BIS Working Papers
Whatever it takes. What’s the impact of a major nonconventional monetary policy intervention?
by Carlo Alcaraz, Stijn Claessens, Gabriel Cuadra, David Marques-Ibanez and Horacio Sapriza
Monetary and Economic Department
JEL classification: demography, ageing, inflation, monetary policy
Keywords: E31, E52, J11
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2017. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISSN 1682-7678 (online)
Whatever it takes. What’s the impact of a major nonconventional monetary policy intervention?
by Carlo Alcaraz, Stijn Claessens, Gabriel Cuadra, David Marques-Ibanez and Horacio Sapriza
We assess how a major, unconventional central bank intervention, Draghi’s “whatever it takes” speech, affected lending conditions. Similar to other large interventions, it responded to adverse financial and macroeconomic developments that also influenced the supply and demand for credit. We avoid such endogeneity concerns by comparing credit granted and its conditions by individual banks to the same borrower in a third country. We show that the intervention reversed prior risk-taking – in volume, price, and risk ratings – by subsidiaries of euro area banks relative to other local and foreign banks. Our results document a new effect of interventions and are robust along many dimensions.
Keywords: Unconventional monetary policy; credit conditions; spillovers
JEL classification: E51; F34; G21
Historically, a leading reason for the emergence of many central banks has been the need to setup institutions that are able to provide extraordinary amounts of liquidity to halt a systemic crisis during periods of financial stress (Goodhart, 1988, Capie, Goodhart, and Schnadt, 1994; see Bindseil, 2018 for a recent review). These interventions involve trade-offs. By providing liquidity to certain institutions, a central bank can stem individual bank runs and prevent failures related to frictions in funding and capital markets. By ensuring the availability of ample liquidity to the overall banking system, a central bank can prevent contagion during periods of high stress and buttress the stability of the overall financial system. Furthermore, support from the central bank can provide financial authorities the time to arrange an orderly resolution of troubled banks. However, liquidity provision can also lead to unintended consequences, such as "risk-shifting", and introduce other distortions, including moral hazard, where banks take excessive risks at the expense of debt holders (Jensen and Meckling, 1976).1 Understanding the exact impact of large scale injections of liquidity on banks, and the economy more generally, is therefore a key question. Our paper sheds light on this by studying how a large, unconventional, central bank intervention affected lending in a third country through the operations of foreign banks residing in that third country.
During the recent financial crises, most major central banks have undertaken unparalleled large-scale interventions to restore financial stability, reestablish financial intermediation, and support the transmission of monetary policy. These actions have involved a dramatic lowering of policy interest rates as well as the adoption of nonconventional monetary policy measures, such as the easing of refinancing operations.2 The largest nonconventional interventions, notably in the United States and Europe, have often involved outright asset purchases, including the direct acquisitions of government and private sector securities. These and other interventions have been credited with containing financial stress, stabilizing the financial system, and, eventually, restarting the real economy (eg, Goodhart, 2014; Financial Crisis Inquiry Report, 2011). At the same time, a number of authors have emphasized and analyzed the distortions that can be triggered by these often massive interventions (BIS, 2014; Acharya and Steffen. 2015; Rajan, 2017; Drechsler, Drechsel, Marques-Ibanez and Schnabl, 2016). The diverse implications of these actions and the potential trade-offs involved make it crucial to understand and accurately quantify their effects. Yet, several challenges must be overcome in order to provide compelling evidence on the effects of these interventions.
The major hurdle for this type of analysis is that it is very difficult to disentangle the effects of such interventions from the factors that led to them in the first place. Extreme financial distress and a worsening macroeconomic outlook are typically among the reasons leading to major liquidity injections and other central bank interventions. However, the presence of these factors and other concurrent forces make it very hard to isolate the impact of the interventions. A more practical, but equally important challenge for conducting impact assessments is the lack of granular data necessary to identify the precise channel through which the intervention affects the banking system and the economy.
To overcome the main hurdle, we study how a large, unconventional central bank intervention in Europe influenced credit conditions of various banks in a third country. Specifically, we focus on European Central Bank's president Mario Draghi's "whatever it takes" speech (WIT henceforth) and analyze its effect on lending standards of the subsidiaries of euro area banks in Mexico.3 We hypothesize that prior to WIT, euro area banks, which suffered heavy losses on their U.S. and European operations and saw their capital depleted, would have had greater incentives to increase risk-taking compared to other institutions (Dewatripont and Tirole, 1994; Holmstrom and Tirole, 1997).4 We then expect that by augmenting their net worth, the WIT speech restored banks' incentives to be safe, which, in turn, should be reflected in reduced risk-taking incentives afterwards.
The increase in banks' net worth induced by WIT likely worked via two related mechanisms. First, by solving a multiple equilibrium problem, WIT prevented a full- scale liquidity crisis, as it dramatically reduced the incentives of debt holders to run from euro area banks (Goldstein and Pauzner, 2005).5 The decline in liquidity risk and uncertainty contributed to strengthen banks' capital position and help preserve their franchise value. Second, WIT also supported the euro area's banks' capital by eliciting a general increase in the prices of the assets held by banks.
Our setting allows us to properly identify the effects of WIT for three reasons. First, we focus on one of the largest central bank interventions ever undertaken, meaning the action was quantitatively sufficiently important that it could be expected to have a significant economic impact. The WIT speech was a response to a severe deterioration in financial and economic conditions in 2011 and early 2012 in the euro area. As the region sank deeper into a sovereign debt crisis and faced a deep recession, banking institutions and sovereign yields came under unprecedented financial stress, especially in peripheral euro area countries. Through WIT, the ECB implicitly promised to provide unlimited support to financial institutions, markets and countries in order to save the Euro. Besides its large scale, the unexpected nature of the announcement helps us to measure its impact, as it ensures that banks and borrowers did not adjust their actions in advance. The surprise associated with the WIT speech was reflected in financial markets dynamics, leading to large effects on the stock prices and credit spreads of banks and sovereigns. After the speech, euro area banks saw their stock prices increase and their spreads on credit default swaps (CDS) narrow (Figures 1 and 2; see also Szczerbowicz, 2015). Sovereign and corporate bond yields of many peripheral euro area countries declined drastically, settling at much lower levels (Figure 3; see also Krishnamurthy et al, 2018).
