Discussion: (0 comments)
There are no comments available.
View related content: Economics
We document the link between increased levels of economic and policy uncertainty and unemployment at the state-level during the 2007-2009 recession. The cross-sectional variation in uncertainty robustly matches the distribution of employment outcomes during this period. When we instrument for this cross-sectional variation using preexisting institutions, we find evidence for a causal role for uncertainty in increasing unemployment. A simple model of hiring and firing under uncertainty rationalizes these results, and the within-state distribution of effects across industries, occupations, and individuals is consistent with this model’s predictions. Together, these results suggest that increased uncertainty contributed to the severity of the Great Recession.
Mr. FITZPATRICK: What is the gentleman’s plan to take care of the unemployment in this country?
Mr. KNUTSON: What is my plan?
Mr. FITZPATRICK: Yes.
Mr. KNUTSON: Reassure industry.
Mr. FITZPATRICK: How?
Mr. KNUTSON: By removing all the uncertainly that you folks have created. Let us assure industry and we will end unemployment in a short time.
United States House of Representatives, April 12, 1935
I Uncertainty and the Great Recession
Did high levels of economic and policy uncertainty contribute to the large and persistent increase in unemployment from 2007 to 2009? This paper presents a broad range of evidence showing that increased levels of uncertainty were an important factor explaining the cross-sectional distribution of employment patterns during the Great Recession. To do so, we develop an indicator-based state-level uncertainty index. Using this index we find that the uncertainty-unemployment relationship is highly robust, and that higher levels of uncertainty created by pre-existing institutions lead to increased unemployment. We then develop a simple model of hiring under uncertainty and show that the employment losses associated with the index were distributed according to the predictions of the model across industries, occupations, and individuals within states.
Macroeconomists have advanced a number of hypotheses to explain the severity of the 2007-2009 decline in employment. These explanations, which are not mutually exclusive, include insufficient demand due to household deleveraging, slow recalculation or adjustment to sector-specific shocks, credit constraints due to problems in the financial sector, and the aforementioned increases in policy and general economic uncertainty. Unfortunately, as is often the case in macroeconomics, distinguishing the differential impact of these amplification channels has not been straightforward.
One aspect of the “Great Recession” that might shed light on the mechanism is the substantial geographic variation in employment losses. The five states most deeply affected by the recession experienced increases in their unemployment rates of 6 percentage points or more from 2006 to 2009 (with the largest increase, in Nevada, exceeding 7.5 percentage points). Conversely, the five states least affected by the downturn saw their unemployment rates increase by less than 2.1 percentage points. Given the importance of this geographic variation, it is desirable that theories of the recession are consistent with this cross-sectional pattern.
The differential effect of the recession across places was not random. In line with a explanation centered around structural sectoral shifts, states with larger housing price run-ups and declines suffered the largest employment losses. More directly on point, overall employment losses across states and counties are highly correlated with employment losses in the construction sector.
A number of important papers have demonstrated that geographic variation in household deleveraging and weaker demand are also correlated with employment losses. Mian and Sufi (2011), in a framework that is underpinned theoretically by Philippon and Midrigan (2011), show that employment losses are most severe in areas with initially high and subsequently falling household debt-to-income ratios. They analyze data from counties with large household balance sheet shocks and claim that lessened aggregate demand was responsible for the majority of the job losses between 2007 and 2009.
Another theory, that credit constraints caused by financial sector problems lengthened the recovery (e.g. Guerrieri and Lorenzoni (2011) and Chodorow-Reich (2013)), does not necessarily predict such wide variety in regional outcomes. However, work by Gozzi and Goetz (2010) and Greenstone and Mas (2012) find that local credit crunches for small businesses did indeed lead to employment and wage losses between 2007 and 2009.
