# Understanding Growth Patterns in US Health Care Expenditures

Understanding Growth Patterns in US Health Care Expenditures Abstract We study the steady upward trend of Health Care Expenditures (HCE) over GDP for a sample of OECD countries between 1970 and 2007. Although the United States is clearly an outlier, almost all of the additional increase in US HCE happened during the 1978–1990 period. We perform two growth accounting exercises to explore sources of variability of HCE over GDP across countries. In the first growth accounting exercise based on value added we find that factor accumulation is unable to replicate the observed growth patterns. We also show that the additional increase in markups in the US corporate medical sector mimics well the ratio of HCE over GDP in the United States. This suggests that differences in the relative price of health care—rather than technology, product quality, and factor accumulation—could explain the divergent growth patterns of HCE over GDP across these countries. In the second growth accounting exercise, we filter out prices from HCE over GDP, and confirm that there is very little variability for the product quality residual to explain the variation in HCE across countries. 1. Introduction One of the most noteworthy stylized facts for the US economy is the pronounced upward trend in the share of health care expenditures (HCE). Indeed, according to OECD data HCE accounted for about 7% of nominal GDP in 1970, whereas it is now about 18%. Mounting US medical expenditures have become an issue of national concern and a continuing challenge for policy makers (e.g., the Clinton Health Care Plan of 1993; the Bush Medicare Prescription Drug, Improvement, and Modernization Act of 2003; and more recently the 2010 Obama Affordable Care Act). We use a comparison group of ten other OECD countries with high quality data: Australia, Canada, Denmark, Finland, France, Germany, Ireland, Japan, Spain, and the United Kingdom. Our quantitative analysis aims to assess deviations of US medical expenditures from common trends over the sample period 1970–2007. We shall focus on the US medical care expenditure gap, which we define as the US share of HCE in nominal GDP over the average share of HCE in nominal GDP of the other OECD countries. Figure 1 illustrates the evolution of the US medical care expenditure gap for the sample period. The United States was among the top health care spenders in 1970 but still far from being an outlier. The US medical care expenditure gap displays a pronounced upward trend over the 1980s. The gap remains flat during the eight-year period of the Clinton presidency, and it goes up slightly at the beginning of the 2000s. Hence, this figure suggests the existence of three well differentiated periods: The 1970–1977 period: The US medical care expenditure gap hovered around 1.25. The 1978–1990 period: The US medical care expenditure gap increased steadily from 1.23 to 1.68. The 1991–2007 period: The US medical care expenditure gap roughly stabilized around 1.70. Figure 1. View largeDownload slide The US medical care expenditure gap 1970–2007. Source: OECD Health Data (June 2014). Figure 1. View largeDownload slide The US medical care expenditure gap 1970–2007. Source: OECD Health Data (June 2014). Of course, the 1978–1990 transition episode is certainly puzzling. Indeed, it does not seem plausible to explain the sharp increase in HCE over such a limited time span by some aggregate variables with smooth long-term trends such as GDP, life expectancy, the size of the elderly population, defensive medicine, the prevalence of some modern health habits and trends (e.g., fast food and obesity), and new medical treatments. The widespread use of better, more advanced technologies and product quality has been blamed for the high share of HCE in GDP in the United States. The evidence supports the idea that new technologies may diffuse quickly among countries with commensurate income levels. Besides, the increase in the US medical expenditure gap occurred just over a decade, and such pattern of growth will be hard to replicate by simple models of technology diffusion. Regarding higher product quality in the US health care sector, the United States typically trails behind in OECD rankings on various health care outcomes and indicators. This is still compatible with the widespread use of more sophisticated and less cost-efficient technologies in the United States. However, our study shows that inflation in the US medical sector is driven by large markups, and price increases do not come along with extra expenditures of resources. Indeed, in our first growth accounting exercise in what follows we show that the US medical care expenditure gap implied by factor accumulation runs in the opposite direction. One major difficulty in uncovering US HCE growth patterns is the high degree of uniformity in the categories of sources and uses of funds. Figure 2(a) breaks down HCE into various categories by source. Some public programs such as Medicare and Medicaid have gained prominence at the expense of private funding and out-of-pocket expenditures, but such extra growth seems too small to account for the general evolution of HCE. Figure 2(b) breaks down HCE into various categories by use. A substantial increase in the categories of other personal health care and prescription drugs is observed, as well as a slight decline in physician and clinical services, and hospital care.1 Again, these changes are just small departures from overall trends in HCE. Thus, for the matter in question, we would need to identify a group of variables that jointly drive all the expenditure categories. Two distinct candidates are prices and technology. Figure 2. View largeDownload slide The evolution of US health care expenditures by sources and uses 1980–2007. Source: US Department of Health and Human Services – Centers for Medicare and Medicaid Services. Figure 2. View largeDownload slide The evolution of US health care expenditures by sources and uses 1980–2007. Source: US Department of Health and Human Services – Centers for Medicare and Medicaid Services. To gain a further understanding of the macroeconomic factors generating the divergent patterns of the US medical care expenditure gap, we perform a growth accounting exercise based on value added. Three main methodological issues motivate this exercise. First, we shall be concerned with growth in factor quantities and this is compatible with cross-country differences in price and quality levels. Second, as observed in Figure 1, most of the incremental growth in US medical expenditures occurred during the 1978–1990 transition period, just over a decade. As discussed in what follows, it is unlikely that a large component of growth in HCE could be attributable to a sharp break in non-observable factors such as technological progress and product quality during such a short period of time. Further, this relatively short transition period is characterized by large increases in nominal and financial variables in the health care sector such as product markups. And third, our main analysis relies solely upon factor accumulation over quantities of inputs; hence, we need not take a stand on pricing policies, monopolistic behavior, the size of the government sector, and many other forms of market intervention and regulation at the aggregate level. We later address product quality and price effects as well as some rather complex measurements issues. In our sample period, capital displays higher growth in the United States whereas labor displays higher growth in the OECD. From our growth accounting exercise we conclude that there should not be much difference in the evolution of the actual shares of HCE in nominal GDP for both economies. That is, the different growth patterns of US health care expenditures cannot be explained by aggregate factor accumulation. The sizable wedge between the contribution of value added and HCE in the United States motivates various supporting exercises. We first look at the evolution of markups, gross profits, and stock returns over a sample of US companies in the health care sector. Excessive levels of profitability and of stock market returns usually happen at times of high growth in HCE, which lends credibility to the idea of existence of monopoly power and other market frictions in the US health care sector not generally observed in the other countries. Moreover, an extensive literature claims that the United States trails behind comparable OECD countries in terms of health outcomes and related health care indicators. Therefore, the different patterns of growth in HCE in the United States cannot be backed by corresponding changes in product quality. We should remark that we analyze variations of HCE over a group of countries with commensurate technological development and data quality, and we consider expenditure shares to condition by cross-country income levels. Several authors have pointed out that early diffusion of new technologies is frequently linked to similar GDP levels (Comin and Hobijn 2004, 2010; Smith et al. 2009). In addition, when we adjust for prices we observe that US real HCE does not grow much faster than US real Total Consumption. In fact, the evolution of US real HCE is fairly well matched by value added from factor accumulation, as well as by similar indicators of health care quantities in the other OECD countries. Hence, our main analysis does not depend on prices, but our basic findings do not change when we filter out prices and common quality trends. To understand the stability of real HCE we explore potential drivers of US medical care prices such as hospital and professional services, and pharmaceutical products. Physicians’ services, however, do not appear to be a major source of the continued inflation in the medical sector. Price markups attest that medical prices are disconnected from average variable costs. Anderson et al. (2003) link the high HCE share in the United States to prices. However, their period of analysis is 1990–2000 in which the US health care expenditure gap remained flat. Hall and Jones (2007) propose a model in which the HCE share goes up with income and technology growth with the end result of increasing life expectancy. As is well known, there has been some leveling of income by European countries. Moreover, US life expectancy has always been slightly below the OECD average. Hence, life expectancy could be instrumental in explaining global increasing trends in HCE, but other factors appear more adequate to replicate the observed differences at the cross-country level over shorter time spans. Some researchers (cf., Anderson and Frogner 2008) argue that even though the United States presents the highest ratios of medical spending among all OECD countries, its residents are not granted the highest value per dollar spent in health care—suggesting a higher level of inefficiency in the United States. Our analysis focuses on medical spending growth rates—rather than levels—within a selected group of OECD countries. In lieu of efficiency, we shall be concerned with the large wedge between the contribution of value added and HCE.2 Our paper suggests that monopoly power by the private sector appears to increase medical care prices, but it is not clear that public intervention might limit public health care expenditure whereas reducing costs as a result of monopsony power since public policies could also lead to over consumption by some parts of the population because of expansionary health care insurance programs. Indeed, the spread of health care insurance has been claimed as an explanatory variable for the growth in US medical expenditures. Finkelstein (2007) estimated that between 1965 and 1970 the introduction of Medicare produced an increase in hospital spending six times larger than a private insurance program would have produced. At roughly the same time, however, many OECD countries did undergo notable expansions of their universal health care systems (e.g., Japan in 1961, Denmark in 1973, and Spain in 1986). Thus, we are led to conclude that expansionary trends in insurance markets cannot account for observed differences in HCE between the United States and the rest of the OECD countries. Some international medical reforms will be echoed in our quantitative studies. Ample literature has linked the rapid increase in HCE in the United States to technological change.3 For example, Di Matteo (2005) finds that technological change accounts for two thirds of health care spending growth over the 1975–2000 period. An excellent review of this literature is presented in Chernew and Newhouse (2012). Again, in our study, we control for the impact of technological change by conditioning on a sample of developed countries with comparable technologies. Chandra and Skinner (2012) claim that simply attaching the label of “technological growth” as the major cause of increases in health care expenditure is not a meaningful way to address the problem. Our macro-level difference-in-difference approach should be of interest for the health micro literature that has tried to answer the same question of excessive HCE in the United States but typically uses more fragmented evidence and does not focus on the 1978–1990 transition period. Chandra and Skinner (2012) follow Hall and Jones (2007) and propose a model with three types of health care goods that differ in cost-efficiency. They suggest that growth in US HCE is mostly driven by the widespread use of the least cost-efficient medical treatments embedded in new technologies. In their view, the adoption of these new technologies comes from the financing of health care costs through the insurance industry. Other health care systems with central budgeting or quantity constraints are not likely to experience rapid growth in these technologies. Garber and Skinner (2008) review some micro studies and claim that both cost and allocation efficiency between health care consumption and other consumption is difficult to answer. They also claim that other countries have shared the enormously valuable improvement in health—enjoyed by Americans in recent decades—at a much lower cost. Our study shows that an increasing US medical care expenditure gap does not stem from an extended use of capital and labor in the United States. Thus, it appears that new technologies (more prevalent in the United States) are typically associated with higher price markups. Hence, large product markups support the idea that market power—rather than the extra cost to generate higher product quality—might be an important determinant of US health care prices. Hospital and Related Services appears to be a driver of inflation in the US health care sector over the 1978–1990 transition period. Health insurance premiums underwent steep increases—considerably higher than the consumer price index (CPI) and the CPI-Medical Care (CPIMC) index. The paper is organized as follows. In Section 2 we highlight some stylized facts about the evolution of labor and capital in the health care sector, and propose a basic growth accounting exercise based on value added. The implied US medical care expenditure gap remains fairly constant over the various time periods. In Section 3 we support this growth accounting exercise by looking at various indicators of health care quality. After filtering out price effects we confirm the stability of US real medical consumption over total consumption. In fact, adjusting for common quality trends and medical care inflation does not change our basic empirical findings about the implied evolution of the US medical care expenditure gap. In Section 4, we reexamine various independent measures of prices and sources of US health care inflation. We conclude in Section 5. 2. A Basic Growth Accounting Exercise for the US Health Care Sector In our first pass to the data, we present a growth accounting exercise using the evolution of labor and capital for a restricted sample of countries. Moreover, for comparison purposes and to circumvent certain issues with some missing data, we also conduct a parallel growth accounting exercise for the US economy only. We find that labor and capital accumulation have no impact on the differential increase of the US medical care expenditure gap. Actually, the predicted gap from factor accumulation runs in favor of the other countries. Over the sample period 1970–2007, employment in the health care sector relative to total employment in the economy increased by 45% in the United States and by 88% in the OECD, whereas average investment in the health care sector relative to total investment in the economy is about 3.7% in the United States and 2.1% in the OECD. Labor income in the health care sector for the United States is about 9% of total labor income, and labor income in the health care sector for the OECD countries is about 8% of total labor income. Therefore, the contribution of value added from factor accumulation to HCE in the United States is relatively small since the share of HCE in GDP is about 18%. 2.1. Evolution of Labor and Capital in The Health Care Sector Table 1 reports various dimensions of the evolution of labor and capital in the health care sector for some selected dates. In row (i) we observe that total physicians’ compensation over HCE has decreased from almost 15% in 1982 to 13.1% in 2000. Physicians’ compensation over the average worker compensation in the economy increased between 1980 and 1990 but ended at a similar ratio in 2000 [row (ii)].4 Row (iii) shows the evolution of the average compensation of a non-physician health care worker with respect to a physician. This ratio remains relatively flat over the sample period. Table 1. Labor and capital statistics. 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% Notes: a. Data from 1982. b. Average over the next 10 years. Sources: AMA, AMGA, Bureau of Labor Statistics, US Census Bureau and OECD Health Data. View Large Table 1. Labor and capital statistics. 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% Notes: a. Data from 1982. b. Average over the next 10 years. Sources: AMA, AMGA, Bureau of Labor Statistics, US Census Bureau and OECD Health Data. View Large Of course, HCE growth has expanded the labor force in the sector. In the 1970–2000 period, the number of active US physicians increased from 3.46 per 1,000 workers to 4.80 per 1,000 workers [row (iv)], whereas the number of active OECD physicians increased from 3.05 per 1,000 workers in 1970 to 5.36 per 1,000 workers in 2000 [row (v)]. The ratio of physician to non-physician health care workers and their income shares have been quite stable in our sample of countries for the periods in which data is available. With the introduction of Medicare and the subsequent expansion of the health care sector there was a considerable increase in the number of non-physician workers in the US health care industry in the late 1960s. For the other OECD countries, we have data for non-physician health care workers only after 1985. OECD and US data show that the ratio of physician to non-physician health care workers remained almost constant for the period 1985–2007. From our own computations from other US data sources, between 1970 and 1980 this ratio decreased from 13% to 8.5%, and between 1980 and 2007 it slowly decreased from 8.5% to 7.3%. In our growth accounting exercise we assume that the ratio of physician health care workers to non-physician health care workers remains flat in both the United States and the OECD countries. When it comes to physical capital in the health care sector,5 growth runs in favor of the United States. Rows (vi) and (vii) report the average investment share in the health care sector in each decade between 1970 and 2000, whereas rows (viii) and (ix) report the average share of aggregate investment. Both the United States and OECD show a decreasing investment pattern in the health care sector, but the US share has always been above—albeit the difference has been diminishing over time. Therefore, labor in the health care sector has grown faster in the OECD whereas the corresponding physical capital investment share is greater in the United States. We now lay down a growth accounting exercise based on factor accumulation to assess the observed growth patterns in the US medical care expenditure gap. 2.2. Growth Accounting Based on Value Added Growth accounting based on value added circumvents the use of aggregate prices that may be picking up quality effects for the provision of goods and services. Because of data availability, in this section we only use seven countries in our sample: Australia, Canada, Germany, Japan, Spain, the United Kingdom, and the United States. 