Credit

Error message

Deprecated function: The each() function is deprecated. This message will be suppressed on further calls in _menu_load_objects() (line 579 of /var/www/drupal-7.x/includes/menu.inc).

The Costs of Corporate Debt Overhang Following the COVID-19 Outbreak

Published by Anonymous (not verified) on Tue, 01/12/2020 - 11:00pm in

Tags 

Credit

Kristian S. Blickle and João A. C. Santos

The Costs of Corporate Debt Overhang Following the COVID-19 Outbreak

Leading up to the COVID-19 outbreak, there were growing concerns about corporate sector indebtedness. High levels of borrowing may give rise to a “debt overhang” problem, particularly during downturns, whereby firms forego good investment opportunities because of an inability to raise additional funding. In this post, we show that firms with high levels of borrowing at the onset of the Great Recession underperformed in the following years, compared to similar—but less indebted—firms. These findings, together with early data on the revenue contractions following the COVID-19 outbreak, suggest that debt overhang during the COVID-recession could lead to an up to 10 percent decrease in growth for firms in industries most affected by the economic repercussions of the battle against the outbreak.

Debt Overhang Defined

Economists argue that firms may experience a debt overhang problem and be unable to finance new worthwhile projects when the face value of a firm’s debt exceeds its payoff. This could be because potential debtors cannot accurately evaluate a company’s investment opportunities, while equity holders are averse to financing projects whose benefits accrue only to existing debt holders. This problem tends to be amplified in downturns because existing projects become less profitable and the cash flows of firms decline.

Debt Overhang and the Great Recession

We analyze the impact of debt overhang on firm growth during the Great Recession in a new paper. We focus on total liabilities relative to cash flow as a measure of debt overhang because liabilities are a broad indicator of "indebtedness." Further, we consider a cash flow-based measure, rather than a pure balance sheet-based measure, because cash flow (and revenue) can be more volatile than a firm’s balance sheet structure and may depend on macroeconomic conditions unrelated to the firm’s business model.

Consistent with the insight discussed above, we find that firms that entered the Great Recession with high levels of debt to cash flow, experience poorer performance in subsequent years. This is evident in the charts below that compare firms’ performance as measured by indexed (a) asset growth, (b) growth of employees, and (c) capital expenditures, in the years after the Great Recession. We split firms based on the ratio of their liabilities to EBITDA (a cash flow proxy) in 2007. In each chart it is clear that firms with high liabilities to cash flow (or even high liabilities and negative cash flow) experience slower growth during the recession. According to our aggregated average estimates, firms with debt overhang experience an asset growth that is 2 percent lower during ordinary times and up to 3 percent lower during the Great Recession than comparable firms without debt overhang.

The Costs of Corporate Debt Overhang Following the COVID-19 Outbreak

We additionally show that the effects of debt overhang during the Great Recession are significantly more pronounced for firms with more significant needs for external funding. We assess firms' need for external funding through several different measures, including (i) the maturity left in their debt at the onset of the crisis, (ii) whether they refinanced their debt in the year leading up to the crisis, (iii) the amount left unused in their credit lines, and (iv) reductions in the size of the credit lines available to them.

Debt Overhang and COVID-19

The COVID-19 outbreak has the potential to have an even bigger effect on the economy through the channel of debt overhang than what we identified in the Great Recession. Corporate sector indebtedness stood at record-high levels at the time of the outbreak in early 2020. Based on data available for the first two quarters of 2020, we see that firms that entered the COVID-19 crisis with high levels of overhang experienced slower or even negative growth during the first half of 2020. Moreover, the economic shutdown that followed the spread of COVID-19 had a significantly negative effect on firms' cash flows. This is likely to have mechanically raised the debt overhang of firms in affected industries during the crisis. For example, firms operating in tourism, restaurant, hospitality, and related industries entered the crisis with an average debt overhang that was twice as high as the average overhang of firms in other industries. Additionally, the revenue contractions were most severe in these industries (following the economic shutdown), meaning debt overhang likely have increased much further. Based on the analysis in our paper, we estimate that the combination of high prior debt overhang and revenue contractions in 2020 could lead firms in these most affected industries to grow 10 percent more slowly in a Great Recession-type crisis than they would in ordinary times.

There is at least one additional factor that is likely to increase the costs of debt overhang in the current crisis: the ownership structure of debt. Debt forgiveness is usually an important mechanism to resolve problems of debt overhang but it may be hampered when debt ownership is diffuse. Forgiveness by any one party presents a free rider problem if any other debtholder does not follow suit. This problem is compounded when loan ownership is dominated by many investors owning small loan shares as in the leveraged loan market. For example, at the end of 2019, there were on average 203 collateralized loan obligations (CLOs) in Ba rated loans, and their average loan share was just 0.24 percent.

Of course, in instances when voluntary debt forgiveness is not feasible, the Chapter 11 bankruptcy process may be a viable option. Bankruptcy courts can help encourage a debt renegotiation and debtor-in-possession (DIP) loans can help a firm circumvent some of the challenges of out-of-court reorganizations. This may be an option for large firms provided that the bankruptcy courts are not overburdened and DIP financing is still available. As for small firms, notwithstanding the restructuring opportunities provided by the Small Business Reorganization Act, their more limited access to DIP financing may make reorganization more difficult. The liquidation or simply reduced long-term growth prospects of small firms with debt overhang represents a sizable risk for the economy because this segment of the corporate sector collectively accounts for a large portion of employment and economic growth.

Kristian S. Blickle


Kristian S. Blickle
is an economist in the Federal Reserve Bank of New York’s Research and Statistics Group.

João A. C. Santos
João A. C. Santos is a senior vice president in the Bank’s Research and Statistics Group.