Second, Mexico's financial system provides a very good setting to disentangle the effect of a central bank intervention on lending conditions from other factors contemporaneously affecting borrowers. This is because at the time of the speech, Mexico's banking system and economy were relatively unaffected by the financial crisis in Europe other than via euro area banks operating in the country. Indeed, around the time of WIT, Mexico did not experience any major bout of financial instability in its banking sector, nor did its central bank need to conduct any major intervention (Bank of Mexico, 2012). Mexico's economy was also largely disconnected from the source of the shock, ie, the euro area.6 Therefore, there are few concerns about other shocks affecting the Mexican banking system at the time besides the events in the euro area. Altogether, this setting - akin to Peek and Rosengren (1997, 2002) - is ideal because it minimizes the risk of reverse feedbacks which can contaminate the empirical results.
Concerns about any other events complicating identification are further addressed by the characteristics of Mexico's banking system. It is a banking system with an important presence of foreign banks, which hold a market share of more than 70%, and include not only European but also other foreign banks (Table 1). Specifically, the foreign banks present in Mexico have their headquarters in Europe, the United Kingdom, Canada and the United States. With these banks exposed differentially to shocks and official policy actions, including those in the euro area, we expect to identify the effects of WIT.
Finally, Mexican authorities collect detailed information on lending relationships between banks and borrowers, which allows us to cleanly measure the effects of any shock on Mexican borrowers. Specifically, we use the Mexican Central Credit Bureau which stores comprehensive data on the universe of bank loans, and we follow the methodology of Khwaja and Mian (2008). We isolate loan supply shocks by studying the differences in lending conditions to the same borrower between WIT (ie euro area) and non-WIT (ie non-euro area) affected banks to multi-bank firms, ie, those firms borrowing from more than one bank including at least of bank from the euro area and another from outside the euro area. This allows us to assess the effect of the announcement on bank lending conditions operating via banks' lending supply (in terms of prices, quantities, and risk ratings) fully controlling for changes in loan demand and borrowers' risk.
We show that following the WIT speech, euro area banks became less aggressive in their lending volumes, interest rate pricing, and risk-taking. While prior to the intervention euro area banks had been pricing their loans more aggressively than other institutions, after WIT they reverted their lending standards back in line with other banks. Our results are confirmed by a number of robustness tests, including winsorizing and changes in time windows, as well as a number of placebo tests.
Our findings show that major central bank liquidity interventions can contribute significantly to altering banks' lending conditions. The results also suggest, however, that even with strict regulation and supervision in place, covering aspects such as changes in (risk) exposures and limits on internal transfers of capital and liquidity between headquarters and subsidiaries, there can be spillovers from headquarters to subsidiaries (and presumably vice-versa). The likely reason is that a bank's risk attitude operates across the whole institution. Thus, when its franchise value is under pressure, say due to worsening circumstances in its home market, the bank aims to preserve its overall profitability by engaging more in risk-taking globally. As this risk-taking might be difficult to detect in real-time, our results suggest the need to pre-emptively adapt cross-border regulation, supervision and resolution of global banks to make these measures well integrated, so as to assure not only that risks do not go undetected, but also that incentives remain well-aligned.
The remainder of this paper is organized as follows. Section 2 reviews related literature on major central bank interventions, including an analysis on the WIT speech. Section 3 describes the institutional setting in Mexico and the dataset employed, while Section 4 provides the empirical model we use. Section 5 presents the main empirical findings and our robustness tests. Section 6 considers some general implications and concludes.
The analytical literature on major central bank interventions has a long history, starting with Thornton (1802). These and other very early contributions already justify why in times of stress, central banks' provision of liquidity to solvent banks, freely and against good collateral at a penalty rate, can help avert bank runs and prevent contagion. More recent theoretical work rigorously shows that banks' inherently unstable nature can give creditors the incentive to run, even on solvent banks: therefore, a liquidity crisis can quickly become a solvency crisis (Diamond and Dybvig, 1983).7 The literature also stresses that inaction on the side of the central bank can lead to contagion, potentially endangering systemic stability (Allen and Gale, 2000, and Freixas et al, 2000). By providing liquidity, a central bank can prevent bank runs from happening, so that the better equilibrium prevails and stability is restored.
A key concern about major central bank liquidity operations has been moral hazard (Bagehot, 1873). The idea is that the provision of liquidity and capital support - as with any type of insurance - modifies the incentives for banks to take preventive actions going forward, thereby increasing the probability of them experiencing stress in the future. Recent work departs from Bagehot's prescriptions and the classical moral hazard literature, by emphasizing that limiting the intervention to lending at a penalty rate to solvent institutions might not be welfare improving.8 These models, which incorporate both liquidity and solvency shocks, conclude that in cases of severe financial stress the central bank should offer loans at lower rates than the market (Rochet and Vives, 2004). It follows that Bagehot's classical principles would have failed in the context of a systemic event, like the 2007-2009 crisis. Instead, under these conditions, lending generously to all banks against risky assets (in practice banks' illiquid portfolios) at favorable rates is argued to be preferable (Freixas, 2009).
There is a paucity of empirical studies on the effects of central bank interventions on financial institutions' behavior, and even fewer on their impact on borrowers. Some of the reasons are straightforward. First, one needs a major systemic crisis combined with a large central bank intervention to potentially observe a meaningful impact. Second, the empirical analysis requires access to very detailed data on the interventions by the central bank complemented with granular information on banks' and matched borrowers' conditions. Such data, however, are not available for many countries and when they do exist, they tend to be highly confidential.
