• No results found

4.4 Empirical Framework and Results

4.4.1 Descriptive Analysis

Table 4.1 shows the distribution of unemployment rates across states, and how it evolved over the 1980-2007 period. Reported numbers are averages of ratios of local unemploy- ment rates to national unemployment rates over the indicated period. There are striking differences across states in unemployment rates which persist over time. Correlations be-

tween different periods are around 0.65-0.70. Southern states tend to remain below national

levels while some states in the Northeast and Midwest remain above national levels. Table 4.2 shows average annual changes in state average wages relative to national wages over the 1980s, 1990s and 2000s for high school and college graduates. Fluctua-

Table 4.1: Average Relative Unemployment State 1980-1989 1990-1999 2000-2007 Alabama 1.28 0.95 0.88 Arizona 0.94 0.96 0.94 Arkansas 1.15 0.99 1.00 California 0.99 1.29 1.15 Colorado 0.89 0.77 0.92 Connecticut 0.65 0.90 0.86 Delaware 0.76 0.77 0.75 Florida 0.89 1.05 0.90 Georgia 0.86 0.87 0.92 Idaho 1.03 0.95 0.85 Illinois 1.19 1.02 1.11 Indiana 1.12 0.76 0.95 Iowa 0.91 0.65 0.76 Kansas 0.70 0.75 0.92 Kentucky 1.22 1.01 1.10 Louisiana 1.39 1.12 1.02 Maine 0.82 1.00 0.88 Maryland 0.77 0.89 0.80 Massachusetts 0.70 1.01 0.91 Michigan 1.49 1.08 1.27 Minnesota 0.82 0.70 0.85 Mississippi 1.39 1.18 1.27 Missouri 0.97 0.87 0.99 Montana 1.00 1.00 0.80 Nebraska 0.61 0.46 0.66 Nevada 0.98 0.96 0.98 New Hampshire 0.57 0.83 0.72 New Jersey 0.83 1.08 0.94 New Mexico 1.14 1.18 0.96 New York 0.91 1.15 1.03 North Carolina 0.84 0.79 1.06 North Dakota 0.72 0.64 0.63 Ohio 1.21 0.96 1.08 Oklahoma 0.90 0.88 0.84 Oregon 1.16 1.02 1.25 Pennsylvania 1.08 1.02 0.98 Rhode Island 0.79 1.10 1.03 South Carolina 0.95 0.93 1.17 South Dakota 0.65 0.58 0.63 Tennessee 1.14 0.94 1.02 Texas 0.96 1.07 1.04 Utah 0.89 0.69 0.82 Vermont 0.69 0.81 0.74 Virginia 0.72 0.77 0.67 Washington 1.17 1.02 1.13 West Virginia 1.67 1.47 1.00 Wisconsin 0.98 0.71 0.92 Wyoming 0.91 0.89 0.72

tions in wages are generally larger during the 1980s compared to other periods for both high school and college graduates. Magnitudes of changes in relative wages of high school graduates are larger than that of college graduates during the 1980s. Figures 4.1 and 4.2 show relative unemployment rates and relative wage changes, and changes in relative un- employment rates and changes in relative wages for high school and college graduates. There is a strong association between relative wage changes and unemployment rates, and correlation between the two series is stronger for high school graduates, especially during the 1980s (-0.54 vs -0.30). Many states with persistent low unemployment rates in the Northeast show increases in high school wages, while those with persistently high unem- ployment rates, such as Michigan, Louisiana, Ohio, Indiana and Illionois, saw decreasing relative wages for high school graduates. Moreover, Figure 4.2 shows that overall pattern of relative wage changes conforms with changes in relative unemployment rates. States with decreasing relative unemployment rates tend to exhibit increasing relative wages. Given that high school graduates show larger fluctuations in their wages compared to college graduates, especially during the 1980s, it could be that they are more sensitive to changes in local labour market conditions.2

Table 4.3 shows the changes in geographic location of employment. Many states in the Midwest such as Michigan, Ohio, Iowa, Indiana, Illinois and Wisconsin, and some in the Northeast such as Massachusetts, Pennsylvania, New York and Connecticut, saw declining employment levels during the 1980s, while several states in the South and West such as California, North Carolina, South Carolina, Texas and Arizona saw increasing employment levels. This pattern generally follows throughout the 1990s and 2000s. Overall, it seems that the distribution of employment shifted from states in the Midwest and Northeast to

2Bound and Holzer (2000)nd that wages of high school graduates and those with less than high school

degree are affected more by local adverse demand shocks compared to college graduates in the 1980s. They define local markets by Metropolitan Statistical Areas, and they look at 10-year changes in wages from 1980 to 1990 using U.S. Census data.

states in the South and West during the 1980-2007 period. Changes in relative employment levels shown in Table 4.3 might reflect supply as well as demand changes, but Tables 4.2 and 4.3 show that most of the time changes in relative employment levels and changes in relative wages are positively associated.

Table 4.4 shows average annual net migration rates for high school and college gradu- ates.3 The pattern of net migration rates generally conforms with that of changes in relative

employment levels in Table 4.3. During the 1980-2007 period, populationflows to many of the Southern states while there are significant population outflows from several states in the Midwest and the Northeast. Magnitudes of population shifts are generally larger for college graduates compared to high school graduates, especially during the 1980s. These population outflows by college graduates may have dominated some of the adverse effects

of local demand shocks in these states. Moreover, for some states populationflows by high school graduates are in opposite direction of what one would expect during the 1980s. For example, Michigan, Ohio and Missouri took a big hit during this period as they are mostly concentrated in manufacturing industry. We see that there are population flows to these states by high school graduates, while there are significant population outflows by college graduates.

As shown in previous studies (Greenwood (1975), Bound and Holzer (2000), Woz- niak (2010)), college graduates are more mobile than high school graduates. However, differences inter-state mobility are decreasing over time. Figure 4.3 shows average annual

inter-state migration rates for high school and college graduates over the 1980s, 1990s and 2000s. There is a decline in mobility for both education groups, but college graduates show a more dramatic decrease in their mobility during the 2000s. Differences in inter-state mi-

3Inter-state mobility data is available in CPS staring in 1982. For a given year, net migration rate for an

education group for a state is calculated by taking the ratio of difference between total number of people who moved into the state and moved out of the state in that year, and the population for that education group in that year. Negative net migration rates indicate population outflows and positive rates imply inflows.

gration rates are around 2.5-2.6% for the 1980-1999 period, while it is around 1.6% in the 2000s.

The descriptive analysis here suggests that most states with improving (deteriorating) labour market conditions saw increasing (declining) relative wages and population inflows (outflows). Fluctuations in high school wages are greater than that of college graduates, who also exhibit larger population outflows from states with deteriorating market condi- tions (e.g. Michigan, Ohio, Wisconsin), especially during the 1980s, a period of economic downturn. Lower magnitudes of relative wage changes for high school graduates after the 1980s is accompanied by decreasing differences in inter-state mobility rates between

high school and college graduates. Declining differences in inter-state mobility and lower

magnitudes of changes in relative wages in the 2000s compared to the 1980s, could sug- gest different implications for the responsiveness of high school graduate wages to local

demand shocks compared to what is found by previous studies (Topel (1986), Bound and Holzer (2000)).

In the next subsection, I present results from log real weekly wage regressions on local demand shocks along with other observable characteristics.