2.2.1. Growth Accounting Using Capital and Labor Data for the United States and the OECD Let us assume that production in the overall economy and in the health care sector evolve according to the following Cobb–Douglas production functions: \begin{eqnarray*} Y&=&AK^{\alpha }L^{1-\alpha },\\ Y_{H}&=&A_{H}K_{H}^{\alpha _{H}}L_{H}^{1-\alpha _{H}}, \end{eqnarray*} where A is the technology residual or total factor productivity (TFP) in the economy, K is the aggregate capital stock, L is total employment, and 0 < α < 1 is the capital income share. Similarly, AH is TFP in the health care sector, KH and LH are capital and labor devoted to health care goods, and 0 < αH < 1 is the capital income share in the health care sector. Then, the evolution of the US medical care expenditure gap depicted in Figure 1 is given by the ratio $$\frac{\left(\frac{p_{H}^{ {\mathit US}}\times Y_{H}^{ {\mathit US}}}{p^{ {\mathit US}}\times Y^{ {\mathit US}}}\right)}{\left(\frac{p_{H}^{ {\mathit OECD}}\times Y_{H}^{ {\mathit OECD}}}{p^{ {\mathit OECD}}\times Y^{ {\mathit OECD}}}\right)},$$ (1) where |$p_{H}^{ {\mathit US}}$| and |$p_{H}^{{\mathit OECD}}$| are the prices in the health care sector and |$p^{ {\mathit US}}$| and |$p^{ {\mathit OECD}}$| are the prices in the rest of the economy in the United States and OECD, respectively. Growth accounting focuses on the contribution of observable production factors. Here, we want to assess the contribution of capital and labor to the US medical care expenditure gap as well as the size of the residual over our sample of countries. We start this exercise with the following identifying assumption over the growth rates of the unobservable factors: $$\left(g_{A,H}^{ {\mathit US}}-g_{A}^{ {\mathit US}}\right)=\left(g_{A,H}^{ {\mathit OECD}}-g_{A}^{ {\mathit OECD}}\right).$$ (2) Additionally, when we analyze the US medical care expenditure gap in this section we will be assuming that $$\left(g_{p,H}^{ {\mathit US}}-g_{p}^{ {\mathit US}}\right)=\left(g_{p,H}^{ {\mathit OECD}}-g_{p}^{ {\mathit OECD}}\right).$$ (3) If both assumptions hold, differential factor accumulation should suffice to explain the US medical care expenditure gap. The rest of the paper provides direct and indirect evidence that the main driver of the observed US medical care expenditure gap is a violation of the equal growth rates for relative prices in equation (3). More precisely, our analysis implies that |$\big(g_{p,H}^{ {\mathit US}}-g_{p}^{ {\mathit US}}\big)>\big(g_{p,H}^{ {\mathit OECD}}-g_{p}^{ {\mathit OECD}}\big)$|⁠. Under equation (3), the US medical care expenditure gap is simply defined as |$\big(Y_{H}^{ {\mathit US}}/Y^{ {\mathit US}}\big)/\big(Y_{H}^{ {\mathit OECD}}/Y^{ {\mathit OECD}}\big)$|⁠, which depends only on the accumulation of inputs in each economy and the technology residuals. The equal growth rate in the residuals [equation (2)] states that the differential increase in technological change in the health care sector between the United States and the OECD is the same as the differential increase in technological change for the aggregate economy between the United States and the OECD. We will later check if this is a good approximation for US data. Now, taking logarithms, differentiating with respect to time, and assuming equal shares in the United States and the OECD, we obtain the following expression for growth in the US medical care expenditure gap from the observable factors: \begin{eqnarray} &&\!\!\!\!{\alpha _{H} \left(g_{K,H}^{ {\mathit US}}-g_{K,H}^{ {\mathit OECD}}\right)+(1-\alpha _{H})\left(g_{L,H}^{ {\mathit US}}-g_{L,H}^{ {\mathit OECD}}\right)}\nonumber \\ &&\quad -\alpha \left(g_{K}^{ {\mathit US}}-g_{K}^{ {\mathit OECD}}\right)-(1-\alpha )\left(g_{L}^{ {\mathit US}}-g_{L}^{ {\mathit OECD}}\right). \end{eqnarray} (4) Calibration. Using OECD data we get a capital to GDP ratio K/Y between 2.0 and 2.5 for a subset of countries with available data. The observed investment to GDP ratio is around 0.20, which together with K/Y = 2.5 yields an annual depreciation rate (δ) of 0.08 in the steady state. After letting K/Y = 2.5 in 1970, we calculate the evolution of K for each country according to the standard law of motion Kt + 1 = It + (1 − δ)Kt. The initial KH is calculated differently since the share of investment in the health care sector is lower than the observed average share in the economy. We first estimate the average investment in health care over HCE (IH/YH) and call this variable φ. Then, we calculate the initial capital in the health care sector as KH, 1970 = (φ/δ)YH, 1970. Once this is estimated, we again apply the standard equation for the capital's law of motion to calculate KH, t for all dates t. Finally, the last parameter we need to calibrate is the capital share α. The real return to capital in the United States has been around 8%.6 Letting K/Y = 2.5 and a net capital return of 8% yields α = 0.4. In our sensitivity analysis we let α range between 0.25 and 0.40. In our benchmark calculation we set α = αH = 0.25. Arguably, the health care sector has a lower K/Y ratio that would lead to a lower capital share than the rest of the economy. Some authors, however, have stressed the externality effects of equipment investment (e.g., De Long and Summers 1992). Hence, the social rate of return may be higher than the private rate of return. Therefore, a higher value for αH could be justified by invoking some plausible externality effects and intangible investments. Results. We now simulate equation (4) under our benchmark calibration for the estimated capital stocks and OECD data on health care employment as a proxy for LH and total employment as a proxy for L. Figure 3 shows the US medical care expenditure gap implied by the model together with its empirical counterpart. We also report the pace of the US medical care capital gap \begin{equation*} \left[\left(K_{H}^{ {\mathit US}}/K^{ {\mathit US}}\right)/\left(K_{H}^{ {\mathit OECD}}/K^{ {\mathit OECD}}\right)\right] \end{equation*} and the US medical care labor gap \begin{equation*} \left[\left(L_{H}^{ {\mathit US}}/L^{ {\mathit US}}\right)/\left(L_{H}^{ {\mathit OECD}}/L^{ {\mathit OECD}}\right)\right]. \end{equation*} The figure shows a relative increase in the stock of capital in the United States and a relative increase in employment in the OECD. Figure 3. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States and OECD. Source: OECD Health Data (June 2014) and authors’ computations. Figure 3. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States and OECD. Source: OECD Health Data (June 2014) and authors’ computations. Observe that the simulated US medical care expenditure gap remains quite flat during the entire period with a slightly declining trend. The gap goes down by 15% for our simulation of equation (4) whereas it goes up by 30% in the data. Further, during the 1978–1990 transition period the gap goes down by 5% for our simulation whereas it goes up by 32% in the data. Sensitivity Analysis. We also consider changes in these baseline values. We let the capital share α range between 0.4 and 0.25, and we let αH differ from α. We let annual depreciation δ range between 0.06 and 0.08. Table 2 shows the percentage change in the simulated US medical care gap for the entire period in panel (a) and for the 1978–1990 transition period in panel (b). For the 1970–2007 period, under all pairs (δ, α), we observe a decrease in the simulated US medical care expenditure gap between 10% and 15%, as opposed to an increase of 30% in the data. Similar results are observed during the 1978–1990 transition period in panel (b). Therefore, the present growth accounting exercise based on value added determines that the US medical care expenditure gap should have decreased during the period 1970–2007 as well as during the 1978–1990 transition period.7 Table 2. Sensitivity analysis. (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 Source: OECD Health Data (June 2014) and authors’ computations. View Large Table 2. Sensitivity analysis. (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 Source: OECD Health Data (June 2014) and authors’ computations. View Large Implied Medical Expenditure Gaps by Country. To get a better understanding of the underlying forces driving the US medical expenditure gap over our sample of countries, we also report the relative contribution of factor accumulation to the medical sector for each country i as compared to the United States. More formally, Figure 4 plots the evolution of the medical expenditure gap as defined by the following equation: \begin{eqnarray} &&\!\!\!\!{\alpha _{H}\left(g_{K,H}^{i}-g_{K,H}^{ {\mathit US}}\right)+(1-\alpha _{H})\left(g_{L,H}^{i}-g_{L,H}^{ {\mathit US}}\right)}\nonumber\\ &&\quad -\alpha \left(g_{K}^{i}-g_{K}^{ {\mathit US}}\right) - (1-\alpha )\left(g_{L}^{i}-g_{L}^{ {\mathit US}}\right), \end{eqnarray} (5) where i stands for every other country in our restricted sample. Figure 4. View largeDownload slide A basic growth accounting exercise—implied medical expenditure gaps by country. Source: OECD Health Data (June 2014) and authors’ computations. Figure 4. View largeDownload slide A basic growth accounting exercise—implied medical expenditure gaps by country. Source: OECD Health Data (June 2014) and authors’ computations. First, the figure shows that most individual medical care expenditure gaps have remained relatively flat as we condition by US value added in the health care sector. The sharp decrease observed for Germany corresponds to the unification of the country after the fall of the Berlin Wall because of a break in the data. Note that this decrease has been tapering off over time. Second, there are two outliers: Canada and Spain. The share of HCE in Canada has been slightly above the OECD average but the contribution of value added appears remarkably low. The Canadian medical system has been widely criticized for being comparatively expensive, having long waiting periods for services, and having limited access to physicians and medical technology, which may justify the low contribution of health care value added in our exercise. At the other extreme, a high contribution of value added in the Spanish medical sector is observed at the beginning of our sample period. Spain presents the highest percentage increase in the share of HCE moving from 3.5% of GDP in 1970 to 8.4% in 2007. Spain underwent various reforms toward higher health care quality and universal insurance for all citizens and residents, as well as an impressive GDP growth at these early dates. 2.2.2. Growth Accounting Using Capital and Labor Data for the United States only As labor and capital data could be of lesser quality in our sample of OECD countries, we can simulate a variant of equation (4) for US data only. That is, we analyze the relative contribution of value added in the US health care sector over overall value added, and we condition for the growth of the share of HCE in the OECD countries. Formally, we consider the evolution of the US medical care expenditure gap (United States only) under the following growth components: $$\alpha _{H}g_{K,H}^{ {\mathit US}}+(1-\alpha _{H})g_{L,H}^{ {\mathit US}}-\alpha g_{K}^{ {\mathit US}}-(1-\alpha )g_{L}^{ {\mathit US}}-g_{Y,H}^{ {\mathit OECD}}+g_{Y}^{ {\mathit OECD}}.$$ (6) The US values for K, KH, α, αH, and δ come from the preceding exercise. Also, L and LH are taken from employment data. For the other OECD countries, we consider the growth in the overall share of HCE. Medical prices face particularly difficult measurement problems, and according to the Boskin Commission Report (1996) the US consumer price index does not take into account a 3% yearly increase in the quality of health care goods, whereas this quality bias was less than 1% for the general index. On these grounds, we may consider that technological progress may have been more pronounced in the medical sector. Hence, our implicit identifying assumption |$g_{A,H}^{ {\mathit US}}=g_{A}^{ {\mathit US}}$| underlying this exercise may not be supported by the data. Figure 5 displays the newly simulated US medical care expenditure gap under equation (6) together with the previously simulated gap under equation (4). The figure illustrates that for the benchmark calibration, there is almost no difference between the present and previous growth accounting exercises. This confirms that the US medical expenditure gap also declines when the contribution of value added is restricted to the US economy. We should note, however, that for some other combination of parameter values used in the sensitivity analysis of Table 2, the simulated US medical care expenditure gap (United States only) is significantly below that obtained from equation (4). Hence, some of these computations suggest that TFP in the US health care sector may have grown faster than in the aggregate economy. Accordingly, the underlying identifying assumption |$g_{A,H}^{ {\mathit US}}=g_{A}^{ {\mathit US}}$| may not be a good approximation of the evolution of TFP in the health care sector. Our general formulation in (4) appears to be more adequate as it conditions for both general TFP trends in the aggregate economy and the health care sector. Figure 5. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States only. Source: OECD Health Data (June 2014) and authors’ computations. Figure 5. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States only. Source: OECD Health Data (June 2014) and authors’ computations. Finally, using the previous values for the growth of labor and capital we compute the growth rate of value added in the health care sector over value added in the economy. Formally, for every date we compute \begin{eqnarray*} \alpha _{\mathit {H}}g_{\mathit{K,H}}^{\mathit {US}}+(1-\alpha _{\mathit {H}})g_{\mathit {L,H}}^{\mathit {US}}-\alpha g_{\mathit {K}}^{\mathit {US}}-(1-\alpha )g_{\mathit {L}}^{\mathit {US}}. \end{eqnarray*} Under our baseline parameters, for the period 1970–2007 the share of HCE grows by 45% in our simulation exercise, whereas it grows by 130% in the data. For the 1978–1990 transition period the simulated share of HCE grows by 17%, whereas it grows by 47% in the data. Therefore, it appears that value added cannot account for the patterns of growth in the share of HCE in the United States, and would require implausible growth in TFP in the health care sector. 2.3. Returns, Yields, and Markups of Publicly Traded Companies in The Health Care Sector The previous growth accounting exercise attests that factor accumulation cannot explain the increase in the US medical care expenditure gap. It follows that our identifying assumptions [equations (2) and (3)] may fail. That is, the pace of technological change and incremental inflation rates in the medical sector may be greater in the United States than in the average OECD country. Comin and Hobijn (2004 and 2010) document that technology spreads relatively quickly across countries with similar levels of development. They argue that the diffusion of new technology became increasingly fast after World War II. After considering financial data we conclude that increasing prices in the US health care sector cannot be associated with additional expenditures in resources. Moreover, Section 3 builds on a vast literature claiming that higher medical prices are not associated with better health outcomes either. Therefore, the sharp break in the US medical care expenditure gap observed during the 1978–1990 period appears to be difficult to justify on the grounds of slow diffusion of technology and increasing costs originating from better product quality in the United States. Several micro studies analyze the degree of competition and efficiency in the US medical care sector. For instance, Dunn and Shapiro (2011) claim that physicians’ market power may bias medical care prices and the quantity of health care services provided. Skinner et al. (2005) argue that Medicare spending appears to be highly inefficient: About 20% of Medicare expenditures do not provide any increase in survival rates or quality of life for the elderly population. Koijen, Philipson, and Uhlig (2016) call attention to a “medical innovation premium” of 4%–6% annually for equity returns of firms in the health care sector, and interpret this premium as compensating investors for government-induced profit risk. As discussed in the introductory section, we do not primarily address these efficiency issues from a micro perspective. Our main goal here is to focus on the large wedge between value added and HCE documented in our previous growth accounting exercise. As already pointed out, physicians did not particularly benefit from the observed HCE growth. But private companies may have been able to take advantage of monopoly power and existing frictions and regulations in the health care market. We consider stock returns, yields, and markups generated by publicly traded companies in the health care sector as compared to those in the aggregate economy. We find that the health care sector is associated with abnormal profits and markups. These financial measures of profitability grow much faster over the 1978–1990 transition period. We retrieve company level data on Total Revenues, and Total Variable Costs from COMPUSTAT, that is, entries REVT and COGS in COMPUSTAT, for the period between January 1970 and December 2007. We retrieve data on returns, prices, and shares outstanding on publicly traded firms in the United States from the Center for Research in Security Prices (CRSP), that is, entries RET, PRC and SHROUT in CRSP. We select our health care companies from the Standard Industrial Classification (SIC) codes as in Kenneth French's webpage within the set of industrial portfolios.8 This is arguably the most used portfolio database in the finance literature. To have an idea of the relevance of our sample, we calculate the ratio of total yearly sales for the health care companies in our portfolio over HCE. This ratio increases from 20% in 1970 to 30% in 2007; hence, our sample appears to be fairly representative. Stock Market Return. At every date, the health care portfolio return is calculated as an average of each company's return (including dividends) weighted by its relative market capitalization in the health care portfolio.9 We also use a market weighted portfolio that includes almost all publicly available companies in the CRSP database.10 Average returns in the health care sector portfolio have soared since the late 1970s. The weighted average yearly return for the 1970–2007 period for this portfolio is about 20.17% whereas that of the market portfolio is about 12.34%. In the finance literature it is customary to consider the metric excess return adjusted by its standard deviation. This metric is known as the Sharpe Ratio and provides a measure of the reward-to-risk. The risk-free rate (the one-month T-bill in our case) is subtracted from the portfolio return to compute the excess return. Under certain regularity conditions, all well diversified portfolios should have the same Sharpe Ratio. For our health portfolio, the Sharpe Ratio is 0.65 for the period 1970–2007 whereas it is 0.43 for the market portfolio. Interestingly, the Sharpe Ratio of the health care portfolio is 1.05 for the 1978–1990 period, whereas it is 0.04 and 0.72 for the 1970–1977 period and the 1991–2007 period, respectively. Yield. We define the yield of a company as the company's gross profit divided by its market capitalization value at a certain date. Gross profit is the difference between total revenue and total variable cost. We focus on gross profit because this is usually considered a clean accounting measure of profitability (cf., Novy-Marx 2013). The aggregate yield is then obtained as a weighted average of individual yields, where the weights are company total revenues over the aggregate sum of total revenues. For our sample period, we get an average aggregate yield of about 36% for the companies in the health care sector, and 70% for all other companies. A lower average yield may mean that companies in the health sector face lower risk and higher expectations of profit growth. In some research in progress, we have documented that lower yields are supported by sizable and stable product markups. Price Markup. We define the price markup of a company as the ratio between total revenue and total variable cost. The aggregate price markup is then obtained as a weighted average of company markups where the weights are company total revenues over aggregate total revenues. For our sample period 1970–2007, we get an average aggregate price markup in the health care sector around 193%, whereas we get an aggregate price markup for all the other companies around 55%. The average yearly growth rate of HCE over nominal GDP was 3.2% during the 1978–1990 period, whereas it was 2.4% in the 1970–1977 period, and 1.6% in the 1991–2007 period. Table 3 compares this incremental growth rate against the following aggregate measures: (i) Annual incremental return of the health care sector portfolio as compared to the remaining market portfolio, (ii) Annual incremental growth rate of the yield of the health care sector portfolio as compared to the remaining market portfolio, (iii) Annual incremental growth rate of the aggregate markup of the health care sector portfolio as compared to the remaining market portfolio, and (iv) Annual incremental growth rate of value added in the health care sector as compared to the aggregate economy. Table 3. Returns, yields and markups of publicly traded companies in the health care sector relative to the rest of the economy. Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Source: CRSP, COMPUSTAT, Kenneth French website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/), and authors’ computations. View Large Table 3. Returns, yields and markups of publicly traded companies in the health care sector relative to the rest of the economy. Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Source: CRSP, COMPUSTAT, Kenneth French website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/), and authors’ computations. View Large Remarkably, for the 1978–1990 and 1991–2007 periods the incremental annual markup growth rate in the health care sector is roughly the same as the incremental annual growth in the share of HCE, whereas there is much less variability in incremental growth of value added. Along these same lines, the incremental yield growth in the health care sector gets much higher during the 1978–1990 transition period. Hence, episodes of high growth of HCE in the United States are associated with higher markups and higher growth rates of profitability. Excess returns also get much higher during the transition period, and stay relatively high afterward even though there is a substantial decrease in the excess growth rate of the yield. This slow adjustment of stock market values typically occurs when high profits are initially perceived as temporary or non-sustainable. To have a better sense for these trends, Figure 6 depicts the evolution of Health Care Sector Markup over Economy Markup, Health Care Sector Gross Profit over Economy Gross Profit, Health Care Sector Value Added over Economy Value Added, and HCE over GDP between January 1970 and December 2007. Panel (a) displays variables in levels, and panel (b) displays 5-year moving averages of corresponding growth rates. For convenience, all levels values are normalized to 100 at the start of the transition period in 1978, and the two vertical bars bracket the 1978–1990 transition period. The following facts are worth noticing from these figures. First, Health Care Sector Markup over Economy Markup mimics HCE over GDP whereas the incremental value added presents milder growth rates. Fluctuations in Health Care Sector Markup over Economy Markup slightly lead but are fairly contemporaneous with those of HCE over GDP. Increases in these two aggregate measures stimulate factor accumulation but this response is not very pronounced. Second, fluctuations in the growth rate of Health Care Sector Gross Profit over Economy Gross Profit move in tandem with that of Health Care Sector Markup over Economy Markup, but tend to have smaller values. In fact, after 1994, relative markups have an upward trend, whereas relative gross profits remain relatively flat. This suggests some kind of cost-containment as discussed in footnote 14. Overall, the present figure is consistent with the notion that growth in Health Care Sector Markup over Economy Markup is strongly linked to growth in HCE over GDP; markups seem a major driver of HCE. Finally, Figure 7 shows the evolution of the Herfindahl–Hirschman Index (HHI) that measures industry concentration; that is, a higher value of the index implies higher industry concentration for the publicly traded companies in the health care sector. Besides this index, panel (a) shows the number of publicly traded companies in the health care sector, whereas panel (b) shows Health Care Sector Markup over Economy Markup. From panel (a) we can observe that even though the number of publicly traded companies was steadily going up since the beginning of the 1978–1990 transition period, the HHI continues to increase until reaching its maximum value in the year 1989. (As discussed in Figure 6, increases in markups may stimulate further economic activity that in this case may translate into entry of firms in the medical sector.) Panel (b) shows that both the HHI and Health Care Sector Markup over Economy Markup went up during the transition period, but afterward these two measures diverge. Hence, industry concentration can only offer a partial account for the evolution of markups. Further research may be necessary (possibly at the micro level) to understand the nature of market frictions that determine the evolution of product markups. Usually, markups are tied down to changes in demand elasticities and in the regulatory environment. Figure 6. View largeDownload slide Health care sector markup over economy markup, health care sector gross profit over economy gross profit, health care sector value added over economy value added, and HCE over GDP. Source: CRSP, COMPUSTAT and authors’ computations. Figure 6. View largeDownload slide Health care sector markup over economy markup, health care sector gross profit over economy gross profit, health care sector value added over economy value added, and HCE over GDP. Source: CRSP, COMPUSTAT and authors’ computations. Figure 7. View largeDownload slide Herfindahl–Hirschman index for the health care sector. Source: CRSP, COMPUSTAT, and authors’ computations. Figure 7. View largeDownload slide Herfindahl–Hirschman index for the health care sector. Source: CRSP, COMPUSTAT, and authors’ computations. In summary, strong positive growth in the ratio of HCE over GDP can be fully backed up by corresponding growth in gross profits and markups. The connection between value added and financial variables appears to be much more tenuous, especially over the 1978–1990 transition period. For a realistic pace of technology diffusion, abnormal profits and markups help us understand that prices in the medical sector may not reflect the extra costs associated with high product quality. 3. Quality and Price Effects To evaluate the wedge between value added and HCE, we now take a different approach and adjust HCE by aggregate prices. Of course, prices may be picking up quality effects but our previous evidence on the evolution of markups and gross profits shows a price disconnect from average costs. We first document some indicators of US health care quality within the OECD countries. After filtering out price effects, we also show that US real health care consumption moves intandem with real GDP, as well as with real health care consumption in the other OECD countries. We then simulate a product quality residual over our sample of OECD countries. This approach does not change our basic findings in the previous section, but it is subject to new measurement issues as it is usually hard to disentangle price and quality effects. 3.1. Quality Our growth accounting exercises rest on a certain identifying assumption for non-observable TFP growth; see equation (2). This assumption appears to be adequate for the 1978–1990 transition period. As already discussed, it is unlikely that TFP changes can have such a sizable differential impact across economic sectors and countries over such a limited time span. We next provide some supporting evidence for our identifying assumption by studying the evolution of product quality at the aggregate level. Various international institutions such as the OECD and the World Health Organization periodically report cross-country rankings on various dimensions of health status, heath care activities, quality of care, and access to care. Generally, the United States lags behind many other OECD countries. In fact, when conditioning for income and health care expenditures the United States appears mostly as an outlier. Figure 8 shows the evolution of some of these quality proxies for health status, which will be expressed as gaps: the corresponding US value over the average value of our sample of OECD countries (excluding the United States), letting 100 be the initial value for our starting year, 1970. We consider the US gap for infant mortality, neonatal mortality, perinatal mortality, and life expectancy at birth.11 We can observe that although the life expectancy gap has remained quite stable, the gap of every other variable has increased—meaning that in fact the quality gap has evolved against the United States. These outcomes may partially be explained by other measures of health care activities and access to care such as the relatively large size of the uninsured population in the United States: Around 12% of the population has permanently been uninsured over the last three decades. Figure 8. View largeDownload slide Quality indicators over time. Source: OECD health data (June 2014). Figure 8. View largeDownload slide Quality indicators over time. Source: OECD health data (June 2014). For narrower quality indicators we have not been able to gather data over the whole sample period. Table 4 reports on the following quality measures for the year 2003: Life expectancy, breast cancer five-year observed and relative survival rates for females, cervical cancer five-year observed and relative survival rates for females, and colorectal cancer five-year observed survival rates for males and females.12 For each of these indicators the coefficient of variation (i.e., the standard deviation over the mean) is usually very small—attesting that there is very little variability in all the quality proxies across the OECD countries.13 The United States has the lowest life expectancy, the maximum survival rates for breast cancer, and is below the maximum for cervical and colorectal cancers. Overall, the assumption of similar health care quality across OECD countries seems to be supported by the data. Table 4. Quality measures in 2003. 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) Note: If there is no data for a country for a specific year but there is for an immediate previous or following year we use that number. View Large Table 4. Quality measures in 2003. 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) Note: If there is no data for a country for a specific year but there is for an immediate previous or following year we use that number. View Large Certainly, all the previous evidence should be merely considered as informative of the health care quality level across countries. Besides, life expectancy can be affected by genetic differences and dietary habits. Similarly, survival rates are not necessarily a good proxy for the quality of life after a surgical procedure. Some indicators may proxy longevity rather than quality. Drösler, Romano, and Wei (2009) report findings on various new quality indicators regarding patient safety across OECD countries. Again, the United States always ranks about average in every category. 3.2. The US Medical Care Price Gap and the Stability of Real HCE over Total Consumption in the United States Besides product quality, health care prices could be driving the US medical expenditure gap. A measure of the consumer price index—medical care (CPIMC) is available for all countries in our sample. These prices along with their data sources are formally explained in the Appendix. Measurement issues concerning the CPIMC will be deferred to the next section. We define the US medical care price gap as the US ratio CPIMC/CPI over the average ratio CPIMC/CPI of the other OECD countries in our sample. Figure 9 plots the US medical care expenditure gap against the price gap; both ratios are normalized to 1 in 1977. Observe that the US medical care price gap mimics quite well the US medical care expenditure gap for the 1978–1990 period in which both increased by about 35%. Then, there is a mild disconnect: The relative price of medical care appears to increase faster than HCE. More precisely, between 1993 and 2007 the US medical care price gap goes from 1.35 to 1.50, whereas the US medical care expenditure gap appears quite flat.14 Indeed, it follows from these definitions that if the US medical expenditure gap grows less than the price gap, then the ratio of real HCE over real GDP will go down in the United States as compared with the other OECD countries. As a matter of fact, the 1990s witnessed a shift to managed care leading to a relative decline of hospital care [Figure 2(b)], and to a decrease of indemnity health insurance plans in favor of HMOs, PPOs, and POSs. Cutler, McClellan, and Newhouse (2000) suggest that for certain treatments and procedures managed care organizations like HMOs may have lowered costs by about 30%–40% as a result of price declines of medical services and treatment intensities. Figure 9. View largeDownload slide The US medical care expenditure gap versus the price gap 1970–2007. Source: Health Care Expenditures are from the OECD Health Data (June 2014). Health Care Price data are taken from each country's official statistics and authors’ computations. Figure 9. View largeDownload slide The US medical care expenditure gap versus the price gap 1970–2007. Source: Health Care Expenditures are from the OECD Health Data (June 2014). Health Care Price data are taken from each country's official statistics and authors’ computations. To go further into the evolution of real consumption in health care in the United States, in Figure 10 we adjust HCE under various price measures. For the 1970–2007 sample period, HCE per capita at constant consumption prices in the United States has increased by 300%. For the same time period, HCE per capita at constant medical care prices (real HCE) in the United States has increased less than 100%. Moreover, there is no significant growth when real HCE per capita is adjusted by real total consumption (TC), but we observe some slight growth in the share of value added. Hence, Figure 10 reports: (i) HCE per capita at constant TC prices; (ii) Real HCE per capita: HCE per capita at constant medical care prices; (iii) Real HCE over real TC; and (iv) US value added in the health care sector over total value added in the economy. Figure 10. View largeDownload slide The evolution of US health care expenditures 1970–2007. Source: US Department of Health and Human Services—centers for Medicare and Medicaid services and the Bureau of Labor Statistics. Figure 10. View largeDownload slide The evolution of US health care expenditures 1970–2007. Source: US Department of Health and Human Services—centers for Medicare and Medicaid services and the Bureau of Labor Statistics. 3.3. A Cross-Country Analysis of The US Health Care Expenditure Gap as Implied by Aggregate Prices We now present a more systematic analysis of real health care consumption across countries. In a separate appendix we detail a simple model for health care demand embedding several sources of technological progress and health care quality. The main purpose of our exercise is to filter out price effects and common trends in technology and product quality under the simplifying assumption of a constant elasticity of substitution between the health care good and the remaining aggregate good in the economy. More precisely, let us assume that qh is health care expenditures (HCE), c is total consumption (TC), q is the relative price of medical care, and |$\widehat {A}$| is a residual term picking up product quality. Then, let $$\frac{\partial {\rm ln}(h_{t}/c_{t})}{\partial {\rm ln}(q_{t})}=\frac{1}{\rho -1},$$ (7) where −∞ < ρ < 1. After appending the residual term, we get: |$\text{ln}(q_{t}h_{t}/c_{t})={\rm ln}(\widehat {A}_{t})+(\rho /(\rho -1))\text{ln}(q_{t})$|⁠. This is therefore our basic equation to track the evolution of the US medical expenditure gap: $$q_{t}h_{t}/c_{t}=\widehat {A}_{t}q_{t}^{\frac{\rho }{\rho -1}}.$$ (8) In our data analysis, qh/c corresponds to the share of health care expenditures in total consumption for the United States and OECD countries, whereas q picks the evolution of the ratio of the medical care price index CPIMC over the CPI for each country. Under the identifying assumption of equal growth rates for the residuals, $$g_{\widehat{A}}^{ {\mathit US}}=g_{\widehat{A}}^{ {\mathit OECD}},$$ (9) we have the following expression for growth in the US medical care expenditure gap based on incremental inflation in the medical sector: $$\left(\frac{\rho }{\rho -1}\right)\left(g_{q}^{ {\mathit US}}-g_{q}^{ {\mathit OECD}}\right).$$ (10) Ringel et al. (2000) report estimates for the price-elasticity of demand for health care consumption of around −0.17. Since the health care good is a relatively small fraction of total consumption, we then get that ρ is approximately equal to −5. Studies consistently find inelastic demands when the share paid by the consumers is relatively low. These point estimates for the elasticity are around our calibrated value (cf., op. cit.). It is worth noticing that fractional payment policies for medical expenses (e.g., some types of coinsurance and out-of-pocket expenditures) will not affect the elasticity of demand. For comparison purposes, we also report computations for ρ = −3 and for ρ = −1, which imply elasticities of demand for health care consumption of around −0.25 and −0.50, respectively. Figure 11 reproduces these computations for the US medical care expenditure gap as implied by (10) as well as the empirical counterpart or actual gap. Observe that for ρ = −5 the simulated US medical care expenditure gap would account for about 107% of the actual gap. That is, over the sample period the implied US health care expenditure gap goes from 1.14 to 1.64 whereas in the data it goes from 1.14 to 1.53. For ρ = −3 the simulated residual would account for about 103% of the actual one, and for ρ = −1 the simulated residual would account for about 92% of the actual one. Because of the evolution of the medical expenditure gap in the last part of the sample, the breaking point is ρ = −2, where the simulated residual growth between 1970 and 2007 is almost the same as the actual one. This is consistent with a price elasticity of demand approximately equal to −0.33. For more inelastic demands, the observable growth in relative prices will overestimate the US medical expenditure gap. Figure 11. View largeDownload slide The US health care expenditure gap as implied by the relative price. Source: Health care expenditures are from the OECD health data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. Figure 11. View largeDownload slide The US health care expenditure gap as implied by the relative price. Source: Health care expenditures are from the OECD health data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. 