Variation in State GDP Growth during the Recovery
The recovery from the U.S. recession has not been uniform across the 50 states. Recent data from the Bureau of Economic Analysis through 2011 show that 20 states still had levels of Gross Domestic Product (GDP) below their 2007 levels. Over this same period, the United States as a whole experienced essentially breakeven growth, with an overall rise of less than one-tenth of a percent.
The spread in growth rates from low to high is quite broad. The bottom five states had declines of 5 percent, while the top 5 states saw GDP expand by over 7 percent from 2007 to 2011. Regions with lagging performance included states hard-hit by the housing bust—including Nevada, Arizona, and Florida—and some manufacturing states like Michigan and Ohio. States with strong output growth included many farm and natural resource states. North Dakota is a particular outlier, having experienced real GDP growth of almost 30 percent over the four-year period.
From a geographic perspective, we see that there were pockets of weaker growth in the Southwest, the Southeast, and the Great Lakes regions, while the Central and Southern Plains and Mountain states showed relative strength.
To formalize the relationships, we examine how GDP growth varies with the structure of a state’s economy going into the last recession. We characterize states by the share of mining and agriculture activity to proxy for states focused on natural resources, by the share of construction activity at the end of the housing boom to proxy for states affected by the housing bubble, and by the share of manufacturing activity prior to the recession.
Looking first at the resource-intensive states, we see that most states have relatively low shares of mining and agricultural activity—these industries comprised no more than 5 percent of any state’s GDP in 2008. However, one can see that states with relatively high shares of such activity had a tendency to experience higher growth. The actual correlation coefficient is 0.49, indicating a modest positive relationship—as does the positive slope of the regression line in the chart. (The correlation statistic ranges between −1 and 1, with values closer to 0 indicating a weaker linear relationship, values closer to 1 a stronger positive linear relationship, and values closer to −1 a stronger negative linear relationship.)
The share of construction activity at the end of the housing boom is negatively correlated to GDP growth between 2007 to 2011. However, the correlation is weaker in this case (−0.28). The housing boom-bust states of Arizona, Florida, and Nevada all had construction shares that were well above the national state average of 4.0 percent, and GDP growth well below the national average. These three states end up driving much of the negative correlation between GDP growth and construction activity.
Finally, and perhaps somewhat surprising, there is very little correlation between the manufacturing share of activity prior to the recession and state GDP growth from 2007 to 2011. The horizontal trend line indicates that differences in manufacturing share do not explain differences in state-level GDP growth.
Some manufacturing-intensive states have experienced below-average growth (such as Ohio and Michigan), and this is linked, in part, to specialization in the automotive sector. Indeed, economic growth over the period tends to be lower in states with higher automotive shares. (The correlation coefficient is −0.27.)
State growth relates directly to changes in the labor market, as well. With respect to employment growth, the link is relatively tight. Employment growth and state GDP growth have a correlation of 0.81 over the 2007 to 2011 period. A key difference between the two variables is that while many states’ GDPs have fully recovered from the recession, few states’ employment levels exceed their 2007 levels. In fact, only three states (Alaska, North Dakota, and Texas) and the District of Columbia have had positive employment growth over the period, while 30 states had positive GDP growth. These patterns reflect the fact that productivity has expanded over the period, allowing firms to increase production with fewer employees.
The relationship between the change in the unemployment rate and state GDP growth, on the other hand, while negative (with a correlation coefficient of −0.52), is clearly not as tight as the relationship between employment growth and state GDP. For example, both Ohio and Massachusetts observed increases in unemployment rates of roughly 3.0 percentage points over the four-year period. However, these states had quite different GDP growth rates, roughly −5.0 percent and +5.0 percent, respectively, and quite different employment growth rates, −6.0 percent and −2.0 percent. Why did Ohio’s unemployment rate change in a similar way to Massachusetts, even though it experienced much weaker GDP and employment growth? In part, it is due to the fact that Ohio’s labor force has declined while Massachusetts’s labor force has grown modestly. With the decline in the overall labor force, Ohio has not had to create as many jobs to bring its unemployment rate down—a point sometimes overlooked when comparing changes in regional unemployment rates.