As such, most studies use aggregate data, making conclusions largely qualitative. Some recent exceptions, where more detailed confidential data from national credit registers are used, show that central bank liquidity injections - such as the extension of liquidity operations - are effective in restoring bank credit supply in the thick of the crisis, particularly for financially constrained banks (Carpinelli and Crosignani, 2017; Andrade, Cahn, Fraisse and Mesonnier, 2018; Alves, Bonfim and Soares, 2016). Other recent work focuses on the unintended consequences of these large liquidity operations either in the United States (Acharya, Fleming, Hrung and Sarkar, 2017) or the euro area (Drechsler et al, 2016) and finds that weaker banks (ie, those with greater leverage) are more likely to borrow more aggressively and in larger amounts from the central bank's liquidity facilities, suggesting that large central bank interventions can induce risk-shifting.
Some studies specifically analyze the WIT's announcement through its impact on financial asset prices. They find a positive outcome (or "bright side") resulting from the intervention, as they show that WIT led to a significant and long-lasting decline in sovereign bond yields, particularly in the euro area periphery (Krishnamurthy et al, 2018). WIT also triggered strong improvements in banks' stock market prices and declines in their CDS spreads (Fiordelisi and Ricci, 2016). Closely related to our work is Acharya, Eisert, Eufinger and Hirsch (2017) who try to quantify the impact of the intervention on banks' lending conditions using data (more limited in coverage) from the euro area syndicated loan market. They find that WIT contributed to lending to insolvent (or zombie) firms at the expense of solvent borrowers, thereby contending that the unintended effects (or "darker side") of the intervention won out.
At the same time, these papers show the challenges mentioned earlier in terms of analyzing the impact of events such as WIT. First, interventions typically happen amid extreme financial stress with many confounding factors occurring simultaneously, including, inter alia, political uncertainty, financial instability, as well as shifts in economic expectations and activity. Thus, any analysis in the same economic region where the intervention takes place has a hard time isolating the effect of the event from these other factors occurring contemporaneously. Another limitation is the lack of data availability. For instance, the commonly used way to account for credit supply (Khwaja and Mian, 2008) is very demanding, as it requires observations on credit (conditions) for the same borrower with at least two banks before and after the intervention. Using public data from the syndicated loan market, often an alternative, restricts the analysis of shifts in loan supply and demand to a small sample, as there are typically only a very limited number of cases where two different loan syndicates lend to the same borrower in the months before and after an event like WIT.
Besides aiming to overcome these endogeneity and data problems, our study contributes to the literature on cross-border banking spillovers (Ongena and Goldberg, 2015). Starting with the seminal papers by Peek and Rosengren (1997, 2000), this literature suggests that distress at the parent (home) country (bank) affects financing conditions and the real economy abroad (host country) via lending standards by foreign subsidiaries. Plenty of evidence, including from recent crises, also shows that global banks transmit liquidity and monetary policy shocks internationally (Cetorelli and Goldberg, 2011, 2012; Buch, Koch, Koetter, 2016; Morais et al, 2016, and Buch et al, 2018). At the same time, domestic banks exhibit more stable lending patterns than foreign banks (Claessens and van Horen, 2014; De Haas, van Lelyveld, 2014).
Altogether, this literature suggests that shocks, including central bank interventions, tend to have a significant impact abroad through global banks. In contrast to our analysis, these papers tend to focus on either the cross-border activities of banks or on the intra-bank (or inter-banking systems) transfers of liquidity or capital in response to shocks. In contrast, in our analysis, spillovers occur because of changes in the riskiness of opportunities at home versus abroad, but without any direct financial transactions between the affected banks and their international operations, including their subsidiaries. As such, we analyze a new channel of spillovers related to relative shifts in franchise values and capitalization at home versus abroad.
In this section we explain why Mexico is a good "laboratory" for testing the effects of a large central bank intervention around the period of the WIT announcement. We do this by reviewing some key institutional features of the Mexican banking system that are relevant for our setting, and by detailing the data used in our analysis.
3.1 Institutional Framework: Mexico
The Mexican banking sector is quite concentrated and has an important presence of European banks coexisting with other foreign institutions and domestic banks. As shown in Table 1, during the 2011 - 2013 period, the seven largest banks represented almost 80 percent of the assets of the Mexican banking system. Of these banks, five were foreign and two domestic. Among the foreign banks, two were from Spain, one from Canada, one from the United Kingdom, and one from the United States. The subsidiaries of these global banks held more than 70 percent of the total assets of the Mexican banking sector. This wide distribution in terms of geographical areas of the ultimate bank ownership, allows us to isolate the impact of a foreign shock - the WIT announcement - on the Mexican banking system.
The importance of foreign banks varies across market segments. Figure 4 shows the share of credit in four categories (non-financial corporations, mortgage, consumer, and government loans) both for the largest ("G7") and all other banks ("non-G7"), with the categories further split by domestic and foreign ownership. The figure shows that large and small banks lend in similar proportions to the corporate sector. In contrast, banks differ significantly in their relative shares of mortgage and consumer lending, with foreign banks much more focused on these types of lending. In order to reduce any biases related to lending specialization connected to the shock, and to ensure comparability across banks, we restrict our analysis to the lending to nonfinancial corporations.
The Mexican banking system was well-capitalized during our period of analysis. Table 2 shows that in June 2012 - just before the WIT event - the capitalization ratio of the banking sector was around 16 percent, mostly in the form of high-quality capital, substantially above the then prevailing regulatory requirement of 8 percent. The table also shows that among the largest banks in the system, both euro area subsidiaries and non-euro area institutions were well-capitalized. During that period, Mexican banks already satisfied the liquidity requirements of Basel III, with little dispersion among banks of similar size, as depicted in Figure 5.13 In short, domestic banks and subsidiaries of foreign banks were all well-capitalized and had satisfactory liquidity buffers at the time of WIT.
In general, and especially during our period of study, the operations of the euro area banks both at home and in Mexico were closely supervised. This meant that there was very limited scope for intra-bank transfers of liquidity and capital. As such, spillovers between the euro area banks and their subsidiaries could have arisen because headquarters altered relative incentives for risk taking due to shifts in perceived growth opportunities at home versus abroad, but not because of headquarters moving financial resources between home and abroad. Indeed data from the Bank of International Settlements show little change in the claims of banks operating in Mexico with their Spanish counterparts around the WIT speech.