A recently popular explanation for the significant duration of the 2007-2009 recession’s recovery is an increase in policy and economic uncertainty. Widely discussed in the popular news amidst analyses of the impact of Federal Reserve policy, health care reform, the rise of the Tea Party movement (see Madestam et al., 2013), debt ceiling disputes and state and federal spending levels, policy and economic uncertainty have also received attention from researchers looking into their possible effects on the U.S. economy during the aftermath of the recession. In a leading paper in this literature, Baker, Bloom, and Davis (2013) create an indicator-based measure of policy uncertainty using newspaper mentions, tax code provision expirations and forecaster disagreement. They show that higher indicator uncertainty from 2008 on was associated with a deeper and longer recession.
In their analysis of news and government documents, Dominguez and Shapiro (2013) look to see how the slow recovery was anticipated, and find that the political “stalemate” in the US contributed to the length of the recession, as did shocks from Europe. Similarly, Bachmann and Sims (2012) establish that consumer and firm “confidence” is of the utmost importance during downturns. Schaal (2011) is able to reproduce many of the dynamics of the Great Recession by introducing uncertainty shocks into a dynamic search model of heterogeneous firms, while Stock and Watson (2012) use a dynamic factor model to establish that heightened uncertainty worsened the recession significantly. That said, in its simplest exposition, the uncertainty channel does not predict a wide spatial distribution of outcomes. This has led some to argue that the policy uncertainty channel is not consistent with a central feature of the recession. For example, Mian and Sufi (2012) claim that “an increase in business uncertainty at the aggregate level does not explain the stark cross-sectional patterns in employment losses we observe,” and Bachmann and Bayer (forthcoming) argue that idiosyncratic firm-level risk shocks that are consistent with the procyclicality of the dispersion of investment rates cannot explain output variations over the business cycle.
This paper serves to counter such claims, and to present cross-sectional evidence in support of the uncertainty channel. We create local measures of Baker-Bloom-Davis type indicator uncertainty from 2006 through 2009. We find that increases in local uncertainty over this period are strongly correlated with the effects of the recession, and that the correlation between uncertainty at the state level and employment losses is highly robust across alternate measures. While there is certainly a feedback loop between economic outcomes and uncertainty, we show that increases in local uncertainty are partially driven by preexisting state institutions, and that these pre-determined uncertainty amplifications cause unemployment increases. The uncertainty channel also remains strongly correlated with unemployment increases in our data in regressions that control for other mechanisms.
Moreover, we show that even when controlling for the aggregate local outcome, uncertainty affects the cross-section of industries, workers, and occupations within states in the manner predicted by a standard model. Our baseline results suggest that if uncertainty levels in all states had been at those of the five states facing the lowest levels of uncertainty in 2009, that would have been associated with a national unemployment rate that was about 1.4 percentage points lower.
The key lessen from these findings taken together is that, like the structural and demand driven channels, the uncertainty explanation is consistent with the geographic pattern of the recession. While it is hard to quantify the exact causal effect of this amplification mechanism, and to separate the impact of uncertainty from that of first-order shocks to the economy, these findings are important in cautioning researchers not to dismiss the uncertainty channel in contributing to the length and depth of the Great Recession. It also suggests that more research on the interaction of multiple channels would prove beneficial.
The remainder of this paper is structured as follows. In Section II, we develop regional measures of uncertainty, document their association with employment outcomes, and use them to construct a regional uncertainty index. We then proceed to show, in Section III, that predetermined state government institutions affect regional uncertainty. By using these institutions to instrument for uncertainty, we show that higher levels of uncertainty cause higher levels of unemployment. After that, in Section IV, we examine the relationship between regional uncertainty and unemployment levels after controlling for competing explanations for high post-2006 levels of unemployment. In Section V, we present a simple model of hiring and firing by firms that face varying levels of uncertainty to derive predictions for the cross-section of employment levels. We confirm these predictions in Section VI at the industry, individual, and occupation level, and present survey-based evidence that provides direct support for the uncertainty mechanism. In Section VII we discuss our results and conclude.
There are no comments available.
1150 17th Street, N.W. Washington, D.C. 20036
© 2015 American Enterprise Institute for Public Policy Research