3.4. Medical Expenditure Gaps by Country To have a better sense of these computations, we now replicate the US medical care expenditure gap for every single country i using ρ = −2. In fact, to adjust directly for actual growth in each individual gap, we consider the evolution of the following growth components for each country i: $$\left(\frac{\rho }{\rho -1}\right) \left(g_{q}^{ {\mathit US}}-g_{q}^{i}\right)-g_{HC}^{ {\mathit US}}+g_{HC}^{i},$$ (11) where |$g_{HC}^{i}$| is the is the growth rate of HCE over Total Consumption for country i. We can interpret expression (11) as the actual evolution of the ratios between the residuals |$\widehat {A}{}_{t}^{i}$| and |$\widehat {A}{}_{t}^{ {\mathit US}}$|⁠. Figure 12 shows the evolution of these ratios for the residuals for each country i in the sample. Except for France and Spain, all these ratios for the individual residuals are fairly stable. Hence, the United States displays a residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| quite in line with the corresponding average of the other OECD countries. Roughly speaking, there are not noticeable differences in the US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| between the 1978 and 1990 transition period and the rest of the sample. For reasons already discussed, the US residual declines after 1992, which suggests either a relative loss in US real health care consumption or an upward bias in the US price index. Figure 12. View largeDownload slide Evolution of the residuals by country as compared to the United States. Source: Health care expenditures are from the OECD Health Data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. Figure 12. View largeDownload slide Evolution of the residuals by country as compared to the United States. Source: Health care expenditures are from the OECD Health Data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. The French National Health Care System initial program was created in 1928 but was not comprehensive (Rodwin and Sandier 1993). France expanded its public health insurance programs at various stages, and it became universal for all its citizens and residents in 2000 (Rodwin 2003). Spain shows sharp trend breaks in 1986 and 2003 corresponding to two major medical reforms. The General Health Law of 1986 recognized the right to health care services for all citizens and foreign residents in Spain, and the Law of Cohesion and Quality of 2003 modernized and broadened the scope of the previous law. In conclusion, three important facts emerge from this study: (i) The US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| behaves quite similarly throughout all three time subperiods considered in our sample. This seems to be in accord with the aforementioned stability of real health care expenditures over real GDP observed in the US economy and the growth patterns of labor and capital in the health care sector. (ii) The pace of the US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| does not differ much from the average pace of the OECD residual |$\widehat {A}{}_{t}^{ {\mathit OECD}}$|⁠. And (iii) sharp breaks in the residuals are usually associated with medical reforms and regulations. Therefore, this growth accounting study appears to indicate that unobservables affecting the residuals like technology and product quality change would not be a major driving force for the pronounced increase in the US medical care expenditures gap over the 1978–1990 transition period. 4. Discussion of US Medical Care Prices As shown in Figure 9, aggregate prices seem to track well the evolution of the US medical expenditure gap. In this section, we discuss some of the limitations of aggregate health care prices. We also compare the evolution of our US health care price index with independent measures of cost and sources of inflation in the medical sector. As illustrated in Figure 13, health care prices in the United States are around 60%–80% higher than in other OECD countries. These differences seem quite commensurate with the size of the US medical care expenditure gap. Figure 13. View largeDownload slide International health care prices over US health care prices 2011. Source: International Federation of Health Plans. 2011 Comparative price report: medical and hospital fees by country. http://ifhp.com/documents/IFHP_PriceReport2010ComparativePriceReport29112010.pdf. Figure 13. View largeDownload slide International health care prices over US health care prices 2011. Source: International Federation of Health Plans. 2011 Comparative price report: medical and hospital fees by country. http://ifhp.com/documents/IFHP_PriceReport2010ComparativePriceReport29112010.pdf. The CPI-medical care (CPIMC) is published by the US Bureau of Labor Statistics (BLS) since 1935. The PCE-health care (PCEHC) is another major price index with data from 1929 and published since 1994 by the US Bureau of Economic Analysis (BEA). The evolution of these two measures is quite similar over the 1978–1990 transition period. As stated in Fixler and Jaditz (2002), there are three main differences between the BLS's CPIMC and the BEA's PCEHC: “First, the two indexes use different formulas. The CPI is a Laspeyres index, whereas the BEA product is a Fisher Ideal index. Second, the two indexes have different underlying concepts. The BLS product measures the prices paid by (urban) consumers, whereas the BEA product measures the prices of final consumption goods, wherever they are purchased. Finally, differences in how the detailed components are implemented lead to differences in how prices are measured and the weights attached to specific series.” After adjusting for aggregate inflation in the economy, Hospital and Related Services has experienced the highest overall growth rate (3.76% annually for the CPIMC over CPI since 1978). Prescription Drugs shows the lowest growth rate before the transition period starts, and it grows at 3% annually for the CPIMC over CPI after 1978. Professional Services appreciates at a slower pace (1.35% annually for the CPIMC over CPI) since 1978. After 1992, for all these categories the PCEHC grows at a slower pace than the CPIMC. In summary, the highest increases relative to average inflation are observed in Hospital and Related Services and within the 1978–1990 transition period. 4.1. Measurement Issues Inflation in the health care sector has been greater than average inflation in the economy in almost every year. As explained in the Boskin Commission Report (1996), the CPIMC and its various subcategories had an estimated annual upward bias of 3% between 1970 and 1995 when adjusting for quality. During the same time period, the overall CPI has an estimated annual upward bias of 0.6%, leading to an estimated yearly bias in the growth rate of the CPIMC over CPI of 2.4%. Technology may impact HCE in various ways. To control for the effects of technological progress our macro-level difference-in-difference approach conditions upon a group of OECD countries with similar health care quality and comparable methodologies in data collection. For many countries public provision of health care is the norm. Accordingly, medical care expenditures are estimated as in any other public good—including the cost of intermediate inputs and labor but not capital rents. Medical care prices in some countries do not depart much from the overall consumer price index. We should also point out that average inflation in the OECD was much greater than in the United States in the 1970s, but inflation has been much lower and more stable in the remaining part of the sample for all these countries. In fact, for both the United States and the OECD countries the CPIMC over the CPI did not grow at all in the 1970s. The US relative price of medical care started to grow in the 1978–1990 transition period. Therefore, the US medical care price gap is not influenced by unequal cross-country inflation trends; when inflation trends diverge between the United States and the OECD, there are no deviations in the relative price of medical care. A major component of the CPIMC is Hospital and Related Services. There is evidence that this component is upward biased because list prices are affected by discounts. Indeed, this bias has been the norm under the expansion of HMOs and PPOs since managed care organizations are able to bargain on hospital's list prices (Dranove et al. 2008). This possible upward bias may have been rather small during the 1978–1990 transition period since HMOs and PPOs just became important in the late 1980s. In 1997 the BLS started to implement corrective measures to the CPIMC-hospital and related services to remove this sampling error (Cardenas 1996). In a recent study, Koechlin, Lorenzoni, and Schreyer (2010) circumvent this upward bias by comparing what they call hospital quasi-prices over several OECD countries. These quasi-prices are defined as negotiated or administrative prices or tariffs on various hospital services items. Still, these authors find that US hospital services are 60% above the average cost of these other countries. Berndt et al. (2000) discuss another source of overstatement of the CPIMC-hospital and related services component: The switch from inpatient procedures to outpatient procedures—mainly because of the cost-containment efforts supported by managed care organizations. The switch from inpatient to outpatient treatments may have increased the average complexity and cost of medical procedures for both types. Indeed, very complex procedures were still left as inpatient care, and those shifted to outpatient care were more complex than the average procedure of this category. Since outpatient and inpatient procedures enter separately and with fixed weights in the index calculation, the possible cost reduction of the shift to outpatient care may have not been reflected in the CPIMC-hospital and related services. This upward bias may have been exacerbated after 1987 with the expansion of managed care organizations. The BLS did not take corrective measures until 1998 (Berndt et al. 2000). Again, this bias should be rather small for the 1978–1990 transition period. A pioneering work by Griliches and Cockburn (1994) found that US pharmaceutical price indexes were upward biased because of a shift from branded to generic drugs. Danzon and Chao (2000) found that the large differences between US prices and those of other countries reported in many studies might be smaller than previously thought. After 1995, the BLS implemented corrective measures for the construction of its pharmaceutical price indexes, including generic drugs (Berndt et al. 2000). It seems that neglecting generic drugs in the price index should have a very limited impact during the 1978–1990 transition period. The market for generic drugs expanded after the Waxman–Hatch Act of 1984 (Berndt and Aitken 2010). The generic drugs’ share of total dispensed prescription drugs in the US retail market represented just 18% in 1984. It increased to 36% by 1994, then 56% by 2004, and 74% by 2009. At the same time, the revenue share of generic drugs has been relatively small. For example, in 2004 generic drugs were 56% of dispensed drugs but its revenue share was only 17%. The revenue share of generic drugs rose to 24% by 2009 (Berndt and Aitken 2010). Hence, during the 1978–1990 transition period the upward bias would seem rather limited. Of course, generic drugs are also dispensed in every OECD country. Danzon and Furukawa (2011) found that generic drugs are more expensive in the United States than in most of the OECD countries right after the branded drug's patent expires (around 75% of the branded drug price). Three years after patent expiration, prices of generic drugs decline more sharply in the United States, and after eight years of patent expiration generic drugs are cheaper in the United States than in all other OECD countries except the United Kingdom. Again, these patterns may further reduce the possible upward CPIMC bias in the 1978–1990 transition period. 4.2. US Medical Prices, Markups, and Health Insurance Premiums We now study the evolution of independent price measures beyond the CPI and other sources of inflation supporting high prices in the US medical care sector. Price increases for medical products and services have come along with hikes in markups and health insurance premiums. Table 5 reports the evolution of several cost measures in the health care sector. Table 5. Independent price measures beyond the CPI and other sources of inflation in the US medical sector. 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 Source: US Bureau of Labor Statistics, US Bureau of Economic Analysis, US Department of Health and Human Services – Centers for Medicare and Medicaid Services, The 2003 GAO Report on Medical Malpractice Insurance, Goldman and McGlynn (2005), and Cohen et al. (2009). Physicians’ compensation data come from authors’ calculations using data from the American Medical Association and the Medical Group Management Association. View Large Table 5. Independent price measures beyond the CPI and other sources of inflation in the US medical sector. 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 Source: US Bureau of Labor Statistics, US Bureau of Economic Analysis, US Department of Health and Human Services – Centers for Medicare and Medicaid Services, The 2003 GAO Report on Medical Malpractice Insurance, Goldman and McGlynn (2005), and Cohen et al. (2009). Physicians’ compensation data come from authors’ calculations using data from the American Medical Association and the Medical Group Management Association. View Large 4.2.1. Aggregate Markup in the Health Care Sector over Aggregate Markup in the Economy A weighted average of Total Revenues over Total Variable Cost of publicly traded companies in the health care sector relative to the remaining publicly traded companies in the US stock market. (Construction of this variable is explained in Section 2.3; source: COMPUSTAT.) 4.2.2. Average Physician Compensation over CPI A measure of the evolution of the average physician's salary in the United States (sources: The American Medical Association for the years 1978–2000 and the Medical Group Management Association for the years 1994–2005). 4.2.3. Cost per Inpatient Day over CPI An independent measure of health care inflation reported by the American Hospital Association (sources: Goldman and McGlynn 2005, and the BLS). 4.2.4. Insurance-Premium Malpractice Index over CPI A measure of malpractice insurance costs for the providers of medical services. It is the average change in the cost of malpractice insurance for three physicians’ specialties: General Medicine, General Surgery, and Obstetrics/Gynecology. Data from 1976 to 1986 are from Danzon (1991). Data from 1987 to 1990 are from Harrington, Danzon, and Epstein (2008). Data from 1991 to 2007 are from authors’ computations from insurance costs for the specialties (Medical Liability Monitor Reports) over 51 states. 4.2.5. Private Health Insurance Premium per Enrollee over CPI The cost of health insurance for a privately insured person (sources: Cohen et al. 2009, the US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.6. Cost of Medicare per Enrollee over CPI Government's expenses per publicly insured elderly person (sources: The US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.7. HCE per Capita over CPI Real HCE per US inhabitant (sources: US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.8. CPIMC over CPI A measure of excess inflation in the US medical care sector with respect to the average inflation in the US economy (source: The BLS). Observe from this table that Health Care Sector Markup over Economy Markup moves remarkably close to the CPIMC over CPI. The CPIMC displays a modest growth in comparison with other measures of medical costs, except for Average Physician Compensation that grows much less. These price measures grow much faster over the 1980–1990 period. The Cost per Inpatient Day over the CPI shows a steep cumulative increase of about 70% over the 1978–1990 transition period, whereas the CPIMC over CPI presents a cumulative increase of around 35% over the same period. Malpractice insurance premiums almost triple during the 1978–1990 transition period. The health economics literature (e.g., Harrington et al. 2008) usually refers to the 1980s as the “crisis” of medical malpractice because of the frequency and severity of claims and the dramatic increase in the cost of malpractice insurance. A main argument against using malpractice as an explanatory variable is that these costs are rather low: About 1.25% of HCE expenditures.15 Private Health Insurance per Enrollee has grown faster than HCE per Capita over the entire sample period. This increment came along with decreasing out-of-pocket expenditures—albeit out-of-pocket expenditures have still grown faster than nominal income per capita. Private Health Insurance per Enrollee over CPI more than doubles over the 1978–1990 transition period. Changes in Private Health Insurance per Enrollee over CPI are contemporaneously correlated with changes in CPIMC over CPI (a correlation coefficient of 0.43). The Cost of Medicare per Enrollee and HCE per Capita have moved together over the sample period. Hence, costs in the private and public sectors may be driven by the same set of common factors. Since the distribution of the population over the categories of uninsured, privately, and publicly insured remained quite stable over the last three decades, it appears that aging and the size of the elderly population cannot possibly account for the observed increase in the US medical care expenditure gap over our sample period. To sum up, except for aggregate markups and average physician compensation, related measures of medical care costs show much steeper increases than the CPIMC over both the entire sample period and the 1978–1990 transition period. On these grounds, the CPIMC seems a conservative measure of health care costs. We should remark that these measures of medical care costs have been obtained from unrelated data sources. 