The International Bank Lending Channel of Monetary Policy Rates and QE: Credit Supply, Reach-for-Yield, and Real Effects

Whether foreign policy moves foreign banks’ local credit supply; whether real effects follow or firms substitute away; whether easing creates an international risk-taking channel through reach-for-yield; whether rates and QE differ; and whether foreign and local banks fund their credit expansion differently. The list is explicit: “(a) whether foreign monetary policy affects the supply of credit from foreign banks to local firms; (b) whether there are real effects associated with foreign monetary policy shocks, including firm investment, employment and survival, or whether local firms are able to reduce these shocks by substituting credit with local banks or with other sources of finance; (c) whether an expansive foreign monetary policy creates an international risk-taking channel by affecting global banks’ reach-for-yield incentives; (d) whether these effects depend on the type of monetary policy used, i.e., policy rates versus QE; and finally, (e) whether foreign and local banks finance differently their local credit expansion” (Section 1). The policy framing comes from named officials: Rey’s (2013) Jackson Hole argument that Fed policy “may have substantial spillovers in emerging markets’ credit cycles, thereby generating an international risk-taking channel of monetary policy,” Fischer’s (2014) warning that both rates and QE matter and that European policy matters too because European banks are strongly globalised, and Rajan’s (2014) alerts from the Reserve Bank of India. The motivating stylised fact is that “strong bank credit growth, especially financed by foreign liabilities, is the most important predictor of financial crises.”

The obstacle is the absence of a credit registry matched to firm and bank data over enough years, especially in an emerging market; Mexico supplies exactly that, plus a foreign banking sector large enough to carry the shock. “Despite the importance of these questions for policy and macro-finance, the identification of foreign monetary policies on the credit and risk-taking channel by foreign banks has been elusive. This has been due to the lack of exploitation of comprehensive credit registry data, matched with firm and bank information, especially in emerging markets, with enough years to analyze monetary policy” (Section 1). Three features of the register matter. It has no minimum loan size, unlike Germany’s 1.5 million euro or Italy’s 75,000 euro thresholds, which elsewhere means “loans to small and even medium size firms may not be included in those credit registers” (fn. 1). It reports loan rates. And foreign ownership is large: US and European banks’ Mexican subsidiaries extend 58 percent of all commercial bank credit. The exogeneity advantage over domestic studies is stated directly: “different from papers that analyze local monetary policy on local credit conditions, we examine European and U.S. monetary policies, which are exogenous to the Mexican economy.” The paper also generalises the setting rather than treating Mexico as sui generis, noting foreign banks hold roughly half the market in emerging and developing countries (Claessens and van Horen, 2012).

All commercial-firm loans monthly with terms and arrears, plus firm balance sheets and bank balance sheets – with two deliberate exclusions. For each loan the paper observes “the issuing bank, the borrower (firm), the outstanding amount, the (annualized) interest rate, both start and ending date of the loan (maturity), the fraction covered by collateral, as well as certain firm information, such as its location, and industry,” tracked monthly so that underperformance is observed by amount and duration (Section 2). Loans to individual entrepreneurs (“personas fisicas con actividad empresarial”) are dropped because “banks may change the classification of loans to individuals with entrepreneurial activity from commercial to consumption loans and vice versa, artificially moving the number of commercial loans in our data,” and because the firm balance-sheet data cover only commercial firms (fn. 13). Banks specialising in consumer lending and niche banking are also excluded, comprising “less than 3 percent of the assets in corporate bank lending” (fn. 12). The loan distribution is heavily skewed: average credit volume is MXN 2,244,000, roughly USD 172,000, while the median loan is close to USD 30,000; the median loan rate is 15 percent and median maturity 36 months; the median loan is uncollateralised though average collateral is 26 percent; and the average default rate – the fraction of loans in arrears more than 90 days – is 7 percent with a median of zero. About 97 percent of loans are peso-denominated, and foreign-currency loans “does not alter our results in any significant way” (fn. 14).

