skip navigation

Thursday, August 28, 2014

[ – ] Text Size [ + ]  |  Print Page

Cascade: No. 58, Summer 2005

Per Capita Personal Income Differences Among States

The Declaration of Independence maintains that “all men are created equal.” Arguably, this is far from the case when it comes to our states. The various regions of our country and the states within them have grown at different rates over time. This variation in the pattern of growth in our regions and states has important implications for professionals in the public and private sectors. The sources that fuel these differences are of particular concern to economists and planners interested in regional economic development and growth. A number of studies have attempted to address the variation in economic performance by state. A recent paper by John E. Connaughton and Ronald A. Madsen of the University of North Carolina at Charlotte adds to this body of literature.1 Connaughton and Madsen's aim is to identify “factors that explain the levels of and differences in state per capita personal income (PCPI) and to examine the question of how the influence of these factors has changed over time.” The following summarizes their study.

Previous Studies

Connaughton and Madsen build upon earlier studies on economic performance across states, with special attention to the factors that give rise to underlying differences. On balance, this research tends to show a convergence in state productivity over time, even though the period studied varies and the variables identified as influencing state performance differ. One study, for example, “specifies differences in the gender composition of the labor force, differences in industrial mix, differences in human capital, and differences in technology or physical capital to explain forces influencing productivity and the rate of convergence among states.” 2 Another study discovers that low taxes and strong support for higher education were influential in explaining the variation in economic growth among states over time. Other researchers find that separating the impact of manufacturing employment from service-sector employment proved enlightening when explaining state performance over time: manufacturing employment had a decreasing role, while service-sector employment played an increasing role.

Despite the useful insights offered by the previous studies, Connaughton and Madsen find that the issue of the impact of states' racial composition on state performance, as measured by per capita personal income, is not well developed. They note that in prior studies “the impact of the racial composition has focused on the intra-state and intra-region income dispersion, not the level of state PCPI.” Moreover, there is a common presumption that the percentage of the population that is black and state per capita income levels are negatively correlated. This notion seems to be somewhat supported by one study that finds a nonwhite population variable to be negative and “weakly significant” in explaining the convergence of state per capita personal income.

However, one researcher disputes this perception and points to a Census Bureau study that indicates that the real per capita income of whites rose 13 percent between 1989 and 1999, while the real per capita income of blacks rose 24 percent. Connaughton and Madsen point out that there is a wide variance in the racial composition of state populations and the impact of this variation may be instrumental in explaining the difference in state per capita personal income levels; but it is unclear if the impact is negative, positive, or neutral.

Data and Methodology

Connaughton and Madsen use state-level decennial census information for the years 1950 through 2000. They employ regression analysis (see table below) to explain the differences in state per capita personal income levels over time.3 Connaughton and Madsen rely on two estimation approaches. First, they estimate the effects of a set of demographic, human capital, and industrial structure variables, and a set of regional dummy variables on the differences in real state per capita personal income levels for each of the years 1950, 1960, 1970, 1980, 1990, and 2000 (the effects of the regional variables will not be reported here).4 This allows Connaughton and Madsen to identify factors that have significant effects in each year as well as to compare the changes in the influence of these factors over time. Second, they estimate a similar regression using the data for all six census years taken together with the addition of a set of variables that reflect the census years.5

They rely on this approach to ferret out any results and patterns that are significantly different from those reflected in the regressions for the individual census years.

Variables, Coefficient Signs, and Statistical Significance
 
Regressions for Census Years
Variables
1950
1960
1970
1980
1990
2000
1950-2000
Percent African American
N
N
N
P
N
P
P
Percent Urban
P*
P
P*
P
P*
P*
P
Percent High School Education
P
P
P
P*
P
P
P
Percent College Education
P*
P*
P*
P*
P*
P*
P*
Percent Employment in Manufacturing Sector
P
P
N
N
N
P
N
Percent Employment in Service Sector
P*
P
P
N
P
P
P*

P = positive sign on coefficient; N = negative sign on coefficient; * = Variable's coefficient is statistically significant

Results and Conclusions

The results of Connaughton and Madsen's statistical analysis are quite revealing. For the 1950 census year, three of the six coefficients of the demographic, human capital, and industrial structure variables are statistically significant. The positive coefficient (33.76) on the percent of the state's population that was classified urban in the given census year “suggests that for each one percent increase in the urban population of a state in 1950, the state real per capita personal income level increased by $33.76…holding all other things constant.� Likewise, the coefficient (64.82) on the percent of the state's adult population that had a four-year college education is both positive and significant as is the coefficient (575.13) on the percent of the state's population employed in the service sector. The results for these variables can be interpreted in a manner similar to that described above.

However, the coefficient on the percent of the state's population classified as African American in the given census year is negative but not significant. This implies that there are no systematic differences in 1950 state real per capita personal income levels associated with the state's racial composition, all other things held constant.

  • 1 “Explaining Per Capita Personal Income Differences Between States,” Review of Regional Studies, 34, 2, 2005. Both Connaughton and Madsen are in the Department of Economics, University of North Carolina at Charlotte.
  • 2 Human capital reflects the measurement of the economic value of one's skill set. One's skill set is composed of and can be enhanced by education, experience, and training.
  • 3 They chose per capita personal income for their measure of state performance because it is widely used in the literature in similar investigations and it has proven to be measured in a consistent manner over time.
  • 4 Connaughton and Madsen chose to represent per capita personal income in the various census years in 2000 dollars so the reader could have the most recent frame of reference to gauge the relative magnitude of the estimated coefficients of the variables and their how they might have changed over time.
  • 5 The 1950 census year was chosen as the omitted category for the set of census dummies.

E-Mail Notification

Find out when information for community development publications and events is released.

Contact Us

Federal Reserve Bank of Philadelphia
Community Development Studies and Education Department
Ten Independence Mall
Philadelphia, PA 19106-1574

(215) 574-6458 – phone
(215) 574-2512 – fax
info.communitydevelopment
@phil.frb.org

View All Contacts