4.3. Microeconomic Evidence of International Health Care Prices Several studies and international institutions have reported marked cross-country differences in health care prices. Figure 13 presents costs of several health care items in five OECD countries as of year 2011. These costs are reported as fractions of the corresponding US cost. Switzerland exhibits the second highest costs: Around 65% of US dollar costs. For the remaining countries in the sample, the unitary prices observed are around one fifth of US costs. In these comparative studies it becomes quite difficult to control for quality across countries. There is, however, a large international market for prescription drugs, and the cross-country variability in wholesale drug prices has been well documented. An early study by Jacoby and Hefner (1971) reported prices for twenty drugs in nine countries. The study confirms a great variation from country to country for a single product by the same manufacturer. Some on-patent drugs were even three times cheaper outside the United States. Using 121 drugs to compare US prices with those of Canada and 76 drugs to compare with the United Kingdom, two GAO reports (GAO 1992, 1994) found that manufacturers prices in the United States were on average 32% higher than in Canada and 60% higher than in the United Kingdom. Later research has expanded the range of sample products to provide accurate price measures. Danzon and Furukawa (2003) find that manufacturer drug prices in eight representative countries are usually between 20% and 40% lower than in the United States. Although on-patent brands may be almost 50% cheaper in some of these countries, generic drugs are usually cheaper right after patent expiration, but later on they become relatively more expensive than in the United States (Danzon and Furukawa 2011). A related study by the US Department of Commerce (2004) on patented prescription drugs reports price indexes that could be 50% lower than their US counterparts (op. cit., p. 38). Hence, even for tradable products of a given quality, medical prices in the United States are much higher. 5. Concluding Remarks In this paper we present a macro-level difference-in-difference approach to study the evolution of US health care expenditures. Since 1980 the United States features the highest health care expenditure share of all the OECD countries. With the approaching retirement of the baby-boom generation, it is feared that the US medical care expenditure gap may take up a new expansionary turn. Thus, managing health care expenditure growth has become a topic of national concern, and a tall order for balanced economic growth. To guide this discussion, we examine a sample of eleven OECD countries with similar income and quality data over the sample period 1970–2007. We define the US medical care expenditure gap as the US medical expenditure share in nominal GDP over the average share of the other OECD countries. We distinguish the following time periods: (i) the 1970–1977 period: The US medical care expenditure gap hovered around 1.25, (ii) the 1978–1990 period: The US medical care expenditure gap increased steadily from 1.23 to 1.68, and (iii) the 1991–2007 period: The US medical care expenditure gap stabilized around 1.70. Hence, the 1978–1990 period stands out as a transition episode with the sharpest increase in the US medical care expenditure gap. One major difficulty in uncovering these growth patterns for US medical expenditures appears to be the high degree of uniformity in the various National Health Expenditure Accounts’ categories by source and use of funds. These regular trends prevail over the 1978–1990 transition period. When adjusting for inflation in the medical sector, we obtain that the ratio of real health care expenditures over real GDP is quite flat over the entire sample period, but we observe some slight growth in the share of US value added. We first perform a growth accounting exercise based on value added to assess the contribution of observable production factors. This exercise focuses on the supply side of the economy, and hence it has the advantage of circumventing obvious issues related to price and quality data as well as the various degrees and forms of government intervention across countries. As is well known, growth accounting methods become most powerful under a stable influence of technological progress across economies, and a similar composition and productivity of labor and capital over time. The ratio of physician to non-physician health care workers has been quite stable in our sample of countries for the periods in which data is available. Also, in spite of global growth in markups, capital, and labor shares, trends have evolved quite similarly within our sample of countries. For the 1970–2007 sample period, labor in the health care sector has grown faster in the OECD, whereas capital in the health care sector has grown faster in the United States. Overall, our growth accounting exercise based on value added indicates that the US medical care expenditure gap should have gone down in the period 1970–2007, whereas it went up by about 30% in the data. Therefore, the divergent patterns of growth of health care expenditures between the United States and the OECD cannot be explained by factor accumulation. We then provide some supporting evidence to account for the wedge between growth in value added and US health care expenditures. Various international institutions such as the OECD and the World Health Organization periodically report cross-country rankings on various dimensions of health status, health care activities, quality of care, and access to care. In many of these rankings the US lags behind many other OECD countries. Hence, the assumption of similar growth in technology and product quality across countries seems quite plausible at the aggregate level. Moreover, it is unclear that technology could account for the sharp break in the trend of the US medical gap over the 1978–1990 transition period—just spanning over a single decade. Indeed, US medical care expenditures are characterized by large corporate markups, gross profits, and stock market returns. Periods of strong positive growth in the ratio of HCE over GDP are backed by contemporaneous and commensurate changes in gross profits and markups. When we filter out price effects we find that the growth of the product quality residual in the US falls into the OECD average. Whereas medical prices in some OECD countries followed the general pace of inflation, US medical inflation has considerably outpaced general inflation. The health economics literature has identified various factors that account for the higher prices of prescription drugs in the United States, which may also extend to the observed high prices of hospital and other medical services. Health insurance premiums present much higher price increases than the CPI-medical care. Increasing charges in hospital and related services appear to be a main driving force in the US health care sector over the 1978–1990 transition period. Finally, our methods impose severe limitations on the form of the health care production function. A statistical-based approach could also identify main drivers of health care expenditure patterns. In fact, in our supporting exercises we have reported several estimates of price elasticities for health care demand, and one would need to check whether or not these elasticities are statistically significant across countries. In unreported results available upon request, we have run panel data regressions for the US medical care expenditure gap with respect to each individual country as the dependent variable and the corresponding US gaps for the following four independent variables: The medical care price, GDP per capita, population 65 years old and over, and life expectancy. In this strongly balanced panel with ten cross-sectional observations and 38 time-series observations, we find that only the US medical care price gap and the US GDP per capita gap are statistically significant. For our sample period, however, the US GDP per capita gap went down by about 10%, and hence it works in the opposite direction. As discussed previously when considering gaps by country, this income gap may pick up a catching-up effect as a result of converging GDP per capita trends in the OECD. Appendix A: A Simple Model of Health Care Expenditures Let us just consider an endowment economy with a representative agent. The economy can also be reinterpreted as a social planning problem for the optimal allocation between a variety of medical services and all the other goods. A key parameter in the model is the price elasticity for health care demand. We should note that even if the health care good is subsidized the price elasticity of demand will not be changed when the representative agent has to pay a constant fraction of medical expenditures. At every time t = 0, 1, …, the economy receives yt units of an aggregate commodity that can be transformed into two types of consumption goods: A composite consumption good c and a variety of health care consumption goods ms for s ∈ [0, σ(at)], where at denotes the technology level at time t and [0, σ(at)] is the mass of available varieties. Preferences are represented by a CES utility function. All health care varieties mts at time t enter symmetrically into a utility aggregator |$M(t)=\big[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\big]^{\frac{1}{\gamma }}$|⁠. The representative agent solves the following budget-constrained maximization problem: \begin{eqnarray*} &&{\max _{c,h,m_{s}} \quad \sum _{t=0}^{\infty }\beta ^{t}\left[\!\lambda c_{t}^{\rho }+(1-\lambda )\left(\phi (a_{t})^{\frac{1}{\rho }}\left[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\right]^{\frac{1}{\gamma }}\right)^{\rho }\!\right]^{\frac{1}{\rho }},}\\ &&\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\text{subj. to:} \quad\,\,\, c_{t}+q_{t}h_{t}=y_{t},\\ &&\,\,\,\, \int _{0}^{\sigma (a_{t})}m_{ts}ds=a_{t}h_{t}, \\ &&\,\,\,\, 0<\beta <1,0<\lambda <1,0<\gamma <1,-\infty <\rho <1, \end{eqnarray*} where qtht represents nominal health care expenditures, qt is the relative price, and ht represents real expenditures. Parameter λ is called the consumption share parameter. Parameter ρ determines the degree of substitution between the composite consumption good c and the health care utility aggregator |$M_{t}=[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds]^{\frac{1}{\gamma }}$|⁠. Parameter γ determines the degree of substitution of the health care varieties ms. Function ϕ(at) allows for shifts in the expenditure share as a result of technological change. An increase in the technology level at may change the composition of expenditures through the following three channels: (i) the price effect: An increase in at lowers the relative price of health care varieties qt/at in terms of the numeraire good. This effect is present in economic growth models of embodied technological change (e.g., Greenwood, Hercowitz, and Krusell 1997). (ii) the productivity effect: An increase in at results in higher productivity because it expands the mass of available varieties [0, σ(at)] to allow for a more efficient production of health care utility |$M(t)=\big[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\big]^{\frac{1}{\gamma }}$|⁠. That is, for 0 < γ < 1, the same utility level Mt can be obtained under lower spending. This effect is present in economic growth models with a continuum of product varieties (cf., Romer 1990). And (iii) the expenditure effect: An increase in at may shift the consumption expenditure share because technological change may expand the domain of application of health care varieties. This effect is reflected in function ϕ(at), and allows for an increase in the health expenditure share under a unitary income elasticity as documented in several studies (Chernew and Newhouse 2012). Note that for an inelastic demand (i.e., ρ < 0) both the price and productivity effects (i)–(ii) may lead to a decrease in the health care expenditure share under an increase in a. Hence, function ϕ(a) builds in some further flexibility to model the effects of a change in at on health care expenditures qtht. There are some other well-known models in which technological change may generate nonlinear Engel curves (e.g., Becker, Philipson, and Soares 2005; Hall and Jones 2007). Although this rich form for the utility function contemplates various channels for the influence of technological change, we will again collect all of these effects as part of the nonobservable residual. Optimality Conditions. The representative agent assumes that the relative price qt and the level of technological change at are exogenously given. Over the optimal solution, consumption must be constant across medical varieties, that is, mts = mt for all s. Then, from the first-order conditions of the agent's optimization problem we obtain the optimal ratio of average consumption of health care varieties mt over the composite consumption of all other goods ct, \begin{equation*} \frac{m_{t}}{c_{t}}=\left(\frac{1}{q_{t}}\right)^{\frac{1}{1-\rho }}\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}\left(a_{t}\phi (a_{t})\sigma (a_{t})^{\frac{\rho -\gamma }{\gamma }}\right)^{\frac{1}{1-\rho }}. \end{equation*} Now, multiplying both terms by relative price qt, we can express the ratio of health care expenditures over total expenditures in non-health care goods as follows: \begin{equation*} \frac{q_{t}h_{t}}{c_{t}}=\left(\frac{1}{q_{t}}\right)^{\frac{\rho }{1-\rho }}\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}a_{t}^{\frac{\rho }{1-\rho }}\phi (a_{t})^{\frac{1}{1-\rho }}\sigma (a_{t})^{\frac{\rho (1-\gamma )}{\gamma (1-\rho )}}. \end{equation*} These equation provides an expression for the evolution of health care expenditures relative to non-health care expenditures as a function of the relative price between the two goods and a residual term |$\widehat {A}{}_{t}$| collecting the underlying quality effects, \begin{equation*} \widehat {A}_{t}=\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}a_{t}^{\frac{\rho }{1-\rho }}\phi (a_{t})^{\frac{1}{1-\rho }}\sigma (a_{t})^{\frac{\rho (1-\gamma )}{\gamma (1-\rho )}}. \end{equation*} Appendix B: Main Definitions Data Sources Health Care Expenditures (OECD Data). “The expenditure on activities that—through application of medical, paramedical, and nursing knowledge and technology—has the goals of: Promoting health and preventing disease; Curing illness and reducing premature mortality; Caring for persons affected by chronic illness who require nursing care; Caring for persons with health-related impairments, disability, and handicaps who require nursing care; Assisting patients to die with dignity; Providing and administering public health; Providing and administering health programs, health insurance and other funding arrangements. With this boundary, general public safety measures such as technical standards monitoring and road safety are not considered as part of expenditure on health. Activities such as food and hygiene control and health research and development are not included in total health expenditure.” (OECD Health Data 2012 Definitions, Sources and Methods; available at http://stats.oecd.org) Medical Care Prices. Medical care prices from 1970 to 1977 for the eleven OECD countries in the sample are from Gillion, Schieber, and Poullier (1985). Remaining data come from the following sources: Australia. Data from 1989 to 2007 corresponds to the health component of the CPI (available from the Australian Bureau of Statistics at http://www.abs.gov.au). Data from 1978 to 1988 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). Canada. Data corresponds to the health and personal care price index component of the CPI (available from Statistics Canada at http://www76.statcan.gc.ca). Denmark. Data from 1996 to 2007 corresponds to the CPI-health price, available at the FRED database (Federal Reserve Bank of St. Louis). Data from 1978 to 1995 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). Finland. Data from 1996 to 2007 corresponds to the CPI-health price, available at the FRED database (Federal Reserve Bank of St. Louis). Data from 1978 to 1995 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). France. The CPI-medical care is the union of the following three price indexes: (i) the health services up to year 1992 (and then discontinued), (ii) the medical services and health care expenditures up to year 1998 (and then discontinued), and (iii) the health services from 1998 to 2007 (available from the French National Institute for Statistics and Economic Studies at http://www.bdm.insee.fr). Germany. Data up to 1983 is from Gillion et al. (1985). Data from 1991 until 2007 corresponds to the health component of the CPI (available from the German Federal Statistics Office at https://www.destatis.de/). Data for the missing period 1984–1990 has been interpolated using data from Schieber et al. (1994). Ireland. Data corresponds to the health subcategory of the CPI, which is available from the Irish Central Statistics Office at http://www.cso.ie/. Japan. Data corresponds to the medical care item of the CPI [available from the Japanese Statistics Bureau, Director-General for Policy Planning (Statistical Standards) and Statistical Research and Training Institute at http://www.stat.go.jp/]. Spain. The CPI-medical care is constructed as the union of several price indexes. From 1977 to 1992 we used the CPI-medicine (IPC-medicina). For the period 1993–2001 we used the average change in five subcategories of the CPI: medical, dental, and non-hospital paramedical services (servicios médicos, dentales y paramédicos no hospitalarios), drugs and other pharmaceutical products (medicamentos y otros productos farmacéuticos), machines, therapeutic material and its repairs (aparatos y material terapéutico y sus reparaciones), hospital care (cuidados en hospitales y similares), and medical insurance (seguros médicos). From 2002 to 2007 we used the average change in the three available subcategories: Drugs, pharmaceutical products and therapeutic material (medicamentos, otros productos farmacéuticos y material terapéutico), medical, dental, and non-hospital paramedical services (servicios médicos, dentales y paramédicos no hospitalarios), and hospital services (servicios hospitalarios). Available from the Instituto Nacional de Estadística at http://www.ine.es/. United Kingdom. Data up to 1983 is taken from Gillion et al. (1985). Data from 1988 until 2007 corresponds to the health component of the CPI (available from the Office for National Statistics at http://www.ons.gov.uk/). Data from the missing period 1984–1987 has been interpolated using data from Schieber et al. (1994). United States. Data corresponds to the CPI-medical care (available from the Bureau of Labor Statistics at http://www.bls.gov/). Acknowledgments We would like to thank an editor and four anonymous reviewers for their constructive comments. Notes The editor in charge of this paper was Dirk Krueger. Footnotes 1 As discussed in what follows, changes in relative prices and public policies can account for some of these expenditure patterns. 2 As documented in Cutler and Ly (2011), three major factors may account for the high level of HCE in the United States: (i) Higher cost of production factors, (ii) Greater share of administrative expenses, and (iii) More generous provision of medical services. As our main goal is to assess the evolution of the US medical care expenditure gap, we shall be concerned with all sources of factor accumulation in the health care sector, and so we focus on growth in value added rather than on actual levels of all these economic variables. 3 The economics literature usually estimates technological change from price data. According to the Boskin Commission Report (1996) the US consumer price index does not take into account a 3% yearly increase in the quality of health care goods. Unfortunately, we were not able to find data on consumption quality for the other OECD countries. 4 We calculate average worker compensation as the ratio of compensation of employees (NIPA Table 1) over full-time and part-time employees (NIPA Table 6.4). These NIPA tables are available at http://www.bea.gov/iTable. 5 Health care investment as percentage of HCE data are available from OECD Health Care Statistics (http://stats.oecd.org/). This item comprises gross capital formation of domestic health care provider institutions (excluding retail sale and other providers of medical goods). Health care investment could be privately or publicly financed. 6 Between 1970 and 2007, for the value-weighted portfolio of all the publicly traded companies in the United States (excluding ADRs and REITs), the yearly gross return was 11.2%, which translates into a yearly rate of return of approximately 8% after adjusting for inflation (data available from the Center for Research in Security Prices). 