Policy rates are residualised against foreign activity to proxy Taylor-rule shocks, QE is central bank balance sheet growth, and the US rate is additionally instrumented. On rates: “we take the residual of the regression of the policy rate of a country on its GDP growth and inflation (thus proxying a Taylor rule-type shock).” Beyond that, “apart from controlling for global shocks via month fixed effects, we also control for foreign economic activity in interactions with our main variables, as this could be a separate channel of influence, including current and expected annual GDP growth and inflation, as well as a measure of financial risk” (fn. 2). The instrumental-variables check is reported separately: the fed funds rate is instrumented “following the instruments suggested in Gertler and Karadi, 2015 for the period June 2001-November 2009, which correspond to the period in which the Fed Funds rate had not reached the zero lower bound,” and “the results using the instrument of the U.S. policy rate are consistent with our findings” (fn. 30). QE is the change in the central bank’s balance sheet as a share of GDP, with the caveat that the instrument differs across central banks because the ECB’s main non-standard policy was liquidity provision rather than asset purchases.

Three nested layers plus a within-period cross-lender comparison, with an explicit argument that the strongest layer does not change the answer. Firmbank effects absorb time-invariant firm and bank heterogeneity and the relationship itself. Adding stateindustryperiod effects removes time-varying borrower shocks. Firmmonth effects go furthest, comparing “loans offered by different banks to the same firm in the same month,” but require multiple simultaneous bank relationships, which only 21 percent of firms have, so “we lose more than half of the observations and some coefficients lose statistical significance” (Sections 1 and 3A). The paper separates sample selection from omitted variables by re-running the milder specification on the same multi-bank firms: “the coefficients that drop by half in column 4 do so because of the sample selection towards larger firms.” The Altonji-style argument is then made explicitly: “in column 4 the estimated coefficients are not statistically different from those of column 3 despite a substantial increase of the R2 (around 43 percentage points). This suggests that our main coefficients on credit supply shocks (foreign banks and monetary policy) are exogenous to unobserved demand proxied by (firm*month) time-varying firm unobservables and observables.” A further identifying point is that “as period fixed effects control for unobserved global shocks, identification also comes in a given month from the differential of monetary policies between Mexico, U.S., U.K., and the Eurozone.”

Each foreign policy moves mainly its own banks’ lending, while the domestic policy rate moves everyone’s – and the paper’s wording is “mainly” rather than “only.” “The three different foreign monetary policy rates affect more strongly credit outcomes of banks from the same country… As for the non-standard monetary policies, the QE of the U.S., U.K. and the Eurozone affects more the credit volume of firms whose loans are from U.S., U.K. or Eurozone banks respectively. In contrast, the Mexican policy rate affects the credit volume of all banks, regardless of their nationality” (Section 3A). The domestic-rate asymmetry gets an explanation: “similar to domestically owned banks, foreign subsidiaries in Mexico have substantial local retail deposits and are therefore also affected by local monetary policy” (fn. 31). Magnitudes differ markedly across source countries: “a 1 standard deviation decrease in the Fed Funds rate raises the average loan volume of U.S. banks in Mexico by 6 percent, and a 1 standard deviation decrease in the monetary policy from U.K. (Eurozone) expands credit by an average of 4.8 (2) percent.” The paper is careful that these are not comparable units without qualification, noting “the standard deviations of interest rates and also of QE are different across the different countries’ monetary policies” (fn. 32).

All of them, with rates moving least and collateral moving the other way – for which the paper offers a valuation explanation rather than a compensation story. “On average, a 1 standard deviation reduction in foreign monetary policy translates into loans not only of larger volume, but also of longer-term maturity, and, for U.S. banks, in lower loan interest rates” (Section 3A). On collateral: “a softening of foreign monetary policy increases collateral by 5.7 percent, which could be due to higher valuation of the collateralized assets when policy is softer. Our main results hold when we control for changes in collateral; that is, our results are robust to controlling for the collateral as right hind side value” (fn. 34). The composite loan-level figures are volume up 2.1 percent, maturity up 6.7 percent and future default up 9.8 percent per one-standard-deviation easing. The paper’s own summary is that “all loan terms are significantly affected, reinforcing the supply driven channel; interestingly, though, the effects are weaker for loan rates” – a pattern it links to the theoretical literature holding that “banks may adjust more volumes than rates in lending, see Stiglitz and Weiss (1981)” (fn. 11).