How to cite this post:

Kristian S. Blickle and João A. C. Santos, “The Costs of Corporate Debt Overhang Following the COVID-19 Outbreak,” Federal Reserve Bank of New York Liberty Street Economics, December 1, 2020, https://libertystreeteconomics.newyorkfed.org/2020/12/the-costs-of-corpo....




Disclaimer

The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

Published by Anonymous (not verified) on Wed, 21/10/2020 - 10:00pm in

Sarah Ngo Hamerling, Donald P. Morgan, and John Sporn

LSE_Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

Did the 2007-09 financial crisis or the regulatory reforms that followed alter how banks change their underwriting standards over the course of the business cycle? We provide some simple, “narrative” evidence on that question by studying the reasons banks cite when they report a change in commercial credit standards in the Federal Reserve’s Senior Loan Officer Opinion Survey. We find that the economic outlook, risk tolerance, and other real factors generally drive standards more than financial factors such as bank capital and loan market liquidity. Those financial factors have mattered more since the crisis, however, and their importance increased further as post-crisis reforms were phased in in the middle of the following decade.

Measuring Credit Standards

The Fed’s Senior Loan Officer Opinion Survey (SLOOS) is a quarterly survey of about eighty large domestic banks and twenty-four foreign banks. The domestic banks, our focus, account for roughly 70 percent of all U.S. banks’ assets. The first question in the survey asks about credit standards for commercial and industrial (C&I) loans:

Over the past three months, how have your bank's credit standards for approving applications for C&I loans or credit lines—other than those to be used to finance mergers and acquisitions—to large and middle-market firms changed?

a. Tightened considerably
b. Tightened somewhat
c. Remained basically unchanged
d. Eased considerably
e. Eased somewhat

The chart below plots the number of banks reporting tightening and easing between 2001 and 2019. Banks tend to cycle from easing to tightening before recessions but that shift had never been so dramatic as during the 2007-09 crisis. Researchers have found that tighter standards strongly predict slowdowns in bank credit growth and economic activity (Lown and Morgan; Basset et al.).

Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

What Drives Credit Standards?

Banks reporting a change in credit standards or terms are asked to rate the importance of various factors in driving the change. The set of reasons offered since at least 2001 is listed in the table below in slightly abridged form (the full text is here). Easing and tightening banks are offered the same set of reasons, except with the directions reversed. For example, the text on bank capital is either “improvement in your bank’s current or expected capital position” or “deterioration …”

The second column below reports the mean share of respondents that ranked the reason as either very or somewhat important. By that (simple) metric, real factors drive standards more than financial factors like bank capital and liquidity. That is notable in light of recent academic literature stressing the importance of banks’ capital strength in driving credit supply (Bernanke and Gertler; Peek and Rosengren). Of course, the hypothesis of a bank capital channel is that capital also matters, along with real factors, not that it matters more. Note also that financial factors were markedly more important during the financial crisis, as shown in the final column.

Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

Columns (3) and (4) report the means separately for banks that reported easing or tightening and column (5) reports the difference in means. The financial factors matter symmetrically, that is, increased concerns about banks’ capital or loan liquidity drive easing standards as much (or as little) as decreased concerns drive lower standards. The real factors, by contrast, matter asymmetrically, with all but one mattering more for tightening than for easing. The exception is competition; increased competition among lenders drives standards downward much more than weaker competition drives them upward. The strong link between increased competition and easing standards is consistent with the long-standing notion that increased competition spurs more risk taking by banks (Carlson, Correia and Luck provide interesting historical evidence on that conjecture; Goetz reviews the literature). However, we’re not aware of theories that predict the asymmetric effects of competition on credit standards.

Different Drivers Since the Crisis?

To see if the drivers have changed since the financial crisis of 2008-09, we compare the average share of banks reporting that a reason was important before the crisis to the average share after the crisis, conditional on the number of banks that changed standards. Technically, we regressed the number of banks saying x was an important reason why they tightened or eased each quarter on a constant, a “post-crisis” indicator, and the number of banks that tightened or eased. Given the asymmetries just noted, we estimated separate models for easing and tightenings. The chart below summarizes the results. The asterisks over the red bars indicate if that reason was significantly more or less important after the crisis.

While real factors still predominate over the cycle, financial factors are more significant since the crisis. On the easing side, the mean share of easing banks citing their capital position rose from 18 percent to 23 percent, eclipsing banks’ own risk tolerance in importance. On the tightening side, both capital and loan market liquidity were more significant drivers; the mean share of tightening banks citing capital concerns increased from 5 percent to 14 percent while the mean share citing liquidity concerns increased from 7 to 18 percent.

Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

To see what’s behind the post-crisis shift, we plot the fraction of tightening banks citing capital and liquidity concerns below. Those concerns remain elevated for several years after the crisis, which partly explains their heightened importance since the crisis. That post-crisis “hangover” is not the whole story, however, since capital and liquidity concerns resurged somewhat in the middle of the decade.

Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis

That resurgence roughly coincides with the finalization of stricter bank capital and liquidity rules in 2013 and 2014, respectively, and their gradual implementation over the next few years. We can’t put too fine a point on this interpretation given our simple analysis, but the timing squares with other recent evidence that those reforms might have unintentionally reduced or reallocated bank credit supply (for example, Roberts, Sarkar, and Shachar; Kovner and Van Tassel). Of course, the policy question is whether any such costs of reforms outweigh the benefits of increased bank resilience and financial stability, a more difficult question indeed.

Hammerling_sarah

Sarah Ngo Hamerling is a senior research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group.

Donald P. Morgan


Donald P. Morgan
is an assistant vice president in the Bank’s Research and Statistics Group.

Sporn_johnJohn Sporn is a collateral value analysis associate in the Bank’s Markets Group.

