For the longitudinal analysis, data on all 50 states over a period of 20 years (1985-2004) are used to examine how the increase in per capita health expenditure affects per capita GSP and unemployment rate. Table 5 provides descriptive statistics for these variables.
|Per capita health care expenditure
(thousands of dollars)
|Per capita GSP (thousands of dollars)||26.62||8.17||11.82||63.00|
|Unemployment rate (percent)||5.44||1.65||2.30||13.40|
Next, the average change in each variable over these 20 years is computed separately for each state, and these mean changes over time (averaged across all 50 states) are presented below in Table 6.
|Change in per capita health care expenditure (thousands of dollars)||50||0.20||0.03||0.14||0.27|
Change in per capita GSP
(thousands of dollars)
Change in unemployment rate
These changes over time in various outcomes are now compared with the change over time in per capita health care expenditure. Specifically, the average change over time for each outcome is plotted against quartiles of the change in per capita health expenditure (Figure 3) over the same period.
Figure 3: Average Change in Aggregate Economic Indicators:
By Quartiles of Change in Per Capita Health Expenditure
Note: Changes in thousands of dollars (per capita GSP), and in percentage points (unemployment rate)
Q1,mean change: per capita GSP--1.06, unemployment--- -0.10; Q2,mean change: per capita GSP--1.06, unemployment--- -0.13; Q3,mean change: per capita GSP--1.15, unemployment--- -0.09; Q4,mean change: per capita GSP--1.23, unemployment--- -0.07
Figure 3 suggests that states that experienced greater increases in per capita health expenditure during 1985-2004 also experienced higher growth in per capita GSP. For example, while the growth in per capita GSP is $1,060 in the bottom two quartiles of the change in per capita health expenditure, it is over $1,200 in the top quartile. For unemployment, we do not find such a clear pattern, although states in the upper two quartiles of increase in per capita health expenditure do seem to have experienced a smaller decline in unemployment.
Finally, multivariable regression models are estimated using longitudinal data. .The regression models include state and year fixed effects that respectively control for time invariant state-level heterogeneity and national trends in health care costs and outcomes. Table 7 reports results from these regressions for each outcome. All regressions are estimated with Huber-Eicker-White (HEW) or robust standard errors to account for possible heteroskedasticity in the data (White, 1980). The regressions are estimated in levels as well as in logs, and the results from both specifications are reported in Table 7.
|(4)Log of unemp.
|Per capita health expenditure||1.842***||0.608***|
|Log of per capita health expenditure||0.101**||0.096|
|Number of States||50||50||50||50|
Robust standard errors in parentheses;
* significant at 10%;
Note: There was missing information on unemployment rate for Connecticut in 1999, resulting in 999 observations instead of 1000.
The estimates in Table 7 suggest a statistically significant and positive relationship between per capita health expenditure and per capita GSP, even after controlling for state and year fixed effects. Specifically, the estimate from the log specification suggests that a ten percent increase in per capita health expenditure leads to a one percent increase in per capita GSP. The results for unemployment rate are somewhat mixed with the estimate from the log specification showing no significant relationship.
In general, these results might suggest that higher health care expenditure was beneficial to economic growth; however, one needs to be careful in interpreting the findings. As reported in the literature review, rising health care costs can have both a positive and a negative impact on the economy, and the results above are possibly due to reverse causation, whereby rising incomes over time lead to a higher proportion of GDP being spent on health care. Several macroeconomic studies have indeed shown such a link between per capita GSP and health care costs. Consequently, this analysis might underestimate the harmful effects of health care cost growth on aggregate economic indicators, if higher incomes do lead to increased expenditure on health care. Therefore, these findings do not rule out the possibility that health care cost growth has a negative effect on the economy, but the evidence so far is insufficient to reach any firm conclusion.