Rates dominate, and the paper offers a specific reason QE may be understated rather than simply weak. “For US monetary policy, which have the largest economic effects, whereas a one standard deviation decrease in Fed Funds rate expands credit volume of U.S. banks by 6 percent and maturity by 9.9 percent, a one standard deviation increase in QE raises volume by only 2.5 percent and maturity by 7.1 percent” (Section 1); the results section gives the US-specific QE figures as “2.6 percent larger in volume, 7.8 percent lengthier, with no change on collateral and 0.3 percentage points lower [rate].” Euro-area QE is the weakest link – insignificant once state and industry time-varying controls enter – and the paper turns that into a testable proposition rather than leaving it unexplained: because QE periods coincide with high risk, “it is possible that the results, especially on QE, may be biased towards zero given the positive correlation between the QE measures and various measures of financial risk. To test for this possibility, we interact our QE measure with sovereign CDS of that same country… We find that indeed QE results are stronger, the lower the CDS of the sovereign where the foreign bank is headquartered in. This may explain why for some results, elasticities from QE are lower than interest rates ones, especially from the Eurozone banks given the Eurozone crisis” (Section 3B). The implication is stated as a general claim about the composition of global liquidity: “when the Federal Reserve, the ECB and the Bank of England expand their balance sheet via nonstandard monetary policies, the U.S. and European banks expand less into Mexico, the higher the risk in the countries where their [headquarters are].”

With a lag, peaking between six and twelve months and fading after twelve to fifteen. Classifying banks as domestic or foreign and regressing credit margins on lags from 3 to 24 months, “when the benchmark one-quarter lagged foreign monetary policy rate declines by 1 standard deviation, loans from foreign banks increase their volume by 1.5 percent, lengthen their maturity by 4.8 percent and increase collateral values by 4.6 percent. Second, the impact of lagged foreign monetary policy on credit supply is somewhat persistent within a range of 3 and 12 months, and declines after 12 months” (Section 3A). The margins differ: “Loan volume, collateral, rates and loan rates coefficients have the maximum absolute value at the 12 month lags, while maturity and the default rate are relatively less persistent. Regarding the coefficients on QE, we also find that they increase for higher lags, especially for volume and defaults.” The summary is that “the speed of transmission of both types of monetary policy shocks becomes weaker after 12 to 15 months, there is some heterogeneity in the speed of transmission across loan margins in the first 12 months, but effects are generally strongest between 6 and 12 months for both monetary rates and QE.” The paper counts this lag analysis among its contributions, noting it contributes “by analyzing the lags of the transmission of monetary policy on loan and firm outcomes” (fn. 11).

Yes, through the banks whose home policy eased – and the paper attributes it to composition rather than to loans deteriorating. “The results suggest that in general, softer monetary policies abroad (standard and non-standard) induce higher future loan default rates of banks from the same country or region. Furthermore, and as we show later, softer monetary policy induces banks to lend relatively more to firms with higher risk as proxied by higher ex-ante loan rates, which also explains our result on defaults” (Section 3A). The measurement convention is stated: default at t+12, or the last available observation if the pair leaves the sample, with these regressions using data only to December 2014, and results “qualitatively similar” at t+6 and t+24 (fn. 35). The paper also tested for asymmetry between easing and tightening and reports a mostly null result with one exception: “we also analyzed asymmetric monetary policy effects, and we do not find statistically different results… though in some other few cases, it may be due to lack of statistical power. One coefficient which is asymmetric is the impact of QE on loan defaults” (fn. 36).