We combine several datasets to conduct our analysis. Our main dataset consists of loan-level information on the credit granted by banks operating in Mexico to Mexican corporate borrowers from July 2009 to December 2013. This database is managed by the Banking Supervisor in Mexico (Comision Nacional Bancaria y de Valores or CNBV) and is available at a monthly frequency. It covers the majority of Mexican bank loans, as the cut-off for reporting is very low, at 5,000 Mexican pesos (about 373 US dollars at the time). We rely on the so-called form R04-C that all banks operating in Mexico are legally required to report to the CNBV, and which contains detailed information on all bank loans granted in Mexico. This loan level information is matched with another database containing the financial statements of all banks. For comparability purposes, we restrict our sample to commercial banks and drop other financial institutions, such as development banks or those institutions belonging to other parts of the financial system. We also focus on the largest banks to ensure a similar pool of borrowers with significant overlap across banks both geographically and across sectors.
Table 4 shows the distribution of all loans across the various types of credit for each of the largest banks operating in Mexico. The table also shows that the majority of loans (about 55%) are so-called "single disposition." It also shows that interest rates are largely similar across loan types, except for syndicated loans, which are much larger and often denominated in US dollars. We therefore focus on "single disposition" loans, but we corroborate our results by also including other types of loans and the totality of Mexican banks. We always drop syndicated loans from our sample, since they do not depend solely on the characteristics of a single bank, and we do not include revolving and multiple non-revolving loans, as they both allow the borrower to have access to multiple loans on pre-arranged conditions.
We include both "straight" as well as "contingent" loans (such as bridge loans). We distinguish between new and previously granted loans. Unlike most national credit registries, the Mexican data covers extensive information on lending rates, including, inter alia, the reference rate, the spread over the reference rate, and the frequency of repricing. In terms of credit quality, we know whether the loan is nonperforming, and if this is the case, the number of days overdue, and its so-called "qualification," ie, an internal risk rating provided by the supervisor.
The database covers loans to companies as well as to physical persons with entrepreneurial activities. We focus our analysis on loans to non-financial companies. The database includes firm-specific information such as borrower's location (state and municipality), main sector of activity, as well as firm's age, size (number of employees) and gross income in previous year.
Finally, we obtain monthly data from Markit and Haver on credit default swap (CDS) spreads and equity prices for banks operating in Mexico during the period of analysis, where in the case of foreign banks the data reflect the assessment of the consolidated operations of the banks and not only of their subsidiaries in Mexico. The list of all the variables used in the study and the corresponding data sources are detailed in Table 3.
In order to isolate the effects of WIT, we proceed in a number of steps. First, we document its impact on the stock prices of banks, in particular of foreign banks operating in Mexico. Figure 1 shows the evolution of stock prices in the years surrounding the WIT speech (with shading for the June-August period around the speech) of the two largest euro area banks operating in Mexico (BBVA and Santander) and compares them to that of HSBC - another European bank operating in Mexico but with headquarters in the United Kingdom, ie, outside the euro area. The figure starkly illustrates how the stock prices of these three banks evolved similarly during 2011, also in line with the broad US index (S&P 500). In the first half of 2012 - ie, the months just before the WIT - however, stock price developments saw a large divergence between euro area and other foreign banks. Euro area banks suffered a further decline in prices of around 30 percent, while HSBC experienced a sustained improvement. After WIT, the downward trend of euro area banks reversed, and their prices came back in sync with those of their peers. In short, the announcement by Draghi in July of 2012 proved to be a turning point for European bank stock prices.
Bank CDS spreads, shown in Figure 2, also highlight that the shocks varied across institutions, with euro area banks being hit harder before Draghi's announcement and benefitting more after it. Figure 3 suggests that the shocks came in large part via changes in sovereign risks, with yields of the peripheral euro area countries increasing markedly and reaching historical heights before the July 2012 announcement, as shown for Spain. Following WIT (marked again together with the previous and following month by the shaded area), Spanish bond yields declined substantially. In contrast, Mexican and US sovereign yields exhibited a declining trend through most of the period.
We next assess the effect of the central bank intervention on lending conditions by banks operating in Mexico. We first study changes in the overall growth rates of bank credit and interest rates. Simple comparisons of aggregate lending by all euro area and all other banks operating in Mexico and t-tests already suggest that WIT led to statistically significant differences in lending conditions between euro area and non-euro banks comparing before and after the event.16 Specifically, we find that the difference in the average monthly growth rate of credit to all Mexican firms between euro area and non-euro area banks decreased significantly, and that the average interest rate charged was some 20 basis points higher for euro area banks relative to non-euro area banks after the event compared to before.
In order to assess causality, our main estimating technique is the difference-indifference (diff-in-diff) methodology procedure of Khwaja and Mian (2008).17 Following this setup, we only include borrowers with lending relationships with several banks, including at least one euro area and one non-euro area bank. This allows us to capture loan supply shocks by fully accounting for any shift in demand conditions or change in borrower's risk. In other words, to be included in our main sample, a firm would need to borrow from at least one bank affected by WIT and from another bank not affected by the shock. We initially apply this methodology using a panel structure, so, in its most demanding form, the specification is as follows:
where Aloanjt is the log difference in general credit (single disposition loans) to nonfinancial firm i by bankj at time t; treatedbankj is a dummy variable that equals 1 if the bank is "treated" by WIT (1 for all euro area banks, and 0 for all other banks); treatmentperiodt a dummy variable that equals 1 in the months immediately after the WIT announcement (August to November 2012); bankj is a bank fixed effect (to account for bank-specific characteristics); iocationk is the geographical location of the firm (to account for local economic conditions); firmi is the firm fixed effect (accounting for borrower characteristics); and periodt is a monthly time dummy (to account for macroeconomic and regulatory factors). Crucially, the firmpperiodt dummy variable controls for firm demand and risk in a given month. We are primarily interested in the coefficient which captures the (differential) effect of WIT on the affected banks, with the null-hypothesis that as the WIT speech increased the franchise value of the euro banks home operations, it reduced the incentives of these banks to engage in risk taking in their Mexican subsidiaries.