How to cite this post:

Sarah Ngo Hamerling, Donald P. Morgan, and John Sporn, “Bank Capital, Loan Liquidity, and Credit Standards since the Global Financial Crisis,” Federal Reserve Bank of New York Liberty Street Economics, October 21, 2020, https://libertystreeteconomics.newyorkfed.org/2020/10/bank-capital-loan-....




Disclaimer

The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

How Do Consumers Believe the Pandemic Will Affect the Economy and Their Households?

Published by Anonymous (not verified) on Sat, 17/10/2020 - 2:00am in

Olivier Armantier, Leo Goldman, Gizem Koşar, Jessica Lu, Rachel Pomerantz, and Wilbert van der Klaauw

How Do Consumers Believe the Pandemic Will Affect the Economy and Their Households?

In this post we analyze consumer beliefs about the duration of the economic impact of the pandemic and present new evidence on their expected spending, income, debt delinquency, and employment outcomes, conditional on different scenarios for the future path of the pandemic. We find that between June and August respondents to the New York Fed Survey of Consumer Expectations (SCE) have grown less optimistic about the pandemic’s economic consequences ending in the near future and also about the likelihood of feeling comfortable in crowded places within the next three months. Although labor market expectations of respondents differ considerably across fairly extreme scenarios for the evolution of the COVID pandemic, the difference in other economic outcomes across scenarios appear relatively moderate on average. There is, however, substantial heterogeneity in these economic outcomes and some vulnerable groups (for example, lower income, non-white) appear considerably more exposed to the evolution of the pandemic.

Measuring Consumer Expectations

The COVID pandemic, dubbed “The Uncertainty Pandemic” by Harvard economist Kenneth Rogoff, has been accompanied by high and pervasive uncertainty about the future path of the virus and the response to it by policymakers, households, and firms. Here we assess consumers’ views about the future course of the pandemic and about its impact on future household economic decisions and outcomes, drawing on the SCE. Since June 2013 the SCE collects information on the economic expectations and behavior of households.

The SCE is designed to be a nationally representative, internet-based survey of about 1,300 U.S. households. In addition to the monthly core questionnaire, special surveys are fielded at regular frequencies on various topics and are occasionally fielded on an ad-hoc basis to answer policy relevant questions in a timely manner. The analysis in this post is based on data collected as part of two special surveys on the pandemic fielded in June (between June 10-June 30) and August (between August 6‑August 21).

Expectations Regarding the Duration of the Pandemic’s Economic Impact

We start with consumers’ beliefs regarding the expected number of weeks it will take for U.S. economic activity to get back to pre-COVID levels. When asked in June, the average expected number of weeks required for economic recovery was 94 weeks. This average increased to 132 weeks (more than 2 years) in August. Even though there are differences in the expectations of respondents, the increase since June in the expected duration of the economic recovery is similar across demographic groups.

Based on the August survey, 32 percent of the respondents in our sample work from home due to the pandemic. This share rises as high as 40 percent for respondents younger than age 40, and 50 percent for respondents with a college degree or with household incomes over $75,000. When asked about their expectations of the earliest date of a return to their previous work location, employed respondents on average report a 36 percent chance for a return within the next three months, 48 percent for returning between three to twelve months, and a 16 percent chance for returning to the office in more than a year, as indicated in the table below. When compared to the June results, we observe a significant deterioration in expectations of going back to the office in less than 3 months (58 percent chance in June versus 36 percent chance in August).

The June and August surveys also elicit information on whether respondents feel comfortable going to crowded places such as movie theaters, concerts, and airports. In both waves, we observe only around 23 percent of respondents reporting currently feeling comfortable going to such crowded places. As expected, this share goes down for older respondents (14 percent for age 60+). Moreover, when we ask when they think they will start feeling comfortable, respondents in the August survey report an average 45 percent chance for feeling comfortable sometime in the next three to twelve months and 43 percent chance of feeling comfortable in more than twelve months. As indicated in the table below, this reflects a considerable deterioration in the expected chance of feeling comfortable in crowded places within the next three months from June to August (19 percent chance in June versus 12 percent chance in August). Except for younger respondents being significantly more optimistic, these expectations are comparable across race, education, income, homeownership and Census region.

How Do Consumers Believe the Pandemic Will Affect the Economy and Their Households?

Summing up, our results show that most respondents do not see a quick recovery from the pandemic and have grown less optimistic since June about the pandemic’s economic consequences ending in the near future and about the likelihood of feeling comfortable in crowded places within the next three months.

Hypothetical Scenarios

How sensitive do consumers expect their future spending, income, and debt repayment to be to the evolution of the pandemic? To address this question, we took an experimental approach as part of the special SCE survey fielded in August 2020. The basic idea is to test how spending, income, and debt repayment expectations respond to different scenarios for the path of the pandemic. Each SCE respondent was asked to consider three hypothetical scenarios for the possible evolution of the COVID pandemic in the United States over the next six months. Under the “baseline” scenario, the levels of new coronavirus cases, deaths, and restrictions on distancing in the United States (including where the respondent currently lives) all remain exactly the same as they currently are today. The coronavirus cases, deaths, and restrictions on distancing all gradually drop to zero over the next six months in the “good” scenario, whereas they double in the “bad” scenario. For each scenario, we ask the respondents what they think would happen to their monthly household spending, income, their ability to make necessary payments, their employment prospects, and chances of applying for government assistance over the next six months.

As indicated in the table below, respondents expect their monthly spending to be $2,883 on average under the baseline scenario. They expect their spending to increase by 4.6 percent to $3,016 under the good scenario and to decrease by 5.9 percent to $2,714 under the bad scenario. Note that, if taken at face value, the 5.9 percent decrease in spending in the bad scenario (in which COVID cases doubled) can be interpreted as a -6 basis point “COVID elasticity of spending.” That is an increase of 1 percent in COVID-19 cases and deaths results in a 0.06 percentage decrease in household spending. In both scenarios, the dollar and percentage change in spending is larger for high income respondents and for those with a college degree.

