6 1,404 1,416 Sevilla −17.34 −14.6 1,643 1,715 Valencia −21.03 −15.4 1,187 1,317 Zaragoza −19.59 −16.5 1,626 1,517 Total −19.42 −16.4 2,072 2,141 Sources: INE (Spanish Statistical Office), Department of Public Works and proprietary data Growth rate: 2011:Q4–2012:Q4 3 Data description We use a unique data set obtained from a housing market intermediary with franchisers in most of the Spanish provinces. This real estate company also possesses its own mortgage brokerage branch. For instance, this company made 6,528 sales in 2012 which was 4 % of the total sales in Spain during that year. Our data is not strictly representative of all the universe of houses sold during the Spanish bubble period. The intermediary that provided the information is not uniformly represented in Spain. It has more branches in large cities and metropolitan areas around large cities. In addition, our sample does not cover the whole distribution of house prices. We missed the upper part of the distribution. For instance, in the city of Barcelona there are no observations on Pedralbes, which is the neighbourhood with the highest house prices. However, this does not seem to affect the average prices. Table 1shows the comparison of the appraisal prices of our dataset with the appraisal prices obtained from the Department of Public Works (DPW),13 for the cities where the housing market intermediary has a very large sample. It corresponds to the second semester of 2012. Appraisal prices are the only variable that we can compare with a population variable (in fact the data of the DPW is not the population of appraisals but it is quite close). The table shows very small deviation in appraisal prices between our sample and the population of appraisals that compiles the DPW. The difference is only 3.2 % for the average of these cities. Therefore, and making clear that we are not claiming that our sample is fully representative of the population of all the transacted properties of the years under study, we believe there are not reasons to expect that the difference would be much larger in other places not included in the table (except for sampling variability). Note that the price statistics of the DPW do not include either 13 “Precios de viviende libre de los municipios mayores de 25000 de habitantes”, Ministerio de Fomento. 123 SERIEs (2014) 5:223–243 237 0 5 10 15 20 25 Percent 0 20 40 60 80 100 120 140 loan to appraisal value (%) Fig. 1 Loan principal to appraisal value two modes correspond to the regulatory thresholds mentioned above.17 Moreover, this bimodal profile is not observed when we look at the distribution of the loan to transaction price ratio. These two pieces of information, over-appraisal and bimodal profile, are giving the evidence of the manipulation on appraisal values that lead to the real estate price and credit bubble. In order to analyze whether the over-appraising mechanism described in this point is related to some characteristics of either the borrower or the loan, we estimate a reduced form equation (using the same variables we used in the previous subsection) of the determinants of the upward bias in appraisal prices, which can be defined as the ratio of the appraisal price and the market price of the dwelling. Results for both the whole sample and boom period are reported in Table 4.Comments are based on the estimation using the subsample (second column of Table 4) given that there are no relevant differences in both sets of estimates. The estimates provided reveal that most of the variables behave accordingly to expectations. With respect to the individual characteristics we expect that upward bias in appraisal prices will be negatively correlated with income, age, university degree and permanent contract.18 All these cases are supposed to have less financial constraints. In those cases, it was more likely that appraisal prices get close to market prices than in the rest of the cases. Otherwise, be a university graduate reduces the upward bias in 4.01 percentage points with respect to primary studies in the boom period. Likewise, being older decreases the upward bias. Finally, a higher real income increases upward bias (from 1.028 points for every 1,000 e). A higher income could be capturing unobserv- 17 This bimodal profile is not observed when we look at the distribution of the loan to transaction price ratio. 