Uncertainty and the
Geography of the Great Recession
Daniel Shoag (Harvard Kennedy School)
Stan Veuger (American Enterprise Institute)
Exploring the Price of Policy Uncertainty, April 2015
United States House of Representatives
April 12, 1935
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 uncertainty that you folks have
created. Let us assure industry and we will end unemployment in a short
time.
Explanations for the Severity of the Great Recession
Macroeconomists have suggested several, non mutually exclusive
hypotheses to explain the severity of the 2007-2009 recession:
Insufficient demand due to household deleveraging
Slow adjustment to structural shocks
Credit constraints due to financial sector problems
Increased policy and economic uncertainty
Regional Variation in Unemployment Rates
There is substantial regional variation in the employment losses
associated with the recession:
The five states most affected saw unemployment rise by over 6 points
The five states least affected saw unemployment rise by under 2.1 points
How does this variation relate to the various explanations introduced
above?
It is correlated with variation in housing price declines and
construction unemployment losses
Mian and Sufi (2011) and Philippon and Madrigan (2011) show that
it is
also correlated with high and steeply declining household debt to
income ratios
Gozzi and Goetz (2010) and Greenstone and Mas (2012) show that
it is
correlated with credit supply and bank balance sheet losses.
Is the Cross Section Inconsistent with the Uncertainty
Hypothesis?
“[A]n increase in business uncertainty at the aggregate level does not
explain the stark cross-sectional patterns in employment losses we
observe” – Mian and Sufi (2012)
This paper:
• Develops state-level uncertainty measures and documents their
association with employment outcomes
Shows that predetermined state government institutions affect
regional uncertainty and unemployment
•
Follows existing literature by examining relationship between
regional uncertainty and unemployment after controlling for other
hypotheses
•
Presents a simple model of hiring and firing that produces correct
predictions for the cross-section of employment levels
•
Part 1
State-Level Uncertainty Measures and State-Level
Unemployment Outcomes
Seven State-Level Measures of Uncertainty
We develop seven distinct measures of state-level uncertainty:
•
Uncertainty in the news
•
Google searches
•
Mid-session budget cuts
•
Mid-session tax increases
•
Revenue deviations
•
Coincident index deviations
•
State credit ratings
The measures are all strongly correlated with state-level changes in the
unemployment rate between 2006 and 2009.
State-Level Uncertainty
A Local Uncertainty Index
Robustness Tests
Placebo Test: Other Time Periods
Placebo Test: Other Outcomes
Part 2
Do Predetermined State Government Features Drive
Both Uncertainty and Unemployment?
Existing Institutions Affect Local Uncertainty
Same Institutions Affected Employment Outcomes
Part 3
An Assessment of Competing Hypotheses
Is Uncertainty ‘Explained’ by Housing?
Is Uncertainty ‘Explained’ by Aggregate Demand?
Is Uncertainty ‘Explained’ by Credit Supply?
Is Uncertainty Independently Important?
Part 4
A Simple Model of Hiring and Firing
Model: Set-up
• Assume an economy with multiple industries indexed by i, each populated with
a continuum of firms
• Firms receive a stochastic productivity draw each period of life that takes a
value of either
• Draw determines whether their optimal production requires high or low
employment
• When a firm’s employment matches its productivity draw, it earns a profit
equal to π; it receives zero otherwise
• Firms’ stochastic productivity draws are persistent, so probability of
receiving the same draw
• Larger value of p reflects less uncertainty
• Firms must pay a fixed cost proportional to profits Ciπ to adjust their
employment state and discount the future at rate β
• Assume that C is sufficiently low that, absent the uncertainty shock, all firms
find it profitable to adjust to their desired level of employment
Model: Steady State
Given our setup, firms solve:
We define population density within industry across states as:
We solve for ergodic distribution of firm population across states
Given that solution, we can calculate the endogenous net hiring occurring
within an adjusting industry each period:
Model: Uncertainty Shock
Now we consider the effect of a temporary (1 period) uncertainty shock
Because the shock is temporary, continuation value functions do not change,
only the odds of reaching them changes
Firms decide to adjust despite temporary shock if induced uncertainty is low
relative to the costs
After a temporary uncertainty shock:
• The adjustment rule for hiring and firing firms is symmetric
• Firms in an industry i continue to adjust if:
• Firms in an industry that does not continue to adjust fail to hire young
workers. Employment in these industries shrink at a rate of
Model: Predictions
We can now identify three cross-sectional implications of an uncertainty shock.
Uncertainty shocks are more likely to reduce employment:
1. In industries with larger baseline turnover
2. In occupations with larger adjustment costs
3. Among younger workers
Part 5
Industries, Occupations, On-the-job Training, and Age
Uncertainty and the Cross-Section of Industries
Uncertainty and On-the-job Training
Uncertainty and Age
Uncertainty and Firm-Level Hiring
Conclusions
The association between local policy uncertainty and local employment
losses is strong, making policy uncertainty a potentially important part of
the narrative explaining the depth and length of the Great Recession
This cross-sectional evidence complements the more firmly established
time series evidence supporting the policy uncertainty driver of poor
economic performance
The microeconomic model of hiring and firing presented here produces
non-trivial predictions regarding the impact of uncertainty that are
confirmed by empirical tests
State Uncertainty Measures Both Underlying Sensitivity to
National Shocks And State Gov’t Amplification
•
For each industry, regress national employment against a trend
and the Bloom, Baker, Davis Index from 1990 onward
•
Create an underlying economic sensitivity measure for each state
by weighting industry sensitivities by state industry shares