As indicated in the table below, respondents on average expect their monthly household income to be $6,811 under the baseline scenario. Respondents only expect a modest increase in their household income of 1.2 percent to $6,896 under the good scenario, and a decrease of 8.1 percent to $6,262 under the bad scenario. In both scenarios, the dollar and percentage change in income is again larger for higher income respondents.

Under the baseline scenario respondents assign an 8 percent chance of missing a minimum debt or rent payment over the next six months, as compared to 5 percent and 10 percent under the good and the bad scenarios, respectively. These differences can be considered fairly modest and lower than one may have expected, especially considering the fairly extreme events the hypothetical scenarios capture. Compared to the baseline scenario, most of the changes in delinquency expectations in the good and bad scenarios are driven by lower income respondents, those under the age of 40, respondents without a college degree, and respondents who experienced a decline in income since the start of the pandemic.

The respondents who are currently employed believe that on average there is a 13 percent chance they may lose their job over the next six months—under the baseline scenario in which levels of new coronavirus cases, deaths, and restrictions on distancing remain the same as they were at the time of the August survey. In contrast, the average probability to lose one’s job is 8 percent in the good scenario and jumps to 19 percent under the bad scenario. Compared to the baseline scenario, most of the differences in expectations in the good and bad scenarios are driven by respondents in the bottom tercile of income.

How Do Consumers Believe the Pandemic Will Affect the Economy and Their Households?

Finally, in the baseline scenario, respondents evaluate the chance that they will apply for an assistance program (such as, food stamps, income assistance, rent or mortgage or other debt repayment assistance programs) over the next six months to be 15 percent on average. The average chance of applying to an assistance program drops to 10 percent in the good scenario and increases to 18 percent in the bad scenario. Compared to the baseline scenario, most of the changes in the good and bad scenarios are driven by respondents in the bottom tercile of income, those below the age of 45, and respondents who self-identify as non-white. In particular, non-white respondents believe there is 29 percent chance that they will apply for an assistance program under the bad scenario. The relatively high likelihood of applying to assistance programs in the different scenarios may contribute to explaining the comparatively low expected delinquency rate sensitivity discussed above.

To sum up, our analysis reveals the importance of both the evolution of the pandemic as well as the continuation of government support programs for household expectations. Labor market expectations appear rather sensitive to the evolution of the pandemic. While our scenarios present fairly extreme differences in the evolution of the COVID pandemic, respondents report moderate responsiveness of their income and spending expectations to the course of the pandemic, and a relatively muted sensitivity for delinquency expectations, which have remained reasonably low and stable throughout the pandemic. There is, however, substantial heterogeneity and some vulnerable groups (for example, lower income, non-white) appear considerably more exposed to the evolution of the pandemic.

The relative insensitivity of delinquency expectations, together with a large sensitivity of expected application rates to government assistance programs, suggest that the forbearance and other assistance programs currently in place have been effective in supporting household financial conditions during the pandemic. It also suggests that their continuation likely would mitigate delinquencies over the coming months.

Olivier Armantierr

Olivier Armantier is an assistant vice president in the Federal Reserve Bank of New York’s Research and Statistics Group.

Leo Goldman

Leo Goldman is a senior research analyst in the Bank’s Research and Statistics Group.

Gizem Koşar

Gizem Koşar is an economist in the Bank’s Research and Statistics Group.

Jessica Lu is a senior research analyst in the Bank’s Research and Statistics Group.

Rachel Pomerantz

Rachel Pomerantz is a senior research analyst in the Bank’s Research and Statistics Group.

Vanderklaauw_wilbert

Wilbert van der Klaauw is a senior vice president in the Bank’s Research and Statistics Group.

How to cite this post:

Olivier Armantier, Leo Goldman, Gizem Koşar, Jessica Lu, Rachel Pomerantz, and Wilbert van der Klaauw, “How Do Consumers Believe the Pandemic Will Affect the Economy and Their Households?,” Federal Reserve Bank of New York Liberty Street Economics, October 16, 2020, https://libertystreeteconomics.newyorkfed.org/2020/10/how-do-consumers-b....




Disclaimer

The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Wir sind die Roboter: can we predict financial crises?

Published by Anonymous (not verified) on Fri, 02/10/2020 - 6:00pm in

Tags 

Credit

Kristina Bluwstein, Marcus Buckmann, Andreas Joseph, Miao Kang, Sujit Kapadia and Özgür Şimşek

Financial crises are recurrent events in economic history. But they are as rare as a Kraftwerk album, making their prediction challenging. In a recent study, we apply robots — in the form of machine learning — to a long-run dataset spanning 140 years, 17 countries and almost 50 crises, successfully predicting almost all crises up to two years ahead. We identify the key economic drivers of our models using Shapley values. The most important predictors are credit growth and the yield curve slope, both domestically and globally. A flat or inverted yield curve is of most concern when interest rates are low and credit growth is high. In such zones of heightened crisis vulnerability, it may be valuable to deploy macroprudential policies.

History and the costs of financial crises

In the history of modern capitalism, financial crises — roughly defined as periods of severe distress and widespread failures in the financial system with significant macroeconomic costs — have been a recurrent theme. For example, the UK experienced regular crises throughout much of the 19th century. But the list of notable crises and crashes is long, encompassing the Tulip, Mississippi and South Sea bubbles, all accompanied by their own stories of manias.

The economic and social costs of these events are often staggeringly high. The estimated average cost of a financial crisis is around 75% of GDP, equivalent to £21,000 for every person in the UK in current terms. Unemployment increases by seven percentage points, house prices drop by 35%, and life expectancy declines by seven months in the years after the crisis.

