...

Comments on Charles, Hurst,  and Notowidigdo Thomas Lemieux, UBC October 18, 2012

by user

on
Category: Documents
21

views

Report

Comments

Transcript

Comments on Charles, Hurst,  and Notowidigdo Thomas Lemieux, UBC October 18, 2012
Comments on Charles, Hurst, and Notowidigdo
Thomas Lemieux, UBC
October 18, 2012
Unemployment rate, Canada vs U.S.
14
12
Canada
10
8
6
4
United States
2
0
Canada
United States
Trends in Employment of Non‐College Men, Canada
0.3
0.2
0.1
0
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Manufacturing
Construction
Manuf+Constr
Exogeneity of the shocks: manufacturing
•
Bartik instrument/manufacturing shock does an excellent job predicting change in manufacturing employment at the MSA level
– R square of .5, first‐stage coefficient of 1.
•
One important finding is that the manufacturing shock also has a large impact on construction employment
– Coefficient half as large as in manufacturing – Suggests that the loss of “good jobs” has large indirect effects on the local economy
•
Raises some potential concerns about the Bartik instrument
– Some manufacturing industries (cement, cabinet making, etc.) may be closely connected to house building. Try without these industries.
– May be picking up long run trends (Cleveland vs San Jose) instead “shocks”. Try departures from prior trends instead of straight changes at the national level for 2000‐07.
– Could try “one level up” in terms of shocks: exchange rate movements, “China syndrome”, etc.
Exogeneity of the shocks: housing
• The large effect of the manufacturing shock on local house prices suggests many other local shocks may be creating a spurious correlation between house prices and overall employment
• Good instrument besides land availability would be great…
• Short of that, more could be done using the boom and bust period
– Assume other local shocks can be captured by local trends from 2000 to 2010
– Can control for this by pooling 2000‐07 and 2007‐2010, and including period and MSA dummies in the regression.
Partial vs. total effect of manufacturing shocks
• It would be important to show both the partial (usual regression coefficient) and total effect (add indirect effect through house prices) of manufacturing shock
• Total effect is what you get without controlling for house prices.
– Wrong thing to do if manufacturing shocks and house prices are both exogenous, but happen to be correlated.
– So you need to assume that the correlation is entirely due to a “causal” effect of manufacturing shocks on house prices
• Appendix Table A12 suggests this matters a lot
Housing shocks and college enrollment
• Interesting but separate issue
• Would be interesting to (eventually…) investigate in more detail short run vs. long run impacts:
– Paper shows evidence of effects on current enrollment
– Do we get a long run effect (lower education) on cohorts that were exposed to larger shocks
• The timing illustrated in Figure 2 is not that convincing since we are looking at a “stock” variable (education > high school for those age 18‐29)
• Slow to build up so we may be catching the impact of earlier shocks
Contemporaneous Enrollment Rates, October CPS, Men
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
1980
1985
1990
18‐19
1995
20‐21
2000
22‐24
25‐29
2005
2010
Contemporaneous Enrollment Rates, October CPS, Women
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
1980
1985
1990
18‐19
1995
20‐21
2000
22‐24
25‐29
2005
2010
Interactions effects between housing booms and manufacturing shocks
• The figures suggest significant interaction effects between the two shocks:
– Figures 7 and 11: Impact of manufacturing shocks on employment is smaller when we have housing booms
– Figure 8: Impact of manufacturing shocks on wages is larger larger when we have housing booms
• Very interesting pattern but not clear this can be captured by the model
Fly UP