Real effects across five distinct firm outcomes, because switching is rare – and the paper documents the switching margin rather than assuming it. The test interacts each country’s policy with the firm’s lagged share of credit from that country’s banks, where “a coefficient that is statistically zero implies that while at the loan level we find that (foreign) monetary policy matters, firms are able to smooth these foreign shocks by switching to other banks or to other forms of credit” (Section 3B). The coefficients are not zero: for firms whose bank credit was entirely with foreign banks, a one-standard-deviation easing raises loan volume by 1.5 percent, maturity by 4.9 percent and collateral by 4.8 percent, cuts the interest rate by 0.8 percent, and raises default by 5.3 percent. On the extensive margin, “a 1 standard deviation reduction in intrateY-fgn reduces firm exit due to loan defaults by 1 percent.” On the intensive margin, “total liabilities of firms (including bank credit) increase by 1.2 percent… while fixed-assets (i.e., net investment) rise by 0.5 percent,” and “employment also increases, but only by 0.3 percent.” The switching premise is documented directly: “only 9 percent of firms switching their main bank from one year to the next,” with switching “positively related with firm size and with the number of bank relations of a firm, and negatively related with loan’s duration and volume,” and similar rates of 8 and 11 percent for small and large firms (fn. 39). One internal inconsistency worth noting for a careful reader: the introduction gives the employment effect as 0.4 percent where Section 3B reports 0.3 percent.

Net investment because it is what matters for productivity – and the firm-level results are explicitly presented as a lower bound. “We analyze net, not gross, investment, which is common in the literature… If investment expenditures just match the depreciation of capital equipment, then gross investment rises; however, net investment is unchanged. Higher net investment, not gross one, is what matters for overall productivity, where net investment is computed as the annual change in fixed tangible assets” (fn. 41). Two caveats bound the firm-level block. On power: “since with this data set we only have few yearly observations for each firm after the QE period started, our results for the impact of non-standard monetary policies on real outcomes could lack statistical power (e.g. for loan outcomes, all QE results are statistically significant in the monthly level data).” On heterogeneity and selection: interacting policy with an indicator for firms with fewer than 50 employees shows “the effects are indeed stronger for smaller firms, while inexistent for large firms. Therefore, and given the somewhat overrepresentation of large firms in Orbis, our results for the firm balance-sheet variables suggest a lower [bound]” (Section 3B).

By splitting firms each period at the volume-weighted average loan rate and re-estimating separately, with the high-yield group verified as riskier ex post. “In each period we calculate the average interest rate charged by banks to all firms (firm-bank observations weighted by loan volume). We then separate our sample into two groups depending on whether their ex-ante cost of credit is above or below this average cost,” with the high-yield group confirmed to “have higher ex-post default rates” (Section 3C). The asymmetry runs across margins. Volume: “a 1 standard deviation decrease in the foreign monetary policy expands loan volume for the high-yield group by an average of 5 percent, and only by 1.3 percent in the low-yield group,” with “effects large for US, UK and Eurozone banks,” while a QE expansion raises high-yield volume by around 1.5 percent “but has no statistically significant effect on the [low-yield group].” Maturity shows the largest gap: “a reduction of 1 standard deviation in the average foreign interest rate lengthens the average loan maturity by 10 percent for firms with high-yield, whereas its effect is negligible among low-yield firms” – though on QE the paper reports the opposite tilt, “on average foreign QE has a stronger, albeit smaller, effect on low-yield firms.” On rates, “a 1 standard deviation reduction of foreign monetary policy translates into a 1.1 percent reduction of the average loan rate of high-yield firms” while low-yield rates do not respond. Collateral is the one margin running against the risk-taking reading: values are “in general higher for low-yield firms when foreign monetary policy is relaxed,” and adjustments “vary substantially depending on the bank’s nationality” (fn. 44).