Given the frequency and cost of financial crises, decision makers in governments and central banks want to avoid crises altogether or at least mitigate their destructive consequences. Predicting financial crises is hard, however. Even people as smart as lsaac Newton lost their fortunes in such events. So, can we accurately predict crises? And if so, what are the most important predictors and how reliable are they? We try to answer these questions using machine learning on a long-run dataset stretching back to the 1870s and including 17 developed countries covering most of world output over that period.

Why machine learning?

While history and economics goes in circles, technology goes uphill. In this sense, machine learning, a set of data-driven prediction techniques, has achieved some impressive feats in recent years. Facebook’s “DeepFace” is as accurate as humans in face recognition. A computer is better in identifying skin cancer from pictures than the average dermatologist. And Google’s AlphaGo beats the world’s strongest player in the highly complex Chinese board game of Go. Machine learning has also been successful in some economic prediction problems such as forecasting recessions and bond risk premia.

When is machine learning successful? Machine learning methods thrive in situations where many different factors play a role, and the relation between these factors and those which we want to predict — e.g. who is in this picture, is it cancer or not — is complex. We are in this kind of situation when trying to predict whether or not there is likely to be a financial crisis in the next year or two.

Machine learning horse race for financial crisis prediction

There are many potential machine learning models that might help to predict crises. We perform a horse race across some popular models, including random forests, support vector machines (SVM) and artificial neural networks among others. We aim to predict financial crises one or two years in advance. This would give policy makers time to react — for example by activating macroprudential policies should a model predict a high chance of a crisis.

Chart 1 shows the performance score for all models in the horse race, as captured by the area under the receiver operating characteristic, a standard metric for binary prediction problems. A model that perfectly distinguishes crisis and non-crisis observations on unseen data gets a score of one, while a model with predictions no better than a coin toss gets a score of 0.5 (the worst possible score). Traditionally, logistic regression is used for predicting crises and is our reference point. The machine learning models, except the decision tree, all outperform the logistic model.

Chart 1: Test performance of the different prediction models

The difference in performance between our best model, extremely randomised trees (a type of random forest), and logistic regression may seem relatively small but it is economically significant. To see this, we calibrate both models to ensure they correctly identify 80% of crises, i.e. we set the proportion of crises we aim to predict correctly. Then, we can compare the false alarm rate, i.e. the proportion of times when the model signals a crisis which doesn’t subsequently happen, as a measure of the cost of unnecessary policy interventions. This false alarm rate is 19% for extremely randomised trees compared to 32% for the logistic model, a substantial reduction.           

Opening the black box: which variables matter for crisis prediction?

Machine learning models are often called “black boxes” because it is hard to understand which variables drive a model’s predictions. This conflicts with the need of decision makers to understand the key economic factors which might be related to the build-up of crises and explain any policy decisions with reference to those. Recent developments like the Shapley value framework help to tackle this issue by assigning a well-defined contribution of how much each variable contributes to a model’s predictions. Concretely, the predicted crisis probability can be decomposed into contributions coming from individual variables, as measured by their Shapley values. These values can then be used to rank variables according to their importance.

Chart 2 shows the mean absolute Shapley values across all predictions for all of the above models apart from the decision tree. The machine learning methods and logistic regression consistently identify the same main predictors for financial crises: domestic and global credit growth, and the domestic and global slope of the yield curve, where the latter refers to cases where the cost of short-term borrowing is relatively high compared to the cost of long-term borrowing.

Chart 2: Model explanations using Shapley decompositions

Other work has previously established that strong credit growth, both domestically and globally, is an important predictor of financial crises.  Our results on the importance of the yield curve are more novel. The yield curve is a well-established leading indicator for recessions. But we find it to be of independent importance in predicting financial crises.

The strong predictive power of our machine learning models may partially stem from the simple and intuitive nonlinear relationships and interactions that they uncover. These help to identify zones of particular vulnerability to future financial crises. For example, we find that crisis probability increases materially at high levels of global credit growth but this variable has nearly no effect at low or medium levels. Similarly, interactions are important — particularly between global and domestic variables. For example, many crises occur in the face of strong domestic credit growth and a globally at or inverted yield curve. We also find that a flat or inverted yield curve domestically is more concerning when nominal interest rates are at low levels. In such environments, financial market players may take on excessive risk in a ‘search-for-yield’ to try to boost their returns — in accordance with some descriptions of financial manias and crashes.

What does our best model say for the UK?

Chart 3 shows the predicted likelihood of a crisis and the factors driving our best machine learning model since 1980. Vertical red shaded bars indicate the observations we would like to predict, one or two years ahead of an actual crisis (grey shaded bars). The black circles indicate the model prediction which is decomposed into the contributions from its four most important variables (Shapley values) from Chart 2.

Chart 3: Decomposition of the machine learning model prediction for the likelihood of financial crises

The model correctly predicts the small banks crisis in 1991 and the Global Financial Crisis 2007–08. But the driving factors of the predictions differ. The former is driven by rapid domestic credit growth and an inverted domestic and global yield curve, whereas the latter crisis is mostly predicted by global credit growth. This shows that the model is flexible enough to account for different types of crises, which is crucial given the long time span our analysis covers and the changes the global economic system underwent during this period.

The way forward

While our analysis does not necessary say the above factors cause financial crises, it does highlight that they make a country more vulnerable to financial crises. There will always be inherently unforeseeable events, such as the economic fallout caused by Covid-19, which remain very challenging for any model to predict. But identifying a financial system as more vulnerable and therefore more likely to amplify such an unexpected shock into a fully-fledged financial crisis is crucial given the enormous economic, political, and social consequences that financial crises entail.

Kristina Bluwstein works in the Bank’s Macroprudential Strategy and Support Division, Marcus Buckmann and Andreas Joseph work in the Bank’s Advanced Analytics Division, Miao Kang works in the Bank’s Data And Statistics Division, Sujit Kapadia works for the European Central Bank and Özgür Şimşek works at the University of Bath.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Debt Relief and the CARES Act: Which Borrowers Face the Most Financial Strain?

