Basolo, Huang / CITIES AND ECONOMIC DEVELOPMENT

335

between the items in the index are strong for our data with the index producing a Cronbach’s alpha of .89.

The regression models also include political factors that reflect the power structure in the city as well as institutional variables. The power factors are the influence of business and the level of sup- port for economic development from elected officials. These variables are the perceptions of respondents to the survey. The institutional variables are the form of government in the city and the existence of formal economic development planning.

Finally, we include several control variables in the models: type of city, central or not,^{4 }1990 population, 1980 to 1990 population growth rate, and geographic region.

# REGRESSION ANALYSES

The analyses consider factors influencing total expenditures on economic development by cit- ies. Approximately 8% of the cities had zero expenditures on economic development, and the dis- tribution of the variable was positively skewed. We chose to model the dependent variable in two ways. First, we created a dichotomous expenditures variable: Either the city spent dollars on eco- nomic development or not. For this model, we use logistic regression.^{5 }Second, we removed the zeros and took the natural log of expenditures to produce a near normally distributed variable. We use linear regression for this analysis. This model is conditional because it considers only cities that spent on economic development.

Prior to multivariate analysis, we inspected the data for potential problems. A few missing val- ues on several independent variables were filled using the relevant mean or mode. The distributions of categorical independent variables were good; however, several of the continuous variables had skewed distributions. We used a natural log or inverse transformation for these variables.

The results of the logistic regression are presented in Table 5. The intercity competition coeffi- cient is negative and statistically significant (p = .05). In other words, as competition increases in a region, cities are less likely to spend any funds on economic development. Elected officials’ sup- port for economic development had a positive relationship with the propensity to spend, and the coefficient is statistically significant (p = .01). The coefficient for economic development plan also was positive with a significance level of .10. A strong mayor-council form of the government is negatively associated with spending on economic development. The coefficient for total popula- tion is positive and highly significant (p = .01). Finally, the South (p = .10) and West (p = .05) regions were less likely to spend any funds on economic development compared to the Midwest.

The linear regression considers only cities that spent funds on economic development. The results from this model are shown in Table 6. The population-needs variable is positive and highly significant (p = .01). Therefore, as the needs in cities increase, cities tend to spend more on eco- nomic development. Similar to the results from the logistic regression model, the coefficients for elected officials’support and formal economic development planning are positive and statistically significant (p = .05). The population and growth rate are positively associated with economic development expenditures, and the coefficients are statistically significant (p = .01). Of the region variables, only the South has a significant coefficient (p = .10), and the estimate has a negative sign, indicating the South spends less on economic development than does the Midwest. Contrary to public choice predictions, the competition variable was not significant. Overall, the model explains 36% of the variation in local economic development expenditures.

DISCUSSION

The analyses offer little support for the public choice thesis. In fact, neither model found inter- city competition positively associated with economic development spending. The logistic model, however, is consistent with Schneider’s (1989) interpretation of public choice and his analytic results. In our analysis, higher levels of competition are related to the absence of spending on local

. . . neither model found intercity competition positively associated with economic development spending. . . . In our analysis, higher levels of competition are related to the absence of spending on local economic development.