We also run similar regressions for the level of interest rates charged, interestjt. For both sets of regressions, we start with a limited set of fixed effects that we then expand progressively to the full set of fixed effects. This means that in the presentation of our regressions, we proceed to specification (1.a) above by progressively adding fixed effects to the equation below: 18
These regressions are run over different windows of time before and after the WIT event to analyze how the behavior of affected banks changes over time. The main period of study extends from March to November 2012. Since Mario Draghi's speech was made on July 26th, the month of July itself is excluded from the sample. The base period is thus March to June 2012, while the treatment period runs from August to November 2012, so it incorporates four months following the event, a period during which banks had sufficient time to change their lending conditions. In our robustness tests, however, we use several alternative periods.
While we focus on changes in lending (conditions) to multi-bank firms, we also consider changes in credit conditions on bank loans to aii firms. Therefore, this estimation includes firms with multiple and single banking lending relationships.19 The estimation is similar to (1.a) except that we cannot use firm*period fixed effects to control for changes in demand at the firm level. Yet, we can use firm fixed effects to control for firm-specific characteristics. In this specification, we also include location*period and industry dummy variables to control for local economic and industry-specific conditions. We thus use the following specification for these regressions:
While specification (2) is less able to disentangle supply effects from demand shocks and is therefore less rigorous than specification (1), it helps assessing the effects of WIT also on credit to new firms, therefore capturing the extensive margin of credit. Finally, in order to further capture the impact of WIT on risk-taking, we build on (1.a) and (2) and run an additional specification in which we use borrowers' credit risk as the left-hand side variable of study, instead of loan amounts or lending rates, keeping the rest of the identification unchanged.
5.1 Main Regression Results
The main results on the effect of WIT on bank lending are presented in Table 5, where the classification variable "Euro" applies to euro area banks, and the treatment effect dummy variable "Post" refers to the period following the WIT speech. Column (1) shows the specification without any fixed effects. Column (2) adds location-specific fixed effects to account for local economic conditions which might impact loan demand. Column (3) has fixed effects accounting for firm-specific characteristics (dropping the location fixed effects). Column (4) includes both firm and location fixed effects; and in column (5) adds firm*treatment fixed effects. Finally, column (6) shows the results with firm*period fixed effects, our most stringent base estimation.
The results show that, compared to other institutions, euro area banks decreased their lending following the WIT speech, as suggested by the consistently negative and highly statistically significant euro area*post variable. In terms of economic magnitude, the reduction is very large, between 0.79 and 0.63 percentage points per month, a very considerable slowdown in growth rates. To put this amount in perspective, this represents about half of the average growth rate in credit for aii banks in the period before the event (January to May 2012). Our findings remain robust to the inclusion of various sets of fixed effects accounting for demand and other shocks affecting firms differently. Thus, these findings provide strong evidence that the WIT speech substantially reduced the incentives of euro area banks to engage in more aggressive lending.
We next conduct a similar set of regressions, but considering now the level of the interest rate charged by banks. Table 6 reports these findings, which are again consistent across specifications. We document that euro area banks adjusted upwards their lending rates following WIT, as the coefficient of the euro area*post dummy interaction variable is positive and highly statistically significant in all six columns. As for lending volumes, the economic impact of this change in euro area banks' behavior is considerable, as the rates go up by between 20 and 29 basis points.
The results imply that, compared to other banks, euro-area banks tightened their lending standards to the same borrowers after WIT.
The analysis so far was limited to multi-bank firms only. This allows us to control well for changes in demand and other firms' characteristics, but a drawback of this specification is that we leave out large amounts of lending, including loans to firms who receive credit from one bank only. This choice could affect our results. For instance, a bank may change lending conditions to firms that borrow exclusively from that bank, and not to firms that also borrow from other banks (multi-bank firms). This can be due to private information linked to relationship-based lending, which tends to be more prevalent for single-bank firms (Degryse and Ongena, 2005).21 In short, banks could respond differently to policy interventions in their lending to single-bank firms compared to their lending to multi-bank firms.
Therefore, we conduct a similar set of regressions, but now including all firms, ie, the specification (2) that considers loans to both multi and single-bank firms. Note that compared to specification (1.a) we need to drop the specification used in column 6 of Tables 5 and 6, as we cannot include firm*period fixed effects, and rather use location fixed effects to account for local economic conditions.
The results for all firms are reported in Table 7 for lending volumes and in Table 8 for interest rates. As with the sample of multi-bank firms, we report results for the differences between euro area and other banks. This analysis also finds that the monthly loan growth of euro area banks is lower relative to other banks after the policy intervention, in this case by about 0.8 percentage point, with results consistent across all five specifications. While weaker, the effects on the extensive margin are very comparable to the ones for multi-bank firms. Note that we show the window with the most conservative results, as all other windows consistently show stronger and more significant coefficients. As such, we can be confident that the effects we find reflect the overall adjustments by banks to the shocks, instead of adjustments in specific lending strategies. In terms of interest rates, Table 8 illustrates that the effects for all loans are also broadly in line with previous results, as euro area banks adjusted their lending rates relatively upwards after the policy event, between 13 and 26 basis points, depending on the specification. Thus, these results show that there is also a significant adjustment in the interest rate for single-bank firms.
We also consider borrowers' credit risk, where we investigate if changes in lending standards were linked to banks actively altering the risk profile of their borrowers. We adopt a similar approach as in previous specifications - (1.a) for multibank and (2) for all borrowers - but now use a credit score as our dependent variable (in which a higher credit score indicates lower risk). This score uses the supervisory agency's estimate of a firm's credit risk, which is derived from a logit model, with the score assigned to each borrower at each point in time.22
To proxy a firm's default, two measures are used: short-term default, defined as either nonpayment or partial payment of the minimum amount required by the bank during the last 3 months; and long-term default, defined as a default registered during the last consecutive 12 months. To estimate the likelihood of such defaults, the credit score includes a measure of credit exposure - the past amount of unpaid loans as a share of total loans, the usage of credit lines, the total amount of the credit line used by the firm in the current period, and regional and bank fixed effects. Using only firms that have sufficient data available, the credit risk score is then estimated using logit regressions for each month on a rolling basis, controlling for heteroscedasticity of errors. The behavior of these out-of-sample credit scores are then analyzed.