Published by Anonymous (not verified) on Wed, 19/08/2020 - 9:00pm in

Rajashri Chakrabarti, Andrew Haughwout, Donghoon Lee, William Nober, Joelle Scally, and Wilbert van der Klaauw

 Which Borrowers Face the Most Financial Strain?

In yesterday's post, we studied the expected debt relief from the CARES Act on mortgagors and student debt borrowers. We now turn our attention to the 63 percent of American borrowers who do not have a mortgage or student loan. These borrowers will not directly benefit from the loan forbearance provisions of the CARES Act, although they may be able to receive some types of leniency that many lenders have voluntarily provided. We ask who these borrowers are, by age, geography, race and income, and how does their financial health compare with other borrowers.

Who is Without Mortgage and Student Debt?

To understand the distribution of borrowers who will be ineligible for the debt relief provisions of the CARES Act, we draw on the New York Fed’s Consumer Credit Panel—a nationally representative sample of Equifax credit report data. Our data set for this post includes a representative 1 percent sample of the nation’s adults with credit records in anonymized form.

The map below shows the (adult) population-weighted geographic (zip code) distribution of the share of the population with neither mortgages nor student debt. Each bin represents 1/4 of the U.S. adult population, so one quarter of adults live in zip codes where less than 60 percent have a student loan or mortgage, while another quarter lives in zip codes in which more than 75 percent have neither type of debt. The darkest-colored bin, where more than 75 percent of people are borrowers with neither a student loan nor mortgage, dominates by land area: 40 percent of all zip codes fall in the top quartile, indicating that these zip codes are less populated than those with higher rates of mortgage and student debt. Spatially, the borrowers without mortgages and student debt largely live in rural areas and they constitute a higher share of the borrowers in rural areas: 74 percent of Americans living in rural areas are in this category, versus 68 percent of people living in MSAs (urban areas).

LSE_2020_debt-relief-heterogeneity-Part 2_map1_art-01

Borrowers with neither mortgages nor student debt skew older than those who have either form of debt more generally. These older borrowers have either paid off their mortgages or never owned a home to begin with. Also they are less likely to have originated a student loan when they were younger. The median age of these borrowers is 56, while the median age of mortgage holders is 51 and that of student debt borrowers is 34. Although the median age of borrowers without any mortgage and student debt, and the median age of mortgage holders are not very different, the age distributions are different.

Partitioning zip codes by income (low-income, mid-income, high-income) and by race (majority-Black, majority-Hispanic, majority-white and mixed) as defined in our previous post, we investigate where these borrowers are concentrated (See the table below). First, distinguishing neighborhoods by income, we find that these borrowers are more concentrated in low-income neighborhoods: 72 percent of the borrowers in low-income neighborhoods have neither mortgage nor student debt whereas in mid-income and high-income neighborhoods these numbers are, respectively, 63 percent and 56 percent.

Distinguishing neighborhoods by race, we find that these borrowers are more concentrated in majority Hispanic and majority Black neighborhoods with 71 percent of borrowers in majority-Hispanic neighborhoods and 67 percent of borrowers in the majority Black neighborhoods. This compares to 62 percent in majority white neighborhoods. This analysis reveals that a large share of borrowers in low income, majority Hispanic, and majority Black neighborhoods will not receive direct relief from the debt moratorium provisions of the CARES Act.

 Which Borrowers Face the Most Financial Strain?

Borrowers who carry loans other than mortgages or student debt have lower credit scores (as captured by Equifax risk scores) than other borrowers, as we see in the table below. The median Equifax risk score for this group is 669 while the median risk score of all individuals with any debt is 700. To further understand the state of financial health of this group of borrowers, below we present the proportions of borrowers in this category who are delinquent on credit card payments by at least 90 days or in collection, conditional on having credit card debt. We find that 11.6 percent of this group of borrowers are more than 90 days delinquent on their credit card debt (last row). This compares with an overall delinquency rate, of 9.9 percent in our sample for credit card debt.

We find that older borrowers with loans other than mortgages and student debt are in better financial health than working-age borrowers. This reflects the fact that older people have often paid off their mortgages and student debt, while younger people without mortgages or student loans often had never owned a home. Regardless, these borrowers will not receive debt relief under the CARES Act forbearance provisions, although they can receive relief from its other provisions, such as direct stimulus checks.

Differentiating by neighborhood income, borrowers without any mortgage or student debt who live in low-income neighborhoods are considerably more likely to be in financial distress than those from high- or mid-income neighborhoods—as captured by both risk score and credit card delinquency. Differentiating by race, we find that borrowers without mortgage and student debt from Black neighborhoods are in markedly worse financial health than those coming from other neighborhoods. The risk score of these borrowers who reside in Black neighborhoods is 574 and their credit card delinquency and collection rate is at 21.8 percent. These contrast with delinquency rates of 16.4 percent and 9.8 percent for the group residing in majority Hispanic and majority white neighborhoods, respectively.

 Which Borrowers Face the Most Financial Strain?

In this post, we have focused on borrowers who carry debt other than mortgages and student loans, and therefore, will not directly benefit from the debt payment moratorium provisions under the CARES Act. We find that these borrowers are likely to be older than those holding mortgage and student debt, and more likely to be concentrated in low income, majority Black and majority Hispanic neighborhoods. They are more likely to be in financial distress than other borrowers, as captured by credit card delinquency rates. Further, as revolving debts are important for smoothing gaps in income, borrowers with already-delinquent credit card accounts are unlikely to be able to lean on these accounts to smooth consumption. Given this, relief from other parts of the Act, such as unemployment insurance and stimulus checks, will therefore be particularly important in mitigating the crisis' impact on the financial health of these borrowers.