Yes, and essentially only for the high-yield group, which is what makes the claim a risk-taking channel rather than a credit-availability channel. “Default rates are more responsive to movements in the monetary policy from the U.S. and the U.K. (both standard and non-standard) and from the Eurozone (mainly non-standard). For instance, a reduction of 1 standard deviation in the foreign interest rate increases the average default for high-yield firms by 11.7 percent and has no significant impact for low-yield firms. Similarly, the expansion in QE also increases the incidence of default. Changes in foreign QE are associated on average with an 8.6 percent increase in the share of bank credit in default among high-yield firms” (Section 3C). The inferential structure is stated with appropriate hedging: “greater risk-taking is associated with ex-ante observable variables (previous high loan rates) and with higher ex-post defaults. The overall evidence suggests an international risk-taking channel of monetary policy through foreign monetary policy rates and QE” – suggests, because the exercise compares subsamples rather than identifying a shift in banks’ risk preferences directly. On QE the euro-area exception recurs: “effects are not significant for Eurozone banks, except for higher ex-post loan defaults,” which the paper again ties to their sovereigns’ CDS (fn. 43).

That foreign banks fund the expansion from abroad and at shorter maturity while lending longer – the paper’s answer to its fifth question. Bank-level regressions “corroborate our loan-level results: when foreign monetary policy becomes more expansive, the total assets and one-year-ahead credit-in-arrears of foreign banks increase relatively more. We find a similar pattern with changes of foreign QE.” On the mechanism: “Compared to domestic banks, foreign banks borrow substantially more, especially from abroad, when foreign monetary policy is softer.” The short-term funding result is reported honestly as economically but not statistically strong: “while not statistically significant, the economic magnitude of the coefficient of foreign monetary policy on short-term liabilities is very strong (the coefficient is high and larger than in the other margins but with substantial higher standard errors), thereby suggesting that foreign banks obtain more short-term funding when foreign monetary policy is softer” (Section 3A). The synthesis is a maturity-mismatch reading: “our results are consistent with foreign banks taking on higher liquidity (partly from abroad) and credit risk (providing more credit and with higher ex-post defaults), and despite that the liabilities are more fragile (i.e., foreign and partially shorter), these banks lend at longer maturities in the asset side.”

Bank commercial characteristics, customer segmentation by export destination, and reliance on a handful of large lenders. On bank characteristics: re-estimating with time-varying bank controls – total size, liquidity and capital ratio – leaves results “not affected in any meaningful way,” and restricting to “the only five large banks (the four foreign and the largest Mexican one)” gives similar results; the paper also notes that “while foreign banks are indeed larger than the average bank, they are very similar to the largest Mexican bank” (Section 3A, fn. 38). On segmentation: “another similar concern is that the results are driven by a segmentation of customers. For example, U.K. banks may be serving firms exporting to the U.K. To control for this hypothesis, we compare our regressions for exporters vs. non-exporters, proxying them by firms in either (i) tradable vs. non-tradable industries following Mian and Sufi (2014) or in (ii) northern vs. southern states, since the northern states have substantially more economic relations with the U.S. as well as a larger share of exports to GDP (39 percent compared to 12 percent).” The results “show not different results for exporters and non-exporters,” which the authors read as unsurprising “as the subsidiaries of foreign banks in Mexico are important across all sectors.”

A case for coordination, framed around the specific point that foreign policy is set without reference to the recipient economy. “The overall results suggest spillovers of core-countries’ monetary policies on emerging markets, both in the foreign monetary softening part (with not only higher credit risk taken by foreign banks, but also higher liquidity risk stemming from higher foreign funding) and in the tightening part (with the negative associated local real effects in terms of lower firm total assets, net investment, employment and survival)” (Section 4). The destabilising asymmetry is then named: “foreign monetary policy is not only key to analyze the international channel and to obtain exogenous variation of monetary policy, as compared to local policy, but moreover it is not determined by the local economic conditions of emerging markets, so a change of foreign policy can be further destabilizing, especially given the foreign bank channel we show in this paper.” The recommendation is hedged as a potential need rather than a demonstrated one: the results “are consistent, among others, with some claims by the Governor Rajan of the Reserve Bank of India (2014) and Jackson Hole’s speech by Rey (2013) on the effects of core countries’ monetary policies on emerging markets’ economies, and thus suggest a potential need for a more coordinated global monetary policy, for example at the G-20 [level].”