7 In unreported simulations available upon request, we estimate αH for each country in our sample using the World Input-Output Database tables (http://http://www.wiod.org) for the year 2000. We found that the capital income share αH in the health care sector for Canada, Germany, Japan, and the United States are quite similar (0.23, 0.21, 0.26, and 0.23 respectively), whereas those of Australia, Spain, and the United Kingdom are significantly lower (0.12, 0.16, and 0.10 respectively). Allowing αH to vary across countries does not change our qualitative results. If anything, it strengthens our conclusion that factor accumulation cannot account for the evolution of the US medical care expenditure gap. 8 See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/det_49_ind_port.html. The list of companies used to construct the health care portfolio and SIC codes is also available from the authors upon request. 9 A firm's market capitalization is the total number of shares outstanding times the price. 10 The return of the market weighted portfolio also comes from Kenneth French database. See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_factors.html. 11 Infant mortality is the yearly number of deaths of children under 1 year of age per 1,000 live births. Neonatal mortality is the yearly number of deaths of children under 28 d of age per 1,000 live births. Perinatal mortality is the yearly number of deaths of children within 1 week of birth (early neonatal deaths) plus fetal deaths of minimum gestation period of 28 weeks or minimum fetal weight of 1,000 grams per 1,000 births (source: http://stats.oecd.org/). 12 Except for life expectancy, the set of countries used to construct the quality estimates in Table 4 is different from the eleven countries in the original comparison group. Since we could not find data for some of the countries in our original sample, we decided to use all the OECD countries with available data. Five-year observed survival rates are defined as the number of people diagnosed with cancer (age 15–99) within a certain period surviving five years after diagnosis, over the number of people diagnosed with the same cancer (age 15–99) within a certain period. Five-year relative survival rates are defined as the observed rate of people diagnosed with cancer (age 15–99) surviving five years after diagnosis, over an expected survival rate of a comparable group from the general population (see http://stats.oecd.org/). 13 Although the value of the standard deviation over the mean for life expectancy is 0.024 for our sample of OECD countries, this same ratio for all countries in the world in 2003 is about 0.145, around six times larger (data available at http://data.un.org/). There is no general data about survival rates for different types of cancer across countries. The American Cancer Society (2008) publishes some statistics for a few developing countries, and there is much more variability in survival rates than in the OECD. For example, the breast cancer 5-year relative survival rates in 2008 for China, India, Thailand, and Uganda are 0.82, 0.52, 0.63, and 0.46, respectively. For the sample of 18 countries reported in Table 4 the minimum is 0.79 and the maximum is 0.89. 14 This is in line with our analysis of financial variables in Section 2.3, where after the 1978–1990 transition period we observe that relative markups between the US health care sector and the rest of the economy have an upward trend whereas relative gross profits remain quite flat. Thus, the evidence suggests that in relative terms, real medical consumption may have declined in the United States because of cost containment whereas prices in the health care sector were still outpacing those of the rest of the economy. 15 Kessler (2011) argues that although the indemnity payments and administrative expenses of the malpractice system amount to less than 1% of health care spending, the costs of defensive medicine are likely to be far greater because neither patients nor physicians bear most of the marginal costs. References American Cancer Society ( 2008 ). Global Cancer Facts and Figures , 2nd ed . American Cancer Society . Anderson G. , Reinhardt U. , Hussey P. , Petrosyan V. ( 2003 ). “ It's the Prices, Stupid: Why the United States is So Different from other Countries .” Health Affairs , 22 , 89 – 105 . Google Scholar Crossref Search ADS PubMed Anderson G. , Frogner B. ( 2008 ). “ Health Spending in OECD Countries: Obtaining Value per Dollar .” Health Affairs , 27 , 1718 – 1727 . Google Scholar Crossref Search ADS PubMed Becker G. , Philipson T. , Soares R. ( 2005 ). “ The Quantity and Quality of Life and the Evolution of World Inequality .” American Economic Review , 95 (1) , 277 – 291 . Google Scholar Crossref Search ADS PubMed Berndt E. , Cutler D. , Frank R. , Griliches Z. , Newhouse J. , Triplett J. ( 2000 ). “ Medical Care Prices and Output .” Handbook of Health Economics , Vol. 1 , edited by Culyer A. , Newhouse J. , Elsevier, Amsterdam, pp. 119 – 180 . Berndt E. , Aitken M. ( 2010 ). “ Brand Loyalty, Generic Entry and Price Competition in Pharmaceuticals in the Quarter Century After the 1984 Waxman-Hatch Legislation .” NBER Working Paper 16431, National Bureau of Economic Research, Cambridge, MA. Boskin Commission Report ( 1996 ). “ Toward a More Accurate Measure of the Cost of Living .” Final Report to the Senate Finance Committee from the Advisory Commission to Study the Consumer Price Index , http://www.ssa.gov/history/reports/boskinrpt.html. Cardenas E. ( 1996 ). “ Revision of the CPI Hospital Services Component .” BLS Monthly Labor Review , 119 , 40 – 48 . Chandra H. Skinner J. ( 2012 ). “ Technology Growth and Expenditure Growth in Health Care .” Journal of Economic Literature , 50 , 645 – 680 . Google Scholar Crossref Search ADS Chernew M. Newhouse J. ( 2012 ). “ Health Care Spending Growth .” Handbook of Health Economics , Vol. 2 , edited by Pauly M. Mcguire T. Barros P. , Elsevier, Amsterdam, pp. 1 – 43 . Cohen R. Makuc D. Bernstein A. Bilheimer L. Powell-Griner E. ( 2009 ). “ Health Insurance Coverage Trends, 1959–2007: Estimates from the National Health Interview Survey .” National Health Statistics Reports 17 . Comin D. Hobijn B. ( 2004 ). “ Cross-Country Technology Adoption: Making the Theories Face the Facts .” Journal of Monetary Economics , 51 , 39 – 83 . Google Scholar Crossref Search ADS Comin D. Hobijn B. ( 2010 ). “ An Exploration of Technology Diffusion .” American Economic Review , 100 (5) , 2031 – 2059 . Google Scholar Crossref Search ADS Cutler D. Ly D. ( 2011 ). “ The (Paper) Work of Medicine: Understanding International Medical Costs .” Journal of Economic Perspectives , 25 (2) , 3 – 25 . Google Scholar Crossref Search ADS PubMed Cutler D. McClellan M. Newhouse J. ( 2000 ). “ How Does Managed Care Do It? ” RAND Journal of Economics , 31 , 526 – 548 . Google Scholar Crossref Search ADS PubMed Danzon P. ( 1991 ). “ Liability for Medical Malpractice .” Journal of Economic Perspectives , 5 (3) , 51 – 69 . Google Scholar Crossref Search ADS Danzon P. Chao L. ( 2000 ). “ Cross-National Price Differences for Pharmaceuticals: How Large, and Why? ” Journal of Health Economics , 19 , 159 – 195 . Google Scholar Crossref Search ADS PubMed Danzon P. Furukawa M. ( 2003 ). “ Prices and Availability of Pharmaceuticals: Evidence from Nine Countries .” Health Affairs , 22 , 521 – 536 . Google Scholar Crossref Search ADS Danzon P. Furukawa M. ( 2011 ). “ Cross-National Evidence on Generic Pharmaceuticals: Pharmacy vs. Physician-Driven Markets .” NBER Working Paper 17226 , National Bureau of Economic Research, Cambridge, MA. De Long B. Summers L. ( 1992 ). “ Equipment Investment and Economic Growth: How Strong is the Nexus? ” Brookings Papers on Economic Activity , 2 , 157 – 199 . Google Scholar Crossref Search ADS Di Matteo L. ( 2005 ). “ The Macro Determinants of Health Expenditures in the United States and Canada: Assessing the Impact of Income, Age Distribution and Time .” Health Policy , 71 , 23 – 42 . Google Scholar Crossref Search ADS PubMed Dranove D. Lindrooth R. White W. Zwanziger J. ( 2008 ). “ Is the Impact of Managed Care on Hospital Prices Decreasing? ” Journal of Health Economics , 27 , 362 – 376 . Google Scholar Crossref Search ADS PubMed Drösler S. Romano P. Wei L. ( 2009 ). “ Health Care Quality Indicators Project: Patient Safety Indicators Report 2009 .” OECD Health Working Paper 47 , Organisation for Economic Co-operation and Development, Paris, France. Dunn A. Shapiro A. ( 2011 ). “ Physician Market Power and Medical-Care Expenditures .” BEA Working Paper 84 , Bureau of Economic Analysis, Washington, DC. Finkelstein A. ( 2007 ). “ The Aggregate Effects of Health Insurance: Evidence from the Introduction of Medicare .” Quarterly Journal of Economics , 122 , 1 – 37 . Google Scholar Crossref Search ADS Fixler D. Jaditz T. ( 2002 ). “ An Examination of the Difference Between the PCE Deflator and the CPI .” BLS Working Paper 361 , Bureau of Labor Statistics, Washington, DC. Garber A. Skinner J. ( 2008 ). “ Is American Health Care Uniquely Inefficient? ” Journal of Economic Perspectives , 22 (4) , 27 – 50 . Google Scholar Crossref Search ADS PubMed Gillion C. Schieber G. Poullier J. ( 1985 ). “ Measuring Health Care, 1960–1983: Expenditure, Costs, and Performance .” OECD Social Policy Studies 2 . Goldman D. McGlynn E. ( 2005 ). “ U.S. Health Care: Facts About Cost, Access, and Quality .” RAND Corporate Publication Series CP-484/1-(1/05) . Greenwood J. Hercowitz Z. Krusell P. ( 1997 ). “ Long-Run Implications of Investment-Specific Technological Change .” American Economic Review , 87 (3) , 342 – 362 . Griliches Z. Cockburn I. ( 1994 ). “ Generics and New Goods in Pharmaceutical Price Indexes .” American Economic Review , 84 (5) , 1213 – 1232 . Hall R. Jones C. ( 2007 ). “ The Value of Life and the Rise in Health Spending .” Quarterly Journal of Economics , 122 , 39 – 72 . Google Scholar Crossref Search ADS Harrington S. Danzon P. Epstein A. ( 2008 ). “ Crises in Medical Malpractice Insurance: Evidence of Excessive Price-Cutting in the Preceding Soft Market .” Journal of Banking & Finance , 32 , 157 – 169 . Google Scholar Crossref Search ADS Jacoby E. Hefner D. ( 1971 ). “ Domestic and Foreign Prescription Drug Prices .” Social Security Bulletin , 34 , 15 – 22 . Kessler D. ( 2011 ). “ Evaluating the Medical Malpractice System and Options for Reform .” Journal of Economic Perspectives , 25 (2) , 93 – 110 . Google Scholar Crossref Search ADS PubMed Koechlin F. Lorenzoni L. Schreyer P. ( 2010 ). “ Comparing Price Levels of Hospital Services Across Countries: Results of Pilot Study .” OECD Health Working Papers 53 , Organisation for Economic Co-operation and Development, Paris, France. Koijen R. Philipson T. Uhlig H. ( 2016 ). “ Financial Health Economics .” Econometrica , 84 , 195 – 242 . Novy-Marx R. ( 2013 ). “ The Other Side of Value: The Gross Profitability Premium .” Journal of Financial Economics , 108 , 1 – 28 . Google Scholar Crossref Search ADS Ringel J. Hocek S. Vollaard B. Mahnovski S. ( 2000 ). “ The Elasticity of Demand for Health Care: A Review of the Literature and Its Application to the Military Health System .” National Defense Research Institute/RAND . Romer P. ( 1990 ). “ Endogenous Technological Change .” Journal of Political Economy , 98 , 71 – 102 . Google Scholar Crossref Search ADS Rodwin V. ( 2003 ). “ The Health Care System under French National Health Insurance: Lessons for Health Reform in the United States .” The American Journal of Public Health , 93 , 31 – 37 . Google Scholar Crossref Search ADS PubMed Rodwin V. Sandier S. ( 1993 ). “ Health Care under French National Health Insurance .” Health Affairs , 12 , 111 – 131 . Schieber G. Poullier J. Greenwald L. ( 1994 ). “ Health System Performance in OECD Countries 1980–1992 .” Health Affairs , 13 , 100 – 112 . Google Scholar Crossref Search ADS PubMed Skinner J. Fisher E. Wennberg J. ( 2005 ). “ The Efficiency of Medicare .” In Analyses in the Economics of Aging . University of Chicago Press , pp. 129 – 157 . Smith S. Newhouse J. Freeland M. ( 2009 ). “ Income, Insurance and Technology: Why Does Health Spending Outpace Economic Growth? ” Health Affairs , 28 , 1276 – 1284 . Google Scholar Crossref Search ADS PubMed US Department of Commerce ( 2004 ). Pharmaceutical Price Controls in OECD Countries . International Trade Administration , Washington, DC . US General Accounting Office ( 1992 ). Prescription Drugs: Companies Typically Charge More in the United States than in Canada . Government Printing Office , Washington, DC . US General Accounting Office ( 1994 ). Prescription Drugs: Companies Typically Charge More in the United States than in the United Kingdom . Government Printing Office , Washington, DC . US General Accounting Office ( 2003 ). Medical Malpractice Insurance: Multiple Factors Have Contributed to Increased Premium Rates . Government Printing Office , Washington, DC . © The Author(s) 2017. Published by Oxford University Press on behalf of European Economic Association. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the European Economic Association Oxford University Press

# Understanding Growth Patterns in US Health Care Expenditures

Journal of the European Economic Association, Volume 17 (1) – Feb 1, 2019
43 pages

Loading next page...

/lp/ou_press/understanding-growth-patterns-in-us-health-care-expenditures-8EWS1MfLlE
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of European Economic Association.
ISSN
1542-4766
eISSN
1542-4774
D.O.I.
10.1093/jeea/jvx059
Publisher site
See Article on Publisher Site

### Abstract

Abstract We study the steady upward trend of Health Care Expenditures (HCE) over GDP for a sample of OECD countries between 1970 and 2007. Although the United States is clearly an outlier, almost all of the additional increase in US HCE happened during the 1978–1990 period. We perform two growth accounting exercises to explore sources of variability of HCE over GDP across countries. In the first growth accounting exercise based on value added we find that factor accumulation is unable to replicate the observed growth patterns. We also show that the additional increase in markups in the US corporate medical sector mimics well the ratio of HCE over GDP in the United States. This suggests that differences in the relative price of health care—rather than technology, product quality, and factor accumulation—could explain the divergent growth patterns of HCE over GDP across these countries. In the second growth accounting exercise, we filter out prices from HCE over GDP, and confirm that there is very little variability for the product quality residual to explain the variation in HCE across countries. 1. Introduction One of the most noteworthy stylized facts for the US economy is the pronounced upward trend in the share of health care expenditures (HCE). Indeed, according to OECD data HCE accounted for about 7% of nominal GDP in 1970, whereas it is now about 18%. Mounting US medical expenditures have become an issue of national concern and a continuing challenge for policy makers (e.g., the Clinton Health Care Plan of 1993; the Bush Medicare Prescription Drug, Improvement, and Modernization Act of 2003; and more recently the 2010 Obama Affordable Care Act). We use a comparison group of ten other OECD countries with high quality data: Australia, Canada, Denmark, Finland, France, Germany, Ireland, Japan, Spain, and the United Kingdom. Our quantitative analysis aims to assess deviations of US medical expenditures from common trends over the sample period 1970–2007. We shall focus on the US medical care expenditure gap, which we define as the US share of HCE in nominal GDP over the average share of HCE in nominal GDP of the other OECD countries. Figure 1 illustrates the evolution of the US medical care expenditure gap for the sample period. The United States was among the top health care spenders in 1970 but still far from being an outlier. The US medical care expenditure gap displays a pronounced upward trend over the 1980s. The gap remains flat during the eight-year period of the Clinton presidency, and it goes up slightly at the beginning of the 2000s. Hence, this figure suggests the existence of three well differentiated periods: The 1970–1977 period: The US medical care expenditure gap hovered around 1.25. The 1978–1990 period: The US medical care expenditure gap increased steadily from 1.23 to 1.68. The 1991–2007 period: The US medical care expenditure gap roughly stabilized around 1.70. Figure 1. View largeDownload slide The US medical care expenditure gap 1970–2007. Source: OECD Health Data (June 2014). Figure 1. View largeDownload slide The US medical care expenditure gap 1970–2007. Source: OECD Health Data (June 2014). Of course, the 1978–1990 transition episode is certainly puzzling. Indeed, it does not seem plausible to explain the sharp increase in HCE over such a limited time span by some aggregate variables with smooth long-term trends such as GDP, life expectancy, the size of the elderly population, defensive medicine, the prevalence of some modern health habits and trends (e.g., fast food and obesity), and new medical treatments. The widespread use of better, more advanced technologies and product quality has been blamed for the high share of HCE in GDP in the United States. The evidence supports the idea that new technologies may diffuse quickly among countries with commensurate income levels. Besides, the increase in the US medical expenditure gap occurred just over a decade, and such pattern of growth will be hard to replicate by simple models of technology diffusion. Regarding higher product quality in the US health care sector, the United States typically trails behind in OECD rankings on various health care outcomes and indicators. This is still compatible with the widespread use of more sophisticated and less cost-efficient technologies in the United States. However, our study shows that inflation in the US medical sector is driven by large markups, and price increases do not come along with extra expenditures of resources. Indeed, in our first growth accounting exercise in what follows we show that the US medical care expenditure gap implied by factor accumulation runs in the opposite direction. One major difficulty in uncovering US HCE growth patterns is the high degree of uniformity in the categories of sources and uses of funds. Figure 2(a) breaks down HCE into various categories by source. Some public programs such as Medicare and Medicaid have gained prominence at the expense of private funding and out-of-pocket expenditures, but such extra growth seems too small to account for the general evolution of HCE. Figure 2(b) breaks down HCE into various categories by use. A substantial increase in the categories of other personal health care and prescription drugs is observed, as well as a slight decline in physician and clinical services, and hospital care.1 Again, these changes are just small departures from overall trends in HCE. Thus, for the matter in question, we would need to identify a group of variables that jointly drive all the expenditure categories. Two distinct candidates are prices and technology. Figure 2. View largeDownload slide The evolution of US health care expenditures by sources and uses 1980–2007. Source: US Department of Health and Human Services – Centers for Medicare and Medicaid Services. Figure 2. View largeDownload slide The evolution of US health care expenditures by sources and uses 1980–2007. Source: US Department of Health and Human Services – Centers for Medicare and Medicaid Services. To gain a further understanding of the macroeconomic factors generating the divergent patterns of the US medical care expenditure gap, we perform a growth accounting exercise based