Rajashri Chakrabarti

Rajashri Chakrabarti is a senior economist in the Federal Reserve Bank of New York’s Research and Statistics Group.

Andrew Haughwout

Andrew F. Haughwout is a senior vice president in the Federal Reserve Bank of New York’s Research and Statistics Group.

Donghoon Lee

Donghoon Lee is an officer in the Bank’s Research and Statistics Group.

Joelle Scally

Joelle Scally is a senior data strategist in the Bank's Research and Statistics Group.

Wilbert van der Klaauw

Wilbert van der Klaauw is a senior vice president in the Bank’s Research and Statistics Group.

How to cite this post:

Rajashri Chakrabarti, Andrew Haughwout, Donghoon Lee, William Nober, Joelle Scally, and Wilbert van der Klaauw. “Debt Relief and the CARES Act: Which Borrowers Face the Most Financial Strain?" August 19, 2020, https://libertystreeteconomics.newyorkfed.org/2020/08/debt-relief-and-th....

Additional heterogeneity posts on Liberty Street Economics:

Heterogeneity: A Multi-Part Research Series




Disclaimer

The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Who Has Been Evicted and Why?

Published by Anonymous (not verified) on Wed, 08/07/2020 - 9:15pm in

Tags 

Credit, Housing

Andrew Haughwout, Haoyang Liu, and Xiaohan Zhang

LSE_Who Has Been Evicted and Why?

More than two million American households are at risk of eviction every year. Evictions have been found to cause prolonged homelessness, worsened health conditions, and lack of credit access. During the COVID-19 outbreak, governments at all levels implemented eviction moratoriums to keep renters in their homes. As these moratoriums and enhanced income supports for unemployed workers come to an end, the possibility of a wave of evictions in the second half of the year is drawing increased attention. Despite the importance of evictions and related policies, very few economic studies have been done on this topic. With the exception of the Milwaukee Area Renters Study, evictions are rarely measured in economic surveys. To fill this gap, we conducted a novel national survey on evictions within the Housing Module of the Survey of Consumer Expectations (SCE) in 2019 and 2020. This post describes our findings.

Our survey evidence complements recent studies based on administrative court data in two ways. First, evictions can happen outside of the court system. Some evidence even suggests that more than half of evictions are informal, without being ordered by a court. Second, within a survey, we can elicit reasons for evictions—for example, loss of income, as opposed to divorce or sale of a property by the landlord—to shed light on the events leading up to evictions.

Our eviction survey module consists of the following questions. First, to encourage truthful reporting by respondents with an eviction history, we ask all respondents whether they know someone who has been evicted since 2006. Then, respondents are asked whether they themselves have been evicted in the past, and if so, the year of eviction. Finally, we elicit reasons for eviction by offering respondents nine predefined options, such as “health issues/medical bills” and “increase in monthly rent or utility costs,” as well as the ability to type in their own reasons if theirs is not covered by the options provided. After merging these survey responses with a rich set of demographics and other outcomes collected from the SCE, we examine the links between eviction, income, homeownership status, and credit access.

Eviction Rate and Income

Our analysis finds that evictions are relatively common in the United States. More than 24 percent of households know someone who has been evicted and about 4 percent of households have been evicted themselves. These rates are substantially higher for current renters, with 37 percent knowing someone evicted since 2006 and 9 percent having an eviction history. For homeowners, these shares are 21 percent and 2 percent, respectively. Our eviction rate is slightly lower than that of a previous study based on the Milwaukee Area Renters Study, which showed that one in eight renters in Milwaukee were evicted between 2009 and 2011. The different eviction rates can potentially be attributed to the differences between samples (national versus local) and different survey years.

Eviction is related to household income in a highly nonlinear way, as shown in the chart below. About 73 percent of the households who have been evicted before have a current income under $50,000. For income bins above $50,000, the eviction rate is always under 3 percent, and gradually declines as income rises. The percentage of households knowing someone who has been evicted is highly correlated with the share of households who have been evicted themselves.

Causes of Eviction

Turning to reasons for eviction, the next chart shows that slightly more than half of respondents with an eviction history selected “job loss/ unemployment,” “reduction in income,” or both as reasons for their eviction, making income or job loss the most prevalent reason.

Besides income loss, about 15 percent of the evictions were driven by property sales, remodeling, or landlords changing the property to a primary residence for themselves. In fact, for households that did not suffer an income loss, sale of the property by the landlord was the most common driver of evictions. (Note that we did not offer these “no‑fault” reasons as options for respondents to choose from, meaning that our survey participants entered them in the “Other” category. This design makes it unlikely that respondents untruthfully used these reasons for their eviction.) Taken together, a meaningful fraction of evictions is no‑fault. Such results motivate recent state and local policies requiring “good cause” for evictions.

Who Has Been Evicted and Why?

Homeownership among Evicted Households

We next study homeownership for households with an eviction history. We find that households with an eviction history are much less likely to be homeowners than other households, with a homeownership rate of 35 percent (regardless of whether their evictions were at-fault or no-fault), compared to an average of 74 percent for our full sample. Put another way, however, this means that more than one-third of households with past evictions are homeowners, suggesting that many evicted households subsequently overcome challenges and become homeowners.

Credit Access for Evicted Households

The next chart presents credit-access results for households, broken down by eviction history and homeownership status.

We can see that among current renters, those with past evictions are less likely to have access to credit cards and auto loans, consistent with previous work showing that eviction negatively affects credit access and consumption for several years. The bottom panel shows that renters with past evictions are more likely than other renters to have applied for credit cards in the past twelve months. Therefore, despite applying more often for credit cards, renters with past evictions are less likely to have them. This suggests that considering only whether renters have credit cards could understate the impact of evictions on credit supply to renters. Looking at student loans, we see that for current renters, there is no significant relationship between access to student loans and eviction history, potentially because student loans are generally not underwritten.

