Ballotpedia:Who Runs the States, Part Three, Methodology
The two major political parties claim that their policies will lead to better outcomes. What does the data show?
At Ballotpedia, we explored these issues in a three-part study, Who Runs the States.
This page contains the section of Part Three pertaining to the Methodology.
Partisanship/Quality of Life Regression Methodology
In our analysis of trends in rankings, we used several different regression models to explore tendencies and potential relationships. In our examination of the aggregate state rankings, we used an ordinary least squares bivariate regression to find correlations between the state performance rankings and the degree to which the state was Republican or Democratic according to our scoring method. The partisanship score functioned as the independent variable, and the aggregate state life quality rankings served as the dependent variable. We did not control specifically for region or population.
Another test of the aggregate rankings used a probit regression, which is designed to show relationships between dependent variables with a dichotomous outcome (0 or 1) and independent variables that may be any real number. All state governments in each year were then coded as one (1) for having a trifecta government (Democratic or Republican) or zero (0) for not having a trifecta government. These values were then regressed with the state’s performance ranking for that year using the probit regression. This probit regression was also performed with a two-year lag on the dependent variable, which correlates a performance ranking in a given year with a trifecta ranking two years prior. This was done to test the difference between having a trifecta or a mixed government on quality of life.
Using this same dichotomous variable for trifecta or non-trifecta governments, we used a panel data series regression, which allows us to identify separate time periods and changes over time, to find trends in the performance of trifecta and non-trifecta governments. With this regression, we were able to incorporate the fixed effects associated with each state and control for more lurking or unaccounted for variables than our earlier analyses. However, the available data in rankings form, relative to other states or types of government, severely restricts the analytical power of this type of regression. For our independent variables, we used both the same-year trifecta ranking and the two-year lagged trifecta ranking for the independent variables in two separate regressions, with performance ranking relative to other states in that year as the dependent variable.
Finally, we used the panel data series regression with fixed effects for the comparison between the overall rankings of the states and the partisan coding of the states in each year. These metrics provide slightly stronger analysis of the data than the previous panel regression, though they do not measure the specific influence of trifectas. In these regressions, the dependent variable was the state performance ranking for each year, and the independent variables were the state governments coded one through nine (1 through 9) as outlined in the “Overall Partisanship” section, both in the same year and lagged two years in separate regressions. Again, measuring coding limits the analytical effectiveness of these regressions, but still provides some sense as to the broader trends in the data.
- Ballotpedia:Who runs the states
- Ballotpedia:Who Runs the States, Part One: State Partisanship
- Ballotpedia:Who Runs the States, Part Two: State Quality of life Index (SQLI)
- Ballotpedia:Who Runs the States, Part Three: Overlaying State Partisanship and State Quality of Life (SQLI)
- Part 1 Full report PDF
- Part 2 Full report PDF
- Part 3 Full report PDF
- State government trifectas