on value added. Three main methodological issues motivate this exercise. First, we shall be concerned with growth in factor quantities and this is compatible with cross-country differences in price and quality levels. Second, as observed in Figure 1, most of the incremental growth in US medical expenditures occurred during the 1978–1990 transition period, just over a decade. As discussed in what follows, it is unlikely that a large component of growth in HCE could be attributable to a sharp break in non-observable factors such as technological progress and product quality during such a short period of time. Further, this relatively short transition period is characterized by large increases in nominal and financial variables in the health care sector such as product markups. And third, our main analysis relies solely upon factor accumulation over quantities of inputs; hence, we need not take a stand on pricing policies, monopolistic behavior, the size of the government sector, and many other forms of market intervention and regulation at the aggregate level. We later address product quality and price effects as well as some rather complex measurements issues. In our sample period, capital displays higher growth in the United States whereas labor displays higher growth in the OECD. From our growth accounting exercise we conclude that there should not be much difference in the evolution of the actual shares of HCE in nominal GDP for both economies. That is, the different growth patterns of US health care expenditures cannot be explained by aggregate factor accumulation. The sizable wedge between the contribution of value added and HCE in the United States motivates various supporting exercises. We first look at the evolution of markups, gross profits, and stock returns over a sample of US companies in the health care sector. Excessive levels of profitability and of stock market returns usually happen at times of high growth in HCE, which lends credibility to the idea of existence of monopoly power and other market frictions in the US health care sector not generally observed in the other countries. Moreover, an extensive literature claims that the United States trails behind comparable OECD countries in terms of health outcomes and related health care indicators. Therefore, the different patterns of growth in HCE in the United States cannot be backed by corresponding changes in product quality. We should remark that we analyze variations of HCE over a group of countries with commensurate technological development and data quality, and we consider expenditure shares to condition by cross-country income levels. Several authors have pointed out that early diffusion of new technologies is frequently linked to similar GDP levels (Comin and Hobijn 2004, 2010; Smith et al. 2009). In addition, when we adjust for prices we observe that US real HCE does not grow much faster than US real Total Consumption. In fact, the evolution of US real HCE is fairly well matched by value added from factor accumulation, as well as by similar indicators of health care quantities in the other OECD countries. Hence, our main analysis does not depend on prices, but our basic findings do not change when we filter out prices and common quality trends. To understand the stability of real HCE we explore potential drivers of US medical care prices such as hospital and professional services, and pharmaceutical products. Physicians’ services, however, do not appear to be a major source of the continued inflation in the medical sector. Price markups attest that medical prices are disconnected from average variable costs. Anderson et al. (2003) link the high HCE share in the United States to prices. However, their period of analysis is 1990–2000 in which the US health care expenditure gap remained flat. Hall and Jones (2007) propose a model in which the HCE share goes up with income and technology growth with the end result of increasing life expectancy. As is well known, there has been some leveling of income by European countries. Moreover, US life expectancy has always been slightly below the OECD average. Hence, life expectancy could be instrumental in explaining global increasing trends in HCE, but other factors appear more adequate to replicate the observed differences at the cross-country level over shorter time spans. Some researchers (cf., Anderson and Frogner 2008) argue that even though the United States presents the highest ratios of medical spending among all OECD countries, its residents are not granted the highest value per dollar spent in health care—suggesting a higher level of inefficiency in the United States. Our analysis focuses on medical spending growth rates—rather than levels—within a selected group of OECD countries. In lieu of efficiency, we shall be concerned with the large wedge between the contribution of value added and HCE.2 Our paper suggests that monopoly power by the private sector appears to increase medical care prices, but it is not clear that public intervention might limit public health care expenditure whereas reducing costs as a result of monopsony power since public policies could also lead to over consumption by some parts of the population because of expansionary health care insurance programs. Indeed, the spread of health care insurance has been claimed as an explanatory variable for the growth in US medical expenditures. Finkelstein (2007) estimated that between 1965 and 1970 the introduction of Medicare produced an increase in hospital spending six times larger than a private insurance program would have produced. At roughly the same time, however, many OECD countries did undergo notable expansions of their universal health care systems (e.g., Japan in 1961, Denmark in 1973, and Spain in 1986). Thus, we are led to conclude that expansionary trends in insurance markets cannot account for observed differences in HCE between the United States and the rest of the OECD countries. Some international medical reforms will be echoed in our quantitative studies. Ample literature has linked the rapid increase in HCE in the United States to technological change.3 For example, Di Matteo (2005) finds that technological change accounts for two thirds of health care spending growth over the 1975–2000 period. An excellent review of this literature is presented in Chernew and Newhouse (2012). Again, in our study, we control for the impact of technological change by conditioning on a sample of developed countries with comparable technologies. Chandra and Skinner (2012) claim that simply attaching the label of “technological growth” as the major cause of increases in health care expenditure is not a meaningful way to address the problem. Our macro-level difference-in-difference approach should be of interest for the health micro literature that has tried to answer the same question of excessive HCE in the United States but typically uses more fragmented evidence and does not focus on the 1978–1990 transition period. Chandra and Skinner (2012) follow Hall and Jones (2007) and propose a model with three types of health care goods that differ in cost-efficiency. They suggest that growth in US HCE is mostly driven by the widespread use of the least cost-efficient medical treatments embedded in new technologies. In their view, the adoption of these new technologies comes from the financing of health care costs through the insurance industry. Other health care systems with central budgeting or quantity constraints are not likely to experience rapid growth in these technologies. Garber and Skinner (2008) review some micro studies and claim that both cost and allocation efficiency between health care consumption and other consumption is difficult to answer. They also claim that other countries have shared the enormously valuable improvement in health—enjoyed by Americans in recent decades—at a much lower cost. Our study shows that an increasing US medical care expenditure gap does not stem from an extended use of capital and labor in the United States. Thus, it appears that new technologies (more prevalent in the United States) are typically associated with higher price markups. Hence, large product markups support the idea that market power—rather than the extra cost to generate higher product quality—might be an important determinant of US health care prices. Hospital and Related Services appears to be a driver of inflation in the US health care sector over the 1978–1990 transition period. Health insurance premiums underwent steep increases—considerably higher than the consumer price index (CPI) and the CPI-Medical Care (CPIMC) index. The paper is organized as follows. In Section 2 we highlight some stylized facts about the evolution of labor and capital in the health care sector, and propose a basic growth accounting exercise based on value added. The implied US medical care expenditure gap remains fairly constant over the various time periods. In Section 3 we support this growth accounting exercise by looking at various indicators of health care quality. After filtering out price effects we confirm the stability of US real medical consumption over total consumption. In fact, adjusting for common quality trends and medical care inflation does not change our basic empirical findings about the implied evolution of the US medical care expenditure gap. In Section 4, we reexamine various independent measures of prices and sources of US health care inflation. We conclude in Section 5. 2. A Basic Growth Accounting Exercise for the US Health Care Sector In our first pass to the data, we present a growth accounting exercise using the evolution of labor and capital for a restricted sample of countries. Moreover, for comparison purposes and to circumvent certain issues with some missing data, we also conduct a parallel growth accounting exercise for the US economy only. We find that labor and capital accumulation have no impact on the differential increase of the US medical care expenditure gap. Actually, the predicted gap from factor accumulation runs in favor of the other countries. Over the sample period 1970–2007, employment in the health care sector relative to total employment in the economy increased by 45% in the United States and by 88% in the OECD, whereas average investment in the health care sector relative to total investment in the economy is about 3.7% in the United States and 2.1% in the OECD. Labor income in the health care sector for the United States is about 9% of total labor income, and labor income in the health care sector for the OECD countries is about 8% of total labor income. Therefore, the contribution of value added from factor accumulation to HCE in the United States is relatively small since the share of HCE in GDP is about 18%. 2.1. Evolution of Labor and Capital in The Health Care Sector Table 1 reports various dimensions of the evolution of labor and capital in the health care sector for some selected dates. In row (i) we observe that total physicians’ compensation over HCE has decreased from almost 15% in 1982 to 13.1% in 2000. Physicians’ compensation over the average worker compensation in the economy increased between 1980 and 1990 but ended at a similar ratio in 2000 [row (ii)].4 Row (iii) shows the evolution of the average compensation of a non-physician health care worker with respect to a physician. This ratio remains relatively flat over the sample period. Table 1. Labor and capital statistics. 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% Notes: a. Data from 1982. b. Average over the next 10 years. Sources: AMA, AMGA, Bureau of Labor Statistics, US Census Bureau and OECD Health Data. View Large Table 1. Labor and capital statistics. 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% 1970 1980 1990 2000 (a) Labor productivity and compensation in the United States (i) Total physicians’ compensation over HCE N/A 14.8%a 14.1% 13.1% (ii) Average physician compensation over average worker compensation N/A 4.95a 5.77 5.16 (iii) Average non-phys. compensation over average physician compensation N/A 11.4%a 11.3% 12.0% (b) International comparisons of labor and capital (iv) US physicians per 1,000 workers 3.46 3.79 4.25 4.80 (v) OECD physicians per 1,000 workers 3.05 4.03 4.96 5.36 (vi) US health care investment over HCEb 7.04% 5.57% 4.85% 4.35% (vii) OECD health care investment over HCEb 5.52% 4.33% 4.25% 4.07% (viii) US aggregate investment over GDPb 22.02% 22.97% 20.98% 21.76% (ix) OECD aggregate investment over GDPb 26.46% 24.34% 23.15% 22.92% Notes: a. Data from 1982. b. Average over the next 10 years. Sources: AMA, AMGA, Bureau of Labor Statistics, US Census Bureau and OECD Health Data. View Large Of course, HCE growth has expanded the labor force in the sector. In the 1970–2000 period, the number of active US physicians increased from 3.46 per 1,000 workers to 4.80 per 1,000 workers [row (iv)], whereas the number of active OECD physicians increased from 3.05 per 1,000 workers in 1970 to 5.36 per 1,000 workers in 2000 [row (v)]. The ratio of physician to non-physician health care workers and their income shares have been quite stable in our sample of countries for the periods in which data is available. With the introduction of Medicare and the subsequent expansion of the health care sector there was a considerable increase in the number of non-physician workers in the US health care industry in the late 1960s. For the other OECD countries, we have data for non-physician health care workers only after 1985. OECD and US data show that the ratio of physician to non-physician health care workers remained almost constant for the period 1985–2007. From our own computations from other US data sources, between 1970 and 1980 this ratio decreased from 13% to 8.5%, and between 1980 and 2007 it slowly decreased from 8.5% to 7.3%. In our growth accounting exercise we assume that the ratio of physician health care workers to non-physician health care workers remains flat in both the United States and the OECD countries. When it comes to physical capital in the health care sector,5 growth runs in favor of the United States. Rows (vi) and (vii) report the average investment share in the health care sector in each decade between 1970 and 2000, whereas rows (viii) and (ix) report the average share of aggregate investment. Both the United States and OECD show a decreasing investment pattern in the health care sector, but the US share has always been above—albeit the difference has been diminishing over time. Therefore, labor in the health care sector has grown faster in the OECD whereas the corresponding physical capital investment share is greater in the United States. We now lay down a growth accounting exercise based on factor accumulation to assess the observed growth patterns in the US medical care expenditure gap. 2.2. Growth Accounting Based on Value Added Growth accounting based on value added circumvents the use of aggregate prices that may be picking up quality effects for the provision of goods and services. Because of data availability, in this section we only use seven countries in our sample: Australia, Canada, Germany, Japan, Spain, the United Kingdom, and the United States. 2.2.1. Growth Accounting Using Capital and Labor Data for the United States and the OECD Let us assume that production in the overall economy and in the health care sector evolve according to the following Cobb–Douglas production functions: \begin{eqnarray*} Y&=&AK^{\alpha }L^{1-\alpha },\\ Y_{H}&=&A_{H}K_{H}^{\alpha _{H}}L_{H}^{1-\alpha _{H}}, \end{eqnarray*} where A is the technology residual or total factor productivity (TFP) in the economy, K is the aggregate capital stock, L is total employment, and 0 < α < 1 is the capital income share. Similarly, AH is TFP in the health care sector, KH and LH are capital and labor devoted to health care goods, and 0 < αH < 1 is the capital income share in the health care sector. Then, the evolution of the US medical care expenditure gap depicted in Figure 1 is given by the ratio $$\frac{\left(\frac{p_{H}^{ {\mathit US}}\times Y_{H}^{ {\mathit US}}}{p^{ {\mathit US}}\times Y^{ {\mathit US}}}\right)}{\left(\frac{p_{H}^{ {\mathit OECD}}\times Y_{H}^{ {\mathit OECD}}}{p^{ {\mathit OECD}}\times Y^{ {\mathit OECD}}}\right)},$$ (1) where |$p_{H}^{ {\mathit US}}$| and |$p_{H}^{{\mathit OECD}}$| are the prices in the health care sector and |$p^{ {\mathit US}}$| and |$p^{ {\mathit OECD}}$| are the prices in the rest of the economy in the United States and OECD, respectively. Growth accounting focuses on the contribution of observable production factors. Here, we want to assess the contribution of capital and labor to the US medical care expenditure gap as well as the size of the residual over our sample of countries. We start this exercise with the following identifying assumption over the growth rates of the unobservable factors: $$\left(g_{A,H}^{ {\mathit US}}-g_{A}^{ {\mathit US}}\right)=\left(g_{A,H}^{ {\mathit OECD}}-g_{A}^{ {\mathit OECD}}\right).$$ (2) Additionally, when we analyze the US medical care expenditure gap in this section we will be assuming that $$\left(g_{p,H}^{ {\mathit US}}-g_{p}^{ {\mathit US}}\right)=\left(g_{p,H}^{ {\mathit OECD}}-g_{p}^{ {\mathit OECD}}\right).$$ (3) If both assumptions hold, differential factor accumulation should suffice to explain the US medical care expenditure gap. The rest of the paper provides direct and indirect evidence that the main driver of the observed US medical care expenditure gap is a violation of the equal growth rates for relative prices in equation (3). More precisely, our analysis implies that |$\big(g_{p,H}^{ {\mathit US}}-g_{p}^{ {\mathit US}}\big)>\big(g_{p,H}^{ {\mathit OECD}}-g_{p}^{ {\mathit OECD}}\big)$|⁠. Under equation (3), the US medical care expenditure gap is simply defined as |$\big(Y_{H}^{ {\mathit US}}/Y^{ {\mathit US}}\big)/\big(Y_{H}^{ {\mathit OECD}}/Y^{ {\mathit OECD}}\big)$|⁠, which depends only on the accumulation of inputs in each economy and the technology residuals. The equal growth rate in the residuals [equation (2)] states that the differential increase in technological change in the health care sector between the United States and the OECD is the same as the differential increase in technological change for the aggregate economy between the United States and the OECD. We will later check if this is a good approximation for US data. Now, taking logarithms, differentiating with respect to time, and assuming equal shares in the United States and the OECD, we obtain the following expression for growth in the US medical care expenditure gap from the observable factors: \begin{eqnarray} &&\!\!\!\!