Turning to homeowners, we see no substantial differences in credit access—for credit cards, auto loans, or student loans—between those with and without an eviction history, suggesting that once an evicted household becomes homeowners, they enjoy access to credit similar to that of other households.

One final note of interest is that when we compare respondents who experienced a no-fault eviction with those who were evicted for income-related reasons, we don’t get a clear message about whether it is eviction or income instability that causes outcomes such as constrained credit access. A tentative read of the data suggests that eviction itself serves as an impediment to credit access and home ownership for those who remain renters. However, a significant share of no-fault evictees achieve homeownership and enjoy high current incomes.

Who Has Been Evicted and Why?

Eviction and Race

Our survey provides some evidence that evictions are concentrated in minority communities. Minority respondents are more likely to report that they have been evicted in the past, although this result becomes statistically insignificant when we account for income and education levels. But minority respondents are 8 percentage points more likely to report knowing someone who has been evicted, a statistically significant result even when we account for income and other demographics.

Conclusion

In summary, based on a novel survey, we find that evictions are fairly common in the United States, with about 9 percent of renters having at least one past eviction. Income loss is the most important cause of eviction, followed by property sales, remodeling, or conversion of the property to a primary residence, with the latter three drivers highlighting the importance of protecting renters from no-fault evictions. We also find that evictions are linked to reduced credit access. Despite these challenges, one-third of evicted households subsequently become homeowners.

Andrew Haughwout

Andrew Haughwout is a senior vice president in the Federal Reserve Bank of New York’s Research and Statistics Group.

Haoyang Liu

Haoyang Liu is an economist in the Bank’s Research and Statistics Group

Xiaohan Zhang is an assistant professor at California State University‑Los Angeles.

How to cite this post:

Andrew Haughwout, Haoyang Liu, and Xiaohan Zhang, “Who Has Been Evicted and Why?,” Federal Reserve Bank of New York Liberty Street Economics, July 8, 2020, 2020, https://libertystreeteconomics.newyorkfed.org/2020/07/who-has-been-evict....

Additional heterogeneity posts on Liberty Street Economics.

Heterogeneity: A Multi-Part Research Series




Disclaimer

The views expressed in this post are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Stiglitz on credit creation by banks

Published by Anonymous (not verified) on Sun, 17/01/2016 - 11:57pm in

For those who think that Joseph Stiglitz doesn't know that banks actually create credit:

"We are not in a corn economy where banks serve as an intermediary between farmers who have excess seed and farmers who want more seed. We are in an economy where banks actually create credit. And that makes a very big difference."
h/t: wonkmonk

Does the concentration of finance matter?

Published by Anonymous (not verified) on Wed, 10/07/2013 - 1:19am in

It may sound like a strange question in light of all the talk about "too big to fail" during the last few years. But, believe or not, the idea that bank concentration has an impact on real economic activity isn't the standard view. Here's from a recent blog post by NY Fed economists Mary Amiti and David Weinstein:

The notion that financial institutions are large relative to the size of economies is not something that plays a prominent role in traditional economic theory. Macroeconomic textbooks tend to treat economies as composed of representative firms that are infinitesimal in size compared to any given market. As a result, positive and negative idiosyncratic shocks [movement in bank loan supply net of borrower characteristics and general credit conditions] to financial institutions cancel out due to the law of large numbers. 

However, this representation stands in stark contrast with the reality of concentration in financial markets. A striking regularity is that a few banks account for a substantial share of an economy’s loans.

Starting from this basis, Amiti and Weinstein have examined Japanese aggregate bank lending data and other aggregates and were able to demonstrate the following: banks matter, bank concentration matters, bank lending matters. No small feat.

On the issue of bank concentration and aggregate lending, they found that

...if markets are dominated by a few financial institutions, cuts in lending due to some change in financial conditions in just a small number of banks have the potential to substantially affect aggregate lending. Moreover, if firms find it hard to find good substitutes for loans like issuing equity or debt, then it is possible for their investment rates to fall as well. 

As for their take on banks' impact on the real economy, the conclusion to their paper (on which their blog post in based) gives a good summary:

Our paper contributes to this literature by providing the first evidence that shocks to the supply of credit affect firm investment rates. We find that even after controlling for firm credit shocks, loan supply shocks are a significant determinant of firm-level investment of loan-dependent firms. This result is particularly surprising because our sample is comprised of listed companies that have, by definition, access to equity markets. Moreover, the fact that so much lending is intermediated through a few financial institutions means that idiosyncratic shocks hitting large financial institutions can move aggregate lending and investment. We show that about 40 percent of the movement in these variables can be attributed to these granular bank shocks. This means that the idiosyncratic fates of large financial institutions are an important determinant of investment and real economic activity.

And the implication for policy, according to Amiti and Weinstein, is significant. Here is the relevant excerpt of their blog post on this point:

...[P]olicymakers without detailed information on the major financial institutions are likely to have a difficult time understanding the causes of lending and investment fluctuations. A large portion of Japan’s aggregate economic fluctuations can be traced to the country’s banking problems. 

While many researchers have focused on the implications of banks being “too big to fail,” we show that even if large banks do not fail, granular bank shocks can have substantial impacts on aggregate investment. 

For example, reductions in bank capital at large financial institutions can cause investment declines by firms that would like to borrow, while recapitalization of the right institutions can stimulate investment. In sum, this study shows that what happens to large financial institutions is important for understanding aggregate investment behavior. 

While their paper looks specifically at Japanese data, the authors suggest that the overall conclusions are relevant to the situation in the US given that it too has a very concentrated banking sector.

Amiti, Mary and David Weinstein, How much do banks shocks affect investment: Evidence from matched bank-firm loan data, NY Fed staff paper 604, March 2013