We undertake the usual diff-in-diff regressions using this variable. As measures of firm credit risk tend to change gradually over time, we extend the periods of comparison: from January to June 2012 (before) and from August 2012 to January 2013 (after). Beyond that adjustment, the regression specification remains the same and explains whether there were statistically significant changes in the estimated credit risk scores before versus after the WIT speech.
The regression results are presented in Table 9 for multibank firms and in Table 10 for all firms. They show that there was an improvement in the credit riskiness of firms lent to by euro area banks compared to other banks following the WIT intervention, with the effect present in both sets of regressions, ie, for multibank firms as well as the full sample of firms. This suggests that following WIT, euro area banks chose to lend to a safer credit portfolio relative to other banks, so WIT seems to have improved their incentives to maintain a prudent risk exposure.
We also ran all the regressions (settings (1.a) and (2)) in panel formats that include bank*period fixed effects and drop the treatment variable (WIT). As before, we did this exercise for our two main dependent variables: loan growth and interest rates. These sets of regressions capture the dynamics of bank lending, by analyzing whether at any time a bank's lending conditions deviated significantly from other banks. That is, in this setting we do not test explicitly for the effect of WIT by comparing treated (euro area) and non-treated (non-euro area) banks on the basis of a single coefficient (WIT), but we run a panel that includes dummy variables for each bank at each point in time.
Consistent with previous findings, this method (not reported) also shows that euro area banks became more prudent after the WIT. They suggest that, in the runup to the WIT, as financial distress in the euro area was reaching its peak, euro area banks loosened their lending standards compared to other institutions. The deterioration was stronger for the bank that saw its stock market capitalization decline more, reinforcing the idea that large declines in banks' capital positions are linked to higher risk taking by banks and that, by contributing to augment banks' capital positions, WIT reversed the increase in risk taking by providing banks better incentive to tighten their lending conditions.
Interestingly, our results suggest that an increase in bank stocks of euro area parent banks meant lower lending abroad. This finding apparently contrasts those obtained by Peek and Rosengren (1997) for Japanese bank subsidiaries in the United States in the early 1990s. However, the two findings likely reflect the widely different responses by the Bank of Japan and the ECB when confronted with a severe bank crisis. While the ECB swiftly provided large amounts liquidity to the banking sector following the default of Lehman Brothers in 2008 (as reflected by the sharp expansion of its balance sheet over time), the Bank of Japan kept its balance sheet fairly constant until late into the Asian crisis.
We next conduct a number of robustness tests, including alternative groupings of treated and non-treated banks, different time windows, and additional econometric exercises. We use our most stringent specification, ie, (1.a) with firm*time fixed effects.
Alternative groupings. To assess the effects of the ECB intervention, so far we have focused on the differences between the euro area (treated) and all other banks (control), with the latter group amalgamating both domestic Mexican banks and noneuro area foreign banks. Yet, it can be argued that the right control group should be domestic banks (ie, Mexican banks) and should not include the non-euro area foreign banks, as the latter are typically global institutions that might (indirectly) also be affected by WIT.
Therefore, we replicate our baseline regression - specification (1.a) - now using two new control groups, one including Mexican banks only, and another limited to non-euro area foreign banks. Apart from the new groupings of treated and control groups, the estimation is the same as our estimations with firm*period fixed effects including only multibank firms (columns 4 and 6 of Tables 5 and 6). We again analyze the impact of WIT on both loan growth rates and interest rates.
We summarize these tests in Table 11, where we report only the treatment*period coefficients, showing different combinations of the treated (columns) and control groups of banks (rows). Panel A reports the results for loan amounts and Panel B for interest rates. The different control groups confirm our main results, in that following WIT, euro area banks lent significantly less and at considerably higher lending rates to the same borrower than domestic Mexican banks, as shown by the statistically significant coefficients of around -0.7 and 0.2, respectively. While the signs and magnitude of these results are consistent with our previous findings when including all non-euro banks as a control, the effect of WIT is not significant for the control group including only non-Mexican foreign banks (last column of panels A and B), presumably because foreign banks operating in Mexico operate globally and might have been affected by WIT, although indirectly and to a far lesser extent than euro area banks.
When splitting among the sample of euro area banks (in rows 2 and 3 using all non-euro banks as controls), we find that the coefficients on lending (Panel A) remain consistently negative and strongly statistically significant for both euro area banks at about 0.8 percentage points, with the WIT leading to a somewhat stronger restriction in the loan supply of euro bank 2. We do not find large differences across euro area banks regarding the impact of WIT on lending rates (Panel B). The panel shows that WIT resulted in higher lending rates by both euro area banks compared to Mexican
banks. Reassuringly, this increase in lending rates is not significant when the set of other (ie, non-euro area) foreign banks is used as the control group. Note that since these regressions consider lending to the same firm, the effects are not due to differences in the (pool of) firms' risk.
We further compare key characteristics of these euro area banks at the time. Both of them were heavily exposed to extreme funding stress due to the euro area crisis, so we review the relative amount of liquidity support that each of these banks received from the European Central Bank (ECB). As the crisis intensified after the default of Lehman Brothers in 2008 and banks' funding markets became more strained in the euro area, particularly for banks in the periphery, the ECB provided banks unlimited liquidity against good collateral.26 Evidence suggests that weakly capitalized banks took out more liquidity and provided riskier collateral than better capitalized banks (Drechsler et al, 2016). Analysis of confidential data on all liquidity operations granted by the ECB to the largest euro area banks shows that both euro area banks operating in Mexico were among the top 25 percent of banks requesting liquidity (as a percentage of total assets).27 Hence, the data support the idea that these banks had relatively high liquidity demands prior to WIT due to strains in funding markets.
Aggregate data on liquidity granted by the ECB shows that following WIT there was a massive reduction in liquidity requested by euro area banks, further supporting the view that WIT dramatically alleviated liquidity constraints in funding markets, in line with the prediction of Goldstein and Pauzner (2005). A more detailed analysis focusing on both euro area banks operating in Mexico shows that the bank that reduced its lending more sharply as a result of WIT (euro bank 2) had greater liquidity needs prior to WIT compared to euro bank 1, as revealed by the respective liquidity requests to the central bank.