18 We estimated the model by controlling also for the dwelling characteristics and the results did not change significantly. 123 238 SERIEs (2014) 5:223–243 Table 4 Estimation results for the appraisal to transaction ratio (ATratio) ATratio (%) Whole Boom Type of contract (ref: permanent) Temporary 2.190** (0.977) 2.595*** (0.919) Labor status (ref: private sector employee or self-employed) Non-occupied −0.907 (2.849) −3.276 (3.818) Public sector employee 1.069 (1.653) 0.466 (2.162) Income 1.021** (0.427) 1.028** (0.432) Marital status (ref: married) Separate, single, widow −1.877 (1.197) −2.259 (1.371) Education (ref: compulsory) Secondary (non compulsory) −3.743*** (1.457) -4.484*** (1.632) University degree −4.181*** (1.492) −4.010*** (1.437) Age −0.615 (0.443) −0.844* (0.491) Age20.005 (0.005) 0.008 (0.006) Number of holders (ref: one) Two −1.375 (1.538) −2.237 (1.648) Three −2.384 (2.147) −2.262 (2.5) Spread 6.844*** (1.226) 5.701*** (1.520) Reference interest rate (ref: RIML) Euribor −6.576*** (1.535) −5.350*** (1.878) Financial institutions (ref: commercial) Non-bank financial institution −10.138*** (1.773) −10.526*** (1.535) Individually rescued 2.092 (1.569) 3.705* (2.111) Owned by FROB −9.561*** (1.319) −10.244*** (1.325) Savings bank −7.658*** (1.839) −10.923*** (1.911) Location (ref: interior) Coastal −2.118 (3.730) −1.489 (4.207) Madrid dummy Yes Yes Year dummies Yes Yes Observations 3,305 2,852 R-squared 0.119 0.121 Constant is included but its coefficient is left unreported. Robust standard errors are in parentheses *,**,*** Indicate statistical significance at the 1, 5 and 10 % level, respectively able characteristics such as higher financial skills, higher future income or higher financial knowledge that makes lenders trust or, even, to provide incentives for overappraising. As expected, a higher spread is correlated with a higher ratio among appraisal and transaction prices. Thus, we have found evidence of higher interest rates in terms of compensation for the risk assumed with a loan with a higher over-appraising (either for increase the total amount of mortgage or to reduce the loan to value to an stan- 123 SERIEs (2014) 5:223–243 239 dard value). In particular, an increase of 1 percentage point in differential results in an increase of 5.7 percentage points in the ratio between appraisal and transaction prices. In addition, when Euribor is the benchmark interest rate, the ratio between appraisal and transaction prices decreases in 5.4 percentage points. This result brings evidence that lenders tends to use RIML for riskier mortgages. By construction, the RIML is not only higher but also less volatile and react less intensively to changes in the market interest rates than the Euribor. This circumstance makes the RIML more advantageous for lenders, especially for riskier borrowers. As a result, we use this variable to capture unobserved characteristics of the riskier borrowers. In this sense we expect negative sign for the Euribor variable. Finally, financial institutions have had different behavior towards risk during the boom period. In this specification we found evidence for a higher upward bias (of 3.7 percentage points) in individually rescued institutions and a smaller bias for nonbank financial companies (10.5 percentage points), FROB owned institutions (10.2 percentage points) and the rest of the savings banks (10.9 percentage points) with respect to commercial banks. As pointed out previously, from this result we cannot infer that commercial banks behavior was riskier than savings and loans and non-bank financial companies. From the models estimated previously we can infer that, for the boom period, non-bank financial companies, FROB owned institutions and individually rescued companies dealt with riskier mortgages, in the sense of higher loan to value, than commercial banks and the rest of savings banks. In the case of individually rescued institutions, not only loans were riskier but we also find a higher upward bias than for other financial institutions (except for commercial banks). 