Systemic risk in the financial sector has been a ubiquitous talking point since the global financial crisis of 2007–09. Research economists have developed many systemic risk indexes aimed at measuring particular aspects of financial distress. These include institution-specific risk indexes that try to capture an individual bank’s sensitivity to financial distress, contagion indexes that try to quantify dependence across financial institutions and their equity returns, as well as liquidity and credit condition indexes, amongst a slew of other risk measures.
On their own, however, individual indexes that aim to measure unique aspects of systemic risk don’t do that well at helping to forecast economic downturns that affect a typical household. Recent work by Bryan Kelly and Stefano Giglio of the University of Chicago Booth School of Business and Seth Pruitt of the W.P. Carey School of Business at Arizona State have found a novel way to combine a large set of systemic risk indexes that show promis for predicting p macroeconomic outcomes that are of real importance for American households, including household consumption and productivity growth.
In the working paper, the authors make a first attempt at aggregating the large set of systemic indexes already present in the academic literature using dimension reduction techniques. They demonstrate that the aggregated index is a significant means for forecasting future shocks. The key to their analysis? They uncover a latent systemic risk factor that is the common driver of each of the individual systemic risk indexes, making it a powerful indicator for forecasting macroeconomic downturns.
The expectation amongst American households has been that regulators and policymakers should be able to police systemic risk and prevent it from spilling over from the financial sector into the real economy. Pursuing that goal, the researchers point out, policymakers ought to focus on those systemic risk indexes that actually provide information that is welfare-relevant for household consumption.
To quantify macroeconomic outcomes in the United States, the researchers focus on industrial production data from the Federal Reserve Board and data from the Chicago Fed National Activity Index (CFNAI), a composite indicator of overall economic activity.
The researchers present evidence that the systemic risk measures are not strongly related to the upper tail of macroeconomic events. This supports prior literature in macroeconomic tail risk that suggests that systemic risk is fundamentally an asymmetric and nonlinear phenomenon.
To operationalize their criteria for welfare-relevant macroeconomic indicators, the scholars use predictive quantile regression, which is able to estimate the impact of systemic risk on the lower tail behavior of macroeconomic shocks. Specifically, the team uses tests at the 50thand 20th quantile percentiles. The 50th percentile of macroeconomic aggregates represents their value in normal times, while the 20th percentile represents negative events that only occur one fifth of the time.
Using this predictive quantile regression methodology yielded key insights. FIrst, the researchers found significant predictive power for the 20thpercentile and little predictability for the median or 50thpercentile of their industrial productivity and CFNAI outcome variables from the large set of systemic risk indexes. Of the 22 systemic risk measures that the researchers study, 19 of these are stronger predictors for tail risk (a bad but rare event) as compared with central tendency (what happens on average).
Second, financial sector volatility is highly predictive of macroeconomic tail risk, while there is no predictive power at all in the volatility of non-financial firms.
This second stylized fact is incredibly important for building knowledge in the macro forecasting literature. Scholars have argued that aggregate equity market volatility is an important predictive component of the business cycle. Kelly, Giglio, and Pruitt make a strong case for why the financial sector, in particular, is more informative for predicting macroeconomic activity. Equity volatility in non-financial sectors, they argue, simply does not possess any predictive power for tail risk events. This finding shows the importance of distinguishing economic uncertainty in the financial sector from turbulence in other industries and sectors
Lastly, Kelly, Giglio, and Pruitt believe they are able to identify the role of monetary policy easing, as measured by the response of the Federal Funds rate to various systemic risk measures. The researchers explore whether policymakers respond to the systemic risk indices by exploring whether these indices predict changes in the Federal Funds rate. Specifically, the researchers utilize quantile regression to forecast the 20th and 50th percentile of Federal Funds rate changes.
The researchers focus their analysis on three predictor variables: financial sector volatility, a variable termed ‘turbulence’ that aims to measure excess financial sector volatility, and their aggregate estimator for the panel of all systemic risk measures. The most eye-opening result is one illustrating that the out-of-sample 20thpercentile predictive coefficient is significantly larger than the median coefficient for the set of three predictors mentioned. The analysis shows that each predictor, in its own right, is especially powerful for predicting large changes in the Federal Funds rate.
The paper sets out to get both academic economists and policymakers thinking about the empirically relevant features of systemic risk. The authors achieve this objective by linking individual–level risk indexes to outcomes on the real side of economic activity, specifically focusing on industrial productivity growth and the CFNAI. When the systemic risk measures are aggregated in an appropriate way, as the researchers validate with their dimension reduction techniques and factor estimation analysis, the result is one that illustrates a systemic risk factor that is strongly related to future macroeconomic outcomes.
The paper bears a weighty conclusion: aggregate systemic risk measure is not just related to any macroeconomic outcome, but specifically related to downside macroeconomic risks, making it particularly relevant to the conversation of how to avert future financial calamities. The paper is noteworthy in the value it adds to the literature on systemic risk measurement, particularly in its demonstration of a new measure of systemic risk that helps to inform policymakers and regulators about financial distress mapping directly onto important macroeconomic outcomes.
—Alex Verkhivker