Other targe euro area banks. As illustrated in Figure 6, stock market data for other large European banks point to a similar pattern to that of the two euro area banks in our sample, showing a large decline in the months prior to the WIT speech followed by a sharp rebound afterwards. This suggests that the incentives applying to the banks in our sample were common to other large euro area institutions.
Using the same setup, we also ran a similar set of regressions but now including lending to all firms, ie, without restricting our sample to those firms borrowing from more than one bank. The results - not shown - are qualitatively similar to those for the multi-bank firms. Following WIT there is a relative decline in loan growth by euro area banks who also tightened their interest rates more than other institutions. The results also show a similar pattern to that of multi bank borrowers, in that the tightening following WIT was more pronounced for the euro bank that experienced a larger improvement in capitalization following the WIT.
Windows of study. Another set of robustness tests alters the window of time used to assess the policy event. It may take longer or shorter for the WIT to influence bank behavior than the window of time we used in our previous specifications. Since we have no strong priors other than informal conversations with supervisors and some previous literature on monetary policy (de Bondt, 2005) as to what is the correct time span is to capture the lag over which banks adjust their lending standards to such shocks, we apply a number of different start and end dates. Tables 12 and 13 report the results comparing euro area banks with all other banks. In Table 12 the specification includes location and firm fixed effects for all firms and multi-bank firms. Table 13 uses again our most stringent regression for multi-bank firms using firm*period fixed effects. The tables only report the coefficients for the treatment*period dummies. In terms of loan growth, results are only marginally affected by the time window. If anything, effects become stronger as the window of time expands. Comparisons are similar for the sample of all firms (Panel B) and when using the specification with firm*period fixed effects (Table 13).
In terms of interest rates, it also appears that shifts in the time window do not change our main conclusion supporting that euro area banks adjusted their interest rate upward after the event. Again, these regression results hold when using other combinations of treatment and control groups, when using all firms, and when using the most rigorous specification in Table 13. As such, our findings consistently suggest that there was a major shift in lending behavior around the time of the WIT speech.
The Tables also show that trimming observations at the 2 and 98 percentiles or at the 5 and 95 percentiles does not change the main results for loan growth rates or interest rates. The size of the coefficients becomes slightly smaller when winsorizing is performed at the 5 and 95th percentile, but the statistical significance largely remains.
We analyze how a major unconventional central bank intervention, Draghi's WIT speech, impacts lending conditions. To avoid endogeneity problems - where local demand shocks and macroeconomic risks are hard to insulate from the effects of the intervention - we conduct our analysis using data for Mexico and assess how banks from the euro area, which are directly affected by WIT, changed their lending conditions in Mexico, a financial system with diverse banking ownership. Our detailed data from a third country allow us to control for shifts in borrowers' demand and riskiness. Comparing local lending conditions to the same borrower with banks largely sheltered from the WIT shock, we show that the intervention significantly reduced incentives for risk shifting among euro area banks.
Our findings show that major central bank liquidity interventions can contribute significantly to altering banks' lending conditions. Our results complement those analyzing international spillovers, such as Peek and Rosengren (1997) for the case of Japan, and many others since. They suggest that, even with strict regulation and supervision covering aspects such as changes in (risk) exposures and limits on internal transfers of capital and liquidity between headquarters and subsidiaries, there can be spillovers from headquarters to subsidiaries (and vice-versa). The likely reason is that a bank's risk attitude operates across the whole institution, so that when the franchise value is under pressure the bank aims to preserve its overall profitability by engaging more in risk-taking globally. As this risk-taking might be difficult to detect in realtime, our results suggest the need to pre-emptively adapt cross-border regulation, supervision and resolution of global banks to make them well integrated, so as to assure not only that risks do not go undetected, but also that incentives remain well- aligned.
Acharya, V, T Eisert, C Eufinger, and C Hirsch (2017): "Whatever it takes: The real effects of unconventional monetary policy", SAFE Working Paper Series 152.
Acharya, V, M J Fleming, W B Hrung and A Sarkar (2017): "Dealer financial conditions and lender-of-last-resort facilities", Journai of Financiai Economics 123, 1 81-107.
Acharya, V and S Steffen (2015): "The "greatest" carry trade ever?: Understanding eurozone bank risks", Journai of Financiai Economics 115 (2), 215-236.
Albertazzi, U and M Bottero (2013): "Foreign bank lending: Evidence from the global financial crisis", NBER internationai Seminar on Macroeconomics 2013, 22-35.
Allen, F and D Gale (2000): "Financial contagion", Journai of Poiiticai Economy 108 (1), 1-33.
Alves, N, D Bonfim and C Soares (2016): "Surviving the perfect storm: the role of the lender of last resort", Banco de Portugai, Working Papers 17.
Andrade, P, C Cahn, H Fraisse and J S Mesonnier (2018): "Can the provision of longterm liquidity help to avoid a credit crunch? Evidence from the Eurosystem's 3-year LTROs", Journai of the European Economic Association, forthcoming.
Bagehot, W (1873): "Lombard Street, A Description of the Money Market", Henry S. King & Co, London.
Bank for International Settlements (2014): "Re-thinking the lender of last resort", BiS Papers 79.
Bank of Mexico (2012): "Reporte sobre el Sistema Financiero", September.
Bindseil, U (2018): "Pre-1800 central bank operations and the origins of central banking", ECB, Frankfurt mimeo.
Buch, C, M Bussiere, L Goldberg and R Hills (2018): "The international transmission of monetary policy", NBER Working Paper 24454.
Buch, C M and L S Goldberg (2015): "International banking and liquidity risk transmission: Lessons from across countries", iMF Economic Review, 63(3) 377-410.
Capie, F, C Goodhart, S Fisher and N Schnadt (1994) "The Future of Central Banking", Cambridge University Press.
Carpinelli, L and M Crosignani (2017): "The effect of central bank liquidity injections on bank credit supply", Board of Governors of the Federai Reserve System Finance and Economics Discussion Series 2017-038.