5 Conclusions We analyze the determinants of real estate and credit bubbles using a unique dataset on mortgage loans in Spain. The dataset contain real estate credit and price conditions (loan principal and spread, and the appraisal and market price) at the mortgage level, matched with borrower characteristics (such as income, labor status and contract) and the lender identity, over the last credit boom and bust. We find that lending standards are softer in the boom than in the bust. Moreover, despite some adjustment in lending conditions in the good times depending on borrower risk, the results suggest too soft lending standards and excessive risk-taking in the boom. Banks with worse corporate governance problems (rescued cajas) soften even more the lending standards. Finally, we analyze the mechanism by which banks could increase the supply of mortgage loans despite of regulatory restrictions on LTVs. The evidence is consistent with banks encouraging real estate appraisal firms to introduce an upward bias in appraisal prices (29 %), to meet loan-to-value regulatory thresholds (40 % of mortgages are just bunched on these limits), thus building-up the credit and the real estate bubble. 123 240 SERIEs (2014) 5:223–243 All in all, the results suggest that the credit and housing bubble were partly driven by bank agency problems (Freixas and Rochet 2008) and not simply by behavioral motives (Gennaioli et al. 2012). This has crucial implications for public policy, in particular microprudential and macroprudential policy (Freixas, Laeven and Peydró, forthcoming). First, our results are consistent with the new prudential policy measures on limiting the ex-ante banks’ incentives to take on excessive risk-taking. Second, the results show that micro loan level data are crucial to identify—ex-ante on real time—excessive credit supply booms. For example, that borrowers who do not work pay identical loan spreads than workers, or that 40 % of LTVs are bunched in the regulatory limits (while, at the same time, appraisal values are substantially higher than transaction prices). Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited. Appendix: Descriptive statistics See Tables A.1,A.2. Table A.1 Characteristics of the loan Mortgage sample Matched sample Boom Bust Whole sample Boom Amount of the loan (e)a189,957 (82,697) 168,202 (76,403) 171,211 (70,513) 174,13 (700,64) Appraisal value (e)a231,054 (86,567) 216,366 (84,753) 195,214 (69,813) 196,318 (68,276) Loan to value (%) 82.52 (19.94) 78.37 (20.08) 87.56 (18.64) 88.56 (18.3) Spread (%) 0.86 (0.40) 0.88 (0.46) 0.85 (0.43) 0.84 (0.43) Appraisal to transaction price (%) 128.93 (22.25) 128.18 (21.84) Transaction price (e)a156,005 (60,073) 157,782 (59,877) Reference interest rate RIML (ref) 17.02 12.47 15.16 16.73 Euribor 82.98 87.53 84.84 83.27 Financial institution Commercial bank 39.19 30.67 39.61 40.25 Savings bank 50.35 48.82 51.62 51.71 Individually rescuedb9.35 7.41 8.56 8.87 FROB owned 29.98 18.27 30.65 32.92 Rest 11.02 23.14 12.41 9.92 Non-bank financial institutions 10.45 20.51 8.77 8.03 123 SERIEs (2014) 5:223–243 241 Table A.1 continued Mortgage sample Matched sample Boom Bust Whole sample Boom Years 2005 29.14 52.95 61.36 2006 50.03 31.29 36.26 2007 20.83 59.07 4.57 2.38 2008 8.98 2.51 2009 16.13 4.6 2010 15.82 4.08 Number of observations 25,041 5,221 3,305 2,852 We present mean and standard deviation (in brackets) for quantitative variables and the distribution in percentage for qualitative variables aVariables are in real terms bCommercial bank “Banco de Valencia” is included in this group Table A.2 Characteristics of the borrower Mortgage sample Matched sample Boom Bust Whole sample Boom Labor status Public sector employee 9.21 15.36 86.63 88.57 Private sector employee or self-employed 88.47 78.17 10.32 9.29 Non-occupied 2.32 6.47 3.06 2.14 Type of contract Permanent 61.43 69.05 62.17 59.82 Temporary 36.26 24.48 34.77 38.04 Marital status Married 36.02 32.64 31.04 32.5 Separate 5.41 7.41 Single 57.69 58.72 68.96 67.5 Widow 0.89 1.23 Education Compulsory 57.59 46.03 54.52 57.47 Secondary (non compulsory) 31.56 39.8 33.4 31.45 University degree 10.84 14.17 12.07 11.08 Number of holders One 28.25 39.13 31.71 28.51 Two 60.14 53.46 57 59.29 Three 11.61 7.41 11.01 12.03 Location Interior 55.33 52.83 57.7 57.26 Coastal 44.67 47.17 42.93 42.74 123 242 SERIEs (2014) 5:223–243 Table A.2 continued Mortgage sample Matched sample Boom Bust Whole sample Boom Region Comunidad De Madrid 25.48 22.47 27.62 26.51 Income (thousande)a1,601 (0.72) 1,627 (0.69) 1,563 (0.65) 1,557 (0.65) Age 33.78 (9.31) 35.8 (9.97) 33.77 (9.18) 33.43 (8.98) Number of observations 25,041 5,221 3,305 2,852 We present mean and standard deviation (in