Measurement Reliability Results .1 Descriptive Analysis

83 for increases and higher costs of providing services. The mean of tax per capita is 2,409 and the standard deviation is 907. The mean revenue per capita is 5,308 with a standard deviation of 1,967. Finally, expense per capita has a mean of 5,164 and a standard deviation of 1,456 all numbers rounded.

4.4.2 Measurement Reliability

Measurement reliability is the extent to which a measure is free from random measurement error. Wang et al 2007 proposes three criteria to test the reliability of these indices. First, correlation should be present between the financial indicators used to make up each index. Second, there should be correlation between the indices themselves. This correlation assesses the extent to which they are measuring different dimensions of the same concepts financial condition and fiscal stress. Third, the extent to which reporting-related differences affect the reliability of individual state financial ratios should be either minimal or correctable. The financial indicators making up each index are highly correlated at a significant level, meeting the first criterion of measurement reliability. The correlation between the three cash solvency index financial indicators is high. The current ratio is highly correlated with the quick ratio r=0.9969, p0.000 and the cash ratio r=0.9691, p0.000. The correlation between the quick ratio and the cash ratio is also high r=0.9749, p0.000. For the budget solvency index, the operating ratio and surplus per capita are highly correlated r=0.9111, p0.000. For the long-run solvency index, the financial indicators are also highly correlated, although the correlation is not positive for two of the three indicators. Long-term liability ratio and net asset ratio are highly negatively correlated r=-0.9730, p0.000. The net asset ratio and long-term liability per capita r=-0.7039, p0.000 are also highly negatively correlated. Long-term liability ratio and long-term liability per capita are highly positively correlated r=0.7563, p0.000. In the final solvency index, service-level, the financial indicators are also 84 highly correlated. Tax per capita is highly correlated with revenues per capita r=0.6843, p0.000 and expenses per capita r=0.6982, p0.000. Revenue per capita and expenses per capita are also highly correlated r=0.8225, p0.000. The second criterion of measurement reliability is that each dimension of solvency is related. Table 4.3 shows the correlation and significance of the association between each of the indices. For the most part, the association between each index is as expected. Cash solvency is positively associated with budget solvency, so states with more current assets are also more likely to have balanced revenues and expenses. Both cash and budget solvency are associated with long-run solvency, indicating that states with higher cash and budget solvency have higher long-run solvency and lower long- term debt levels. Service-level solvency is also significantly related to long-run solvency. The association between service-level solvency and cash and budget solvency is less clear. The relationship between cash solvency and service-level solvency is not significant and the relationship between budget solvency and service-level solvency is significant, but negatively associated. The data suggests that states with higher Table 4.3: Correlation Matrix for Cash, Budget, Long-run and Service-level Indices Cash Budget Long-run Service-level Cash 1.000 Budget 0.1102 0.0277 1.000 Long-run 0.2625 0.0000 0.2381 0.0000 1.000 Service-level -0.0475 0.3440 -0.1981 0.0010 0.3054 0.0000 1.000 Correlation coefficient reported with the significance level in parentheses. 85 budget solvency have lower service-level solvency. One interpretation of this relationship is that states with budget balance have higher taxes per capita, expenses per capita or revenue per capita. In this setting, having higher per capita indicators is not a problem because these states realize budget balance without the need to raise taxes or revenues. However, for states with low budget solvency, low service-level solvency presents a real problem. In this situation, a state would not have much room to improve budget solvency by increasing revenues without potentially cutting expenditures significantly. An additional measure of the reliability related to these 11 financial indicators as a measure of fiscal stress is Cronbach’s alpha. Cronbach’s alpha for the standardized values of these 11financial indicators is 0.7037 including all 50 states. Based on previous research on municipal and state financial condition and using CAFRs, a range of Cronbach’s alpha values between 0.600 and 0.800 are considered acceptable Wang et al 2007. The final examination of measurement reliability concerns differences in how each state records infrastructure assets, as allowed by the GASB 34 financial reporting model. As discussed in the previous chapter, states have the option to record infrastructure assets using typical depreciation or using a modified approach. States also had the option of delaying retroactive reporting on the value of all infrastructure assets until fiscal year 2006. For states that deferred the retroactive reporting of infrastructure assets, this delay will affect the size of their net assets, total assets, and unrestricted assets Wang et al 2007. Based on a review of state CAFRs, only four states elected to defer reporting infrastructure assets: Alabama, California, Montana, and Rhode Island. The small number of states that chose to defer reporting reduces the likelihood that this reporting difference systematically affects the dataset; however, a t-test is used to assess whether the affected financial indicators are significantly different between these states and those that did not defer reporting. The financial indicators likely to be affected by this reporting difference 86 are the net assets ratio and the long-term liability ratio. Since states only deferred reporting up to 2006 and most states started reporting assets retroactively before then, the t-tests are conducted on the difference in indicator means by year. For 2002, when all four states deferred; the mean difference for the net assets ratio between states that deferred and those that did not was 0.1321, not a statistically significant difference t=0.9695, p=0.3372. For the long-term liability ratio, in 2002 the mean difference was -0.04144, also not statistically significant difference t=-0.3583, p=0.7217. For 2003, when all four states deferred reporting, the mean difference between net assets ratios was 0.2061, not statistically significant t=1.4054, p=0.1663. The same was true for the long- term liability ratio t=-0.9938, p=0.3253. In 2004 and 2005, only two states still deferred retroactive reporting. The statistical significance for the mean difference between the 48 states that reported retroactive assets and those that did not ranged from marginally significant to significant at the 0.05 level for both the net assets ratio and long-term liability ratio. However, it is unclear if this should be interpreted as solely due to the deferred reporting, or due to other differences between these two states, California and Rhode Island, and the other 48 states. Overall, these results suggest that the practice of deferred reporting has a relatively limited impact on the financial indicators. However, to correct for any systematic differences, controlling for deferred reporting practices is warranted. The second difference in state reporting of infrastructure assets is how they report depreciation of infrastructure assets. In lieu of depreciating these assets, GASB 34 offers states the option of a modified approach in which they do not record depreciation but rather must maintain assets at an approved level and report on their ability to meet this criteria. According to Wang et al 2007, states using the modified approach may have lower expenses and higher net assets. This may affect the operating ratio, surplus per capita, net asset ratio, and expenses per capita. Twenty-three states chose the modified approach. No state changed their approach to recording infrastructure depreciation 87 between 2002 and 2009. There was a significant difference between the means of the net assets ratio t=-3.597, p=0.000 and expenses per capita t=5.236, p=0.000. The impact of modified approach is not significant for the operating ratio t=-0.3024, p=0.7625 and surplus per capita t=0.3426, p=0.7320. The mean net assets ratio is higher for states using the modified approach, while the mean expenses per capita is lower. By testing the means of the long-run and service-level solvency indices for states that use the modified and traditional approach, we find that the two groups are significantly different. States using the modified approach have a higher mean long-run solvency index value t=-4.14, p0.000 and a higher service-level solvency index value t=-4.81, p0.000. Dropping these indicators is an option for eliminating these systematic differences. However, the expenses per capita indicator provides information on the cost of providing services, without which the revenues and taxes per capita are biased. The net assets ratio captures a state’s ability to pay off long-term liabilities. In order to keep these indicators and protect against bias, the approach to recording state infrastructure asset depreciation will either be added as a dummy variable to regression analysis or fixed effects will be used to capture systematic differences.

4.4.3 Measurement Validity