The Effect of Health Care Cost Growth on the U.S. Economy. Regression Analysis


To further build upon the graphical analysis presented above, multivariate regression models were estimated. The key explanatory variable in the model is the interaction between the log of medical care price index and benefit share in 1987. A negative coefficient on this variable implies that industries that provided a larger share of compensation as benefits experienced a larger decline (or smaller increase) in output or employment due to rising prices.  The regressions also include year and industry fixed effects that control for secular time trends and time invariant observed and unobserved industry characteristics respectively. Since benefit share in 1987 does not vary within the same industry, its main effect is not identified in these models with industry fixed effects. Similarly, the main effect of medical care price index is not identified since it does not vary across industries in the same year and our models control for year fixed effects. In alternate models controls for sector- specific time trends and also changes in unionization status and labor productivity were included. As before, the regressions were carried out in levels as well as in logs. The results from the log specification are presented and discussed in this report.10 All models were estimated with Huber-Eicker-White (HEW) or robust standard errors to account for possible heteroskedasticity in the data (White, 1980).

In the first set of regressions, industry and year fixed effects are included apart from the interaction between the log of medical care price index and benefit share in 1987 among explanatory variables. Columns (1) – (3) in Table 10 report results from these models. The coefficient estimate on the interaction between medical care prices and share of benefits is negative and significant for all outcomes. Thus, the estimates suggest that industries with higher benefit shares were hit harder by rising health care costs.

Note that the level of unionization within an industry could affect the share of benefits, and also affect employment and output. Similarly, labor productivity could affect aggregate output and employment. Omitting these variables could therefore bias the regression estimates. Hence, additional controls for unionization (measured by percentage of workers who are affiliated to unions) and labor productivity (measured by value added per worker) are now included into the model. Unionization is measured at the sectoral level, and productivity is measured at the level of the specific industry. These results are reported in columns (4) – (6) of Table 10.

Even after controlling for unionization and productivity, rising health care costs still seem to have a sustained adverse effect on industries that provide a larger share of compensation as benefits. In addition, unionization seems to be positively associated with employment growth although it has no significant effect on output or value added. Labor productivity has a small but significant negative effect on employment, and a small yet significant positive effect on output and value added.

Another source of bias is that sectors of the economy with high benefit shares might be experiencing an economic downturn for external reasons but this downturn coincided with rising medical care prices.  To account for this source of bias, additional regressions are estimated that control for sector-specific time trends. Five such sectors were defined – (1) manufacturing, (2) finance and services, (3) wholesale and retail trade, (4) transportation, communications and utilities, and (5) agriculture, mining and construction. These results are reported in columns (7) – (9) of Table 10. The results clearly show that even after taking sector specific time trends into account, industries with a higher share of benefits suffer greater employment and output loss due to rising medical prices. Once again, labor productivity has a significant negative effect on employment, and a significant positive effect on output. The estimates also show a significant and negative effect of unionization on employment and output.11

Overall, the regression results indicate that industries that bear a larger burden of the rise in health care costs due high benefit shares were hit harder by rising health care costs. It is likely that the labor contracts that led to high benefit shares of these industries were negotiated during a period when health care costs were not rising significantly. Consequently, these industries did not anticipate the rapid increase in health care costs when negotiating these contracts that promised large benefits. In addition, the tax exempt status of employer contributions towards health insurance premiums implied that employers were reluctant to reduce benefits in times of moderate health care cost inflation.12 Some industries are now realizing the adverse effect of these high benefit shares and are renegotiating labor contracts to reduce benefits and thus mitigate the adverse economic consequences of health care cost inflation. For example, General Motors is currently negotiating an arrangement with the United Auto Workers whereby the latter would assume responsibility for billions of dollars of retiree health care costs (Carty, 2007). Thus, in the long run some of the adverse economic consequences of health care cost inflation might be mitigated as industries reduce benefits and wages to pass some of the burden of health care cost inflation to their employees.

  (1)Log of
(2)Log of
(3)Log of
value added
to GDP
(4)Log of
(5)Log of
(6)Log of
value added
to GDP
(7)Log of
(8)Log of
(9)Log of
value added
to GDP
Table 10: Regression results
Benefit share in 87* Log of
 medical care price index
-0.158*** -0.097*** -0.113*** -0.130*** -0.101*** -0.112*** -0.065*** -0.062*** -0.066***
[0.018] [0.014] [0.016] [0.020] [0.015] [0.017] [0.018] [0.016] [0.017]
Unionization       0.009** 0.000 0.003 -0.020*** -0.018*** -0.017***
      [0.004] [0.004] [0.004] [0.007] [0.006] [0.006]
Labor productivity
(value added per worker)
      -0.001*** 0.000*** 0.000** -0.001*** 0.001*** 0.001***
      [0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Constant 7.369*** 11.899*** 11.191*** 7.218*** 11.885*** 11.105*** 7.600*** 12.078*** 11.320***
[0.054] [0.041] [0.046] [0.121] [0.115] [0.116] [0.142] [0.120] [0.127]
Observations 741 741 741 741 741 741 741 741 741
Number of Industries 39 39 39 39 39 39 39 39 39
R-squared 0.15 0.69 0.63 0.26 0.70 0.64 0.35 0.73 0.68

Robust standard errors in brackets.

* significant at 10%;
** significant at 5%;
*** significant at 1%

View full report


"report.pdf" (pdf, 309.5Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®