Cetorelli, N and L S Goldberg (2011): "Global banks and international shock transmission: Evidence from the crisis", iMF Economic Review 59(1), 41-76.
Cetorelli, N and L S Goldberg (2012): "Follow the money: Quantifying domestic effects of foreign bank shocks in the great recession", American Economic Review 102(3), 213-18.
Corsetti, G, B Guimaraes and N Roubini (2006): "International lending of last resort and moral hazard: A model of IMF's catalytic finance", Journal of Monetary Economics 53(3), 441-471.
De Bondt, G J (2005): "Interest rate pass-through: Empirical results for the euro area", German Economic Review 6(1), 37-78.
Degryse, H and S Ongena (2015): "Distance, lending relationships, and competition", Journal of Finance 60, 1, 231-266.
De Haas, R and N Van Horen (2013): "Running for the exit? International bank lending during a financial crisis", Review of Financial Studies 26(1), 244-285.
Dewatripont, M and J Tirole (1994): "The Prudential Regulation of Banks", Walras Pareto lectures, MIT press.
Diamond, D W and P H Dybvig (1983): "Bank runs, deposit insurance and liquidity", Journal of Political Economy 91, 401-419.
Diamond, D W and R G Rajan (2011): "Fear of fire sales, illiquidity seeking, and credit freezes", Quarterly Journal of Economics 126, 557-591.
Domanski, D, R Moessner and W Nelson (2014): "Central banks as lenders of last resort: experiences during the 2007-10 crisis and lessons for the future", Bank for International Settlements Working Paper 79.
Eser, F, M Carmona, S Iacobelli, S and M Rubens (2012): "The use of the Eurosystem's monetary policy instruments and operational framework since 2009", European Central Bank Occasional Paper Series 135.
Financial Crisis Inquiry Report (2011) Final Report of the National Commission on the Causes of the Financial and Economic Crisis in the United States.
Fiordelisi, F, and O Ricci (2016): "Whatever it takes: An empirical assessment of the value of policy actions in banking", Review of Finance, 20, 6, 2321-2347.
Freixas, X, B Parigi and J Rochet (2000): "Systemic risk, interbank relations, and liquidity provision by the central bank", Journal of Money, Credit and Banking 32 (3), 611-38.
Freixas, X, B Parigi and J C Rochet (2004): "The lender of last resort: A 21st century approach", Journal of the European Economic Association 2, 1085-1115.
Freixas, X (2009): "Monetary policy in a systemic crisis", Oxford Review of Economic Policy 25, 4, 630-653.
Goldstein, I and A Pauzner (2005): "Demand-deposit contracts and the probability of bank runs" Journal of Finance 60, 1293-1327.
Goodhart, C (1988): "The Evolution of Central Banks", MIT Press.
Goodhart, C and G Illing (2002): "Financial Crises, Contagion, and the Lender of Last Resort: A Reader", Oxford University Press.
Goodhart, C A E (2014): "Lessons for monetary policy from the Euro-area crisis", Journal of Macroeconomics 39(B), 378-382.
Haas, R, and I Lelyveld (2014): "Multinational banks and the global financial crisis: Weathering the perfect storm?", Journal of Money, Credit and Banking 46 (s1), 333364.
He, Z, A Krishnamurthy and K Milbradt (2018): "Intermediary asset pricing and the financial crisis", Annual Review of Financial Economics, forthcoming.
Holmstrom, B and J Tirole (1997): "Financial Intermediation, loanable funds, and the real sector", The Quarterly Journal of Economics 112, 3, 663-691.
Jensen, C and H Meckling (1976): "Theory of the firm: managerial behavior, agency costs and ownership structure", Journal of Financial Economics 3, 305-360.
Jimenez, G, S Ongena, J L Peydro and J Saurina (2012): "Credit supply and monetary policy: Identifying the bank balance-sheet channel with loan applications", American Economic Review 102(5), 2301-2326.
Khwaja, A I and A Mian (2008): "Tracing the impact of bank liquidity shocks: Evidence from an emerging market", American Economic Review 98(4), 1413-42.
Klein, M W, J Peek and E S Rosengren (2002): "Troubled banks, impaired foreign direct investment: The role of relative access to credit", American Economic Review 92(3), 664-682.
Krishnamurthy, A, S Nagel, A Vissing-Jorgensen (2018): "ECB policies involving government bond purchases: Impact and channels", Review of Finance 22(1), 1-44.
Martin, A (2006): "Liquidity provision vs. deposit insurance: preventing bank panics without moral hazard", Economic Theory 28, 197-211.
Morais, B, J L Peydro, and C Ruiz Ortega (2015): "The international bank lending channel of monetary policy rates and quantitative easing: credit supply, reach-for- yield, and real effects", Journal of Finance, forthcoming.
Peek, J, and E S Rosengren (1997): "The international transmission of financial shocks: The case of Japan", American Economic Review 87(4), 495-505.
Rosengren, E S, and J Peek (2000): "Collateral damage: Effects of the Japanese bank crisis on real activity in the United States", American Economic Review 90(1), 30-45.
Peek, J, and E S Rosengren (2005): "Unnatural selection: Perverse incentives and the misallocation of credit in Japan", American Economic Review 95(4), 1144-1166.
Popov, A, and G F Udell (2012): "Cross-border banking, credit access and the financial crisis", Journal of International Economics 87, 147-161.
Rajan, R G (2017): "Was unconventional monetary policy a success?" Keynote address at the 2017 Asian Monetary Policy Forum.
Rochet, J C and X Vives (2004): "Coordination failures and the lender of last resort: Was Bagehot right after all?" Journal of the European Economic Association 2, 11161147.
Schnabl, P (2012): "The international transmission of bank liquidity shocks: Evidence from an emerging market", Journal of Finance 67(3), 897-932.
Szczerbowicz, U (2015): "The ECB unconventional monetary policies: have they lowered Market borrowing costs for banks and governments?", International Journal of Central Banking 11(4), 91-127.
Thornton, H (1802): "An Enquiry into the Nature and Effects of the Paper Credit of Great Britain", Nabu Press, London.