brackets) for quantitative variables and the distribution in percentage for qualitative variables aVariables are in real terms References Acharya V, Schnabl P (2010) Do global banks spread global imbalances? ABCP during the financial crisis of 2007–09. IMF Econ Rev 58:37–73 Ben-David I (2011) Financial constraints and inflated home prices during the real-estate boom. Am Econ J Appl Econ 3(3):55–78 Bernanke BS (1983) Nonmonetary effects of the financial crisis in the propagation of the great depression. Am Econ Rev 73(3):257–276 Besley T, Meads N, Surico P (2013) Risk heterogeneity and credit supply: evidence from the mortgage market. NBER chapters. In: NBER macroeconomics annual 2012, volume 27 National Bureau of Economic Research Inc. Bover O, Casado JM, Costa S, Du Caju P, McCarthy Y, Sjerminska E, Tzamourani P, Villanueva E, Zavadil T (2014) The distribution of debt across euro area countries: the role of individual characteristics, institutions and credit conditions, ECB working paper 1639 Cecchetti SG, Mohanty M, Zampolli F (2011) The real effects of debt. BIS working papers no 352 Cunat V, Garicano L (2010) Did good cajas extend bad loans? The role of governance and human capital in Cajas’ portfolio decisions. In: Bentolila S, Boldrin M, Diaz-Gimenez J, Dolado JJ (eds) la Crisis De la Economía Española: Análisis Económico De la Gran Recesión. Colección monografías fedea. Fedea, Madrid, pp 351–398 Fannie M (2005) Uniform residential appraisal report. Form 1004 Fannie M (2007) Guide to underwriting with DU Freixas X, Laeven L, Peydró J-L (forthcoming) Systemic risk, crises and macroprudential policy. MIT Press, Cambridge Freixas X, Rochet J (2008) Microeconomics of banking. MIT Press, Cambridge Garcia-Montalvo J (2009) Financiación inmobiliaria, burbuja crediticia y crisis financiera: lecciones a partir de la recesión 2008–2009. Papeles de la Economía Española vol 122, pp 66–87 Garcia-Montalvo J, Raya J (2012) What’s the value of Spanish real estate? Span Econ Finan Outlook 4:22–28 Gennaioli N, Shleifer A, Vishny R (2012) Neglected risks, financial innovation, and financial fragility. J Finan Econ 104(3):452–468 Gourinchas PO, Obstfeld M (2012) Stories of the twentieth century for the twenty-first. Am Econ J Macroecon 4(1):226–265 Jiménez G, Ongena S, Peydró J-L, Saurina J (2012) Credit supply and monetary policy: identifying the bank balance-sheet channel with loan applications. Am Econ Rev 102(5):2301–2326 Jiménez G, Ongena S, Peydró J-L, Saurina J (2013) Hazardous times for monetary policy: what do twentythree million bank loans say about the effects of monetary policy on credit risk-taking? Econometrica Jiménez G, Ongena S, Peydró J-L, Saurina J (2013) Macroprudential policy, countercyclical bank capital buffers and credit supply: evidence from the spanish dynamic provisioning experiments. Barcelona GSE working paper no 1315 123 SERIEs (2014) 5:223–243 243 Keys B, Mukherjee T, Seru A, Vig V (2010) Did securitization lead to lax screening? Evidence from subprime loans 2001–2006. Quart J Econ 125(1):307–362 Kindleberger CP (1978) Manias, panics, and crashes: a history of financial crises. Basic Books, New York LaCour-Little M, Malpezzi S (2003) Appraisal quality and residential mortgage default: evidence from Alaska. J Real Estate Finan Econ 27(2):211–233 Lang W, Nakamura L (1993) A model of redlining. J Urban Econ 33(2):223–234 Maddaloni A, Peydró J-L (2011) Bank risk-taking, securitization, supervision, and low interest rates: evidence from the euro area and U.S. lending standards. Rev Finan Stud 24(6):2121–2165 Mian A, Sufi A (2009) The consequences of mortgage credit expansion: evidence from the 2007 mortgage default crisis. Quart J Econ 124(4):1449–1496 Myers S (1977) Determinants of corporate borrowing. J Finan Econ 5(2):147–175 Nakamura L (2010) How much is that home really worth? Appraisal bias and house-price uncertainty. Bus Rev Q1:11–22 Peydró J-L (2012) Comment on Risk heterogeneity and credit supply: evidence from the mortgage market. NBER chapters. In: NBER macroeconomics annual 2012, volume 27 National Bureau of Economic Research Inc. Reinhart CM, Rogoff KS (2009) This time is different: eight centuries of financial folly. Princeton University Press, Princeton Schularick M, Taylor AM (2012) Credit booms gone bust: monetary policy, leverage cycles and financial crises, 1870–2008. Am Econ Rev 102(2):1029–1061 123