{\alpha _{H} \left(g_{K,H}^{ {\mathit US}}-g_{K,H}^{ {\mathit OECD}}\right)+(1-\alpha _{H})\left(g_{L,H}^{ {\mathit US}}-g_{L,H}^{ {\mathit OECD}}\right)}\nonumber \\ &&\quad -\alpha \left(g_{K}^{ {\mathit US}}-g_{K}^{ {\mathit OECD}}\right)-(1-\alpha )\left(g_{L}^{ {\mathit US}}-g_{L}^{ {\mathit OECD}}\right). \end{eqnarray} (4) Calibration. Using OECD data we get a capital to GDP ratio K/Y between 2.0 and 2.5 for a subset of countries with available data. The observed investment to GDP ratio is around 0.20, which together with K/Y = 2.5 yields an annual depreciation rate (δ) of 0.08 in the steady state. After letting K/Y = 2.5 in 1970, we calculate the evolution of K for each country according to the standard law of motion Kt + 1 = It + (1 − δ)Kt. The initial KH is calculated differently since the share of investment in the health care sector is lower than the observed average share in the economy. We first estimate the average investment in health care over HCE (IH/YH) and call this variable φ. Then, we calculate the initial capital in the health care sector as KH, 1970 = (φ/δ)YH, 1970. Once this is estimated, we again apply the standard equation for the capital's law of motion to calculate KH, t for all dates t. Finally, the last parameter we need to calibrate is the capital share α. The real return to capital in the United States has been around 8%.6 Letting K/Y = 2.5 and a net capital return of 8% yields α = 0.4. In our sensitivity analysis we let α range between 0.25 and 0.40. In our benchmark calculation we set α = αH = 0.25. Arguably, the health care sector has a lower K/Y ratio that would lead to a lower capital share than the rest of the economy. Some authors, however, have stressed the externality effects of equipment investment (e.g., De Long and Summers 1992). Hence, the social rate of return may be higher than the private rate of return. Therefore, a higher value for αH could be justified by invoking some plausible externality effects and intangible investments. Results. We now simulate equation (4) under our benchmark calibration for the estimated capital stocks and OECD data on health care employment as a proxy for LH and total employment as a proxy for L. Figure 3 shows the US medical care expenditure gap implied by the model together with its empirical counterpart. We also report the pace of the US medical care capital gap \begin{equation*} \left[\left(K_{H}^{ {\mathit US}}/K^{ {\mathit US}}\right)/\left(K_{H}^{ {\mathit OECD}}/K^{ {\mathit OECD}}\right)\right] \end{equation*} and the US medical care labor gap \begin{equation*} \left[\left(L_{H}^{ {\mathit US}}/L^{ {\mathit US}}\right)/\left(L_{H}^{ {\mathit OECD}}/L^{ {\mathit OECD}}\right)\right]. \end{equation*} The figure shows a relative increase in the stock of capital in the United States and a relative increase in employment in the OECD. Figure 3. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States and OECD. Source: OECD Health Data (June 2014) and authors’ computations. Figure 3. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States and OECD. Source: OECD Health Data (June 2014) and authors’ computations. Observe that the simulated US medical care expenditure gap remains quite flat during the entire period with a slightly declining trend. The gap goes down by 15% for our simulation of equation (4) whereas it goes up by 30% in the data. Further, during the 1978–1990 transition period the gap goes down by 5% for our simulation whereas it goes up by 32% in the data. Sensitivity Analysis. We also consider changes in these baseline values. We let the capital share α range between 0.4 and 0.25, and we let αH differ from α. We let annual depreciation δ range between 0.06 and 0.08. Table 2 shows the percentage change in the simulated US medical care gap for the entire period in panel (a) and for the 1978–1990 transition period in panel (b). For the 1970–2007 period, under all pairs (δ, α), we observe a decrease in the simulated US medical care expenditure gap between 10% and 15%, as opposed to an increase of 30% in the data. Similar results are observed during the 1978–1990 transition period in panel (b). Therefore, the present growth accounting exercise based on value added determines that the US medical care expenditure gap should have decreased during the period 1970–2007 as well as during the 1978–1990 transition period.7 Table 2. Sensitivity analysis. (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 Source: OECD Health Data (June 2014) and authors’ computations. View Large Table 2. Sensitivity analysis. (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 (a) 1970–2007 period (b) 1978–1990 transition period Observed change in US medical expenditure gap: 30% Observed change in US medical expenditure gap: 32% δ α αH Predicted change in gap (%) δ α αH Predicted change in gap (%) 0.08 0.25 0.25 −15.0 0.08 0.25 0.25 −5.4 0.06 0.25 0.25 −15.0 0.06 0.25 0.25 5.9 0.08 0.4 0.25 −13.3 0.08 0.4 0.25 −4.7 0.06 0.4 0.25 −13.3 0.06 0.4 0.25 −5.2 0.08 0.4 0.4 −9.6 0.08 0.4 0.4 −17.0 0.06 0.4 0.4 −10.0 0.06 0.4 0.4 −2.5 Source: OECD Health Data (June 2014) and authors’ computations. View Large Implied Medical Expenditure Gaps by Country. To get a better understanding of the underlying forces driving the US medical expenditure gap over our sample of countries, we also report the relative contribution of factor accumulation to the medical sector for each country i as compared to the United States. More formally, Figure 4 plots the evolution of the medical expenditure gap as defined by the following equation: \begin{eqnarray} &&\!\!\!\!{\alpha _{H}\left(g_{K,H}^{i}-g_{K,H}^{ {\mathit US}}\right)+(1-\alpha _{H})\left(g_{L,H}^{i}-g_{L,H}^{ {\mathit US}}\right)}\nonumber\\ &&\quad -\alpha \left(g_{K}^{i}-g_{K}^{ {\mathit US}}\right) - (1-\alpha )\left(g_{L}^{i}-g_{L}^{ {\mathit US}}\right), \end{eqnarray} (5) where i stands for every other country in our restricted sample. Figure 4. View largeDownload slide A basic growth accounting exercise—implied medical expenditure gaps by country. Source: OECD Health Data (June 2014) and authors’ computations. Figure 4. View largeDownload slide A basic growth accounting exercise—implied medical expenditure gaps by country. Source: OECD Health Data (June 2014) and authors’ computations. First, the figure shows that most individual medical care expenditure gaps have remained relatively flat as we condition by US value added in the health care sector. The sharp decrease observed for Germany corresponds to the unification of the country after the fall of the Berlin Wall because of a break in the data. Note that this decrease has been tapering off over time. Second, there are two outliers: Canada and Spain. The share of HCE in Canada has been slightly above the OECD average but the contribution of value added appears remarkably low. The Canadian medical system has been widely criticized for being comparatively expensive, having long waiting periods for services, and having limited access to physicians and medical technology, which may justify the low contribution of health care value added in our exercise. At the other extreme, a high contribution of value added in the Spanish medical sector is observed at the beginning of our sample period. Spain presents the highest percentage increase in the share of HCE moving from 3.5% of GDP in 1970 to 8.4% in 2007. Spain underwent various reforms toward higher health care quality and universal insurance for all citizens and residents, as well as an impressive GDP growth at these early dates. 2.2.2. Growth Accounting Using Capital and Labor Data for the United States only As labor and capital data could be of lesser quality in our sample of OECD countries, we can simulate a variant of equation (4) for US data only. That is, we analyze the relative contribution of value added in the US health care sector over overall value added, and we condition for the growth of the share of HCE in the OECD countries. Formally, we consider the evolution of the US medical care expenditure gap (United States only) under the following growth components: $$\alpha _{H}g_{K,H}^{ {\mathit US}}+(1-\alpha _{H})g_{L,H}^{ {\mathit US}}-\alpha g_{K}^{ {\mathit US}}-(1-\alpha )g_{L}^{ {\mathit US}}-g_{Y,H}^{ {\mathit OECD}}+g_{Y}^{ {\mathit OECD}}.$$ (6) The US values for K, KH, α, αH, and δ come from the preceding exercise. Also, L and LH are taken from employment data. For the other OECD countries, we consider the growth in the overall share of HCE. Medical prices face particularly difficult measurement problems, and according to the Boskin Commission Report (1996) the US consumer price index does not take into account a 3% yearly increase in the quality of health care goods, whereas this quality bias was less than 1% for the general index. On these grounds, we may consider that technological progress may have been more pronounced in the medical sector. Hence, our implicit identifying assumption |$g_{A,H}^{ {\mathit US}}=g_{A}^{ {\mathit US}}$| underlying this exercise may not be supported by the data. Figure 5 displays the newly simulated US medical care expenditure gap under equation (6) together with the previously simulated gap under equation (4). The figure illustrates that for the benchmark calibration, there is almost no difference between the present and previous growth accounting exercises. This confirms that the US medical expenditure gap also declines when the contribution of value added is restricted to the US economy. We should note, however, that for some other combination of parameter values used in the sensitivity analysis of Table 2, the simulated US medical care expenditure gap (United States only) is significantly below that obtained from equation (4). Hence, some of these computations suggest that TFP in the US health care sector may have grown faster than in the aggregate economy. Accordingly, the underlying identifying assumption |$g_{A,H}^{ {\mathit US}}=g_{A}^{ {\mathit US}}$| may not be a good approximation of the evolution of TFP in the health care sector. Our general formulation in (4) appears to be more adequate as it conditions for both general TFP trends in the aggregate economy and the health care sector. Figure 5. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States only. Source: OECD Health Data (June 2014) and authors’ computations. Figure 5. View largeDownload slide A basic growth accounting exercise using labor and capital data for the United States only. Source: OECD Health Data (June 2014) and authors’ computations. Finally, using the previous values for the growth of labor and capital we compute the growth rate of value added in the health care sector over value added in the economy. Formally, for every date we compute \begin{eqnarray*} \alpha _{\mathit {H}}g_{\mathit{K,H}}^{\mathit {US}}+(1-\alpha _{\mathit {H}})g_{\mathit {L,H}}^{\mathit {US}}-\alpha g_{\mathit {K}}^{\mathit {US}}-(1-\alpha )g_{\mathit {L}}^{\mathit {US}}. \end{eqnarray*} Under our baseline parameters, for the period 1970–2007 the share of HCE grows by 45% in our simulation exercise, whereas it grows by 130% in the data. For the 1978–1990 transition period the simulated share of HCE grows by 17%, whereas it grows by 47% in the data. Therefore, it appears that value added cannot account for the patterns of growth in the share of HCE in the United States, and would require implausible growth in TFP in the health care sector. 2.3. Returns, Yields, and Markups of Publicly Traded Companies in The Health Care Sector The previous growth accounting exercise attests that factor accumulation cannot explain the increase in the US medical care expenditure gap. It follows that our identifying assumptions [equations (2) and (3)] may fail. That is, the pace of technological change and incremental inflation rates in the medical sector may be greater in the United States than in the average OECD country. Comin and Hobijn (2004 and 2010) document that technology spreads relatively quickly across countries with similar levels of development. They argue that the diffusion of new technology became increasingly fast after World War II. After considering financial data we conclude that increasing prices in the US health care sector cannot be associated with additional expenditures in resources. Moreover, Section 3 builds on a vast literature claiming that higher medical prices are not associated with better health outcomes either. Therefore, the sharp break in the US medical care expenditure gap observed during the 1978–1990 period appears to be difficult to justify on the grounds of slow diffusion of technology and increasing costs originating from better product quality in the United States. Several micro studies analyze the degree of competition and efficiency in the US medical care sector. For instance, Dunn and Shapiro (2011) claim that physicians’ market power may bias medical care prices and the quantity of health care services provided. Skinner et al. (2005) argue that Medicare spending appears to be highly inefficient: About 20% of Medicare expenditures do not provide any increase in survival rates or quality of life for the elderly population. Koijen, Philipson, and Uhlig (2016) call attention to a “medical innovation premium” of 4%–6% annually for equity returns of firms in the health care sector, and interpret this premium as compensating investors for government-induced profit risk. As discussed in the introductory section, we do not primarily address these efficiency issues from a micro perspective. Our main goal here is to focus on the large wedge between value added and HCE documented in our previous growth accounting exercise. As already pointed out, physicians did not particularly benefit from the observed HCE growth. But private companies may have been able to take advantage of monopoly power and existing frictions and regulations in the health care market. We consider stock returns, yields, and markups generated by publicly traded companies in the health care sector as compared to those in the aggregate economy. We find that the health care sector is associated with abnormal profits and markups. These financial measures of profitability grow much faster over the 1978–1990 transition period. We retrieve company level data on Total Revenues, and Total Variable Costs from COMPUSTAT, that is, entries REVT and COGS in COMPUSTAT, for the period between January 1970 and December 2007. We retrieve data on returns, prices, and shares outstanding on publicly traded firms in the United States from the Center for Research in Security Prices (CRSP), that is, entries RET, PRC and SHROUT in CRSP. We select our health care companies from the Standard Industrial Classification (SIC) codes as in Kenneth French's webpage within the set of industrial portfolios.8 This is arguably the most used portfolio database in the finance literature. To have an idea of the relevance of our sample, we calculate the ratio of total yearly sales for the health care companies in our portfolio over HCE. This ratio increases from 20% in 1970 to 30% in 2007; hence, our sample appears to be fairly representative. Stock Market Return. At every date, the health care portfolio return is calculated as an average of each company's return (including dividends) weighted by its relative market capitalization in the health care portfolio.9 We also use a market weighted portfolio that includes almost all publicly available companies in the CRSP database.10 Average returns in the health care sector portfolio have soared since the late 1970s. The weighted average yearly return for the 1970–2007 period for this portfolio is about 20.17% whereas that of the market portfolio is about 12.34%. In the finance literature it is customary to consider the metric excess return adjusted by its standard deviation. This metric is known as the Sharpe Ratio and provides a measure of the reward-to-risk. The risk-free rate (the one-month T-bill in our case) is subtracted from the portfolio return to compute the excess return. Under certain regularity conditions, all well diversified portfolios should have the same Sharpe Ratio. For our health portfolio, the Sharpe Ratio is 0.65 for the period 1970–2007 whereas it is 0.43 for the market portfolio. Interestingly, the Sharpe Ratio of the health care portfolio is 1.05 for the 1978–1990 period, whereas it is 0.04 and 0.72 for the 1970–1977 period and the 1991–2007 period, respectively. Yield. We define the yield of a company as the company's gross profit divided by its market capitalization value at a certain date. Gross profit is the difference between total revenue and total variable cost. We focus on gross profit because this is usually considered a clean accounting measure of profitability (cf., Novy-Marx 2013). The aggregate yield is then obtained as a weighted average of individual yields, where the weights are company total revenues over the aggregate sum of total revenues. For our sample period, we get an average aggregate yield of about 36% for the companies in the health care sector, and 70% for all other companies. A lower average yield may mean that companies in the health sector face lower risk and higher expectations of profit growth. In some research in progress, we have documented that lower yields are supported by sizable and stable product markups. Price Markup. We define the price markup of a company as the ratio between total revenue and total variable cost. The aggregate price markup is then obtained as a weighted average of company markups where the weights are company total revenues over aggregate total revenues. For our sample period 1970–2007, we get an average aggregate price markup in the health care sector around 193%, whereas we get an aggregate price markup for all the other companies around 55%. The average yearly growth rate of HCE over nominal GDP was 3.2% during the 1978–1990 period, whereas it was 2.4% in the 1970–1977 period, and 1.6% in the 1991–2007 period. Table 3 compares this incremental growth rate against the following aggregate measures: (i) Annual incremental return of the health care sector portfolio as compared to the remaining market portfolio, (ii) Annual incremental growth rate of the yield of the health care sector portfolio as compared to the remaining market portfolio, (iii) Annual incremental growth rate of the aggregate markup of the health care sector portfolio as compared to the remaining market portfolio, and (iv) Annual incremental growth rate of value added in the health care sector as compared to the aggregate economy. Table 3. Returns, yields and markups of publicly traded companies in the health care sector relative to the rest of the economy. Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Source: CRSP, COMPUSTAT, Kenneth French website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/), and authors’ computations. View Large Table 3. Returns, yields and markups of publicly traded companies in the health care sector relative to the rest of the economy. Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Growth rates Health care sector returns minus economy returns (%) Health care sector markup over economy markup (%) Health care sector yield over economy yield (%) Health care sector value added over economy value added (%) HCE over GDP (%) Average 1970–1977 period 0.6 −0.5 1.7 0.9 2.4 Average 1978–1990 period 10.7 3.0 9.5 1.4 3.2 Average 1991–2007 period 9.1 1.8 2.7 0.9 1.6 Source: CRSP, COMPUSTAT, Kenneth French website (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/), and authors’ computations. View Large Remarkably, for the 1978–1990 and 1991–2007 periods the incremental annual markup growth rate in the health care sector is roughly the same as the incremental annual growth in the share of HCE, whereas there is much less variability in incremental growth of value added. Along these same lines, the incremental yield growth in the health care sector gets much higher during the 1978–1990 transition period. Hence, episodes of high growth of HCE in the United States are associated with higher markups and higher growth rates of profitability. Excess returns also get much higher during the transition period, and stay relatively high afterward even though there is a substantial decrease in the excess growth rate of the yield. This slow adjustment of stock market values typically occurs when high profits are initially perceived as temporary or non-sustainable. To have a better sense for these trends, Figure 6 depicts the evolution of Health Care Sector Markup over Economy Markup, Health Care Sector Gross Profit over Economy Gross Profit, Health Care Sector Value Added over Economy Value Added, and HCE over GDP between January 1970 and December 2007. Panel (a) displays variables in levels, and panel (b) displays 5-year moving averages of corresponding growth rates. For convenience, all levels values are normalized to 100 at the start of the transition period in 1978, and the two vertical bars bracket the 1978–1990 transition period. The following facts are worth noticing from these figures. First, Health Care Sector Markup over Economy Markup mimics HCE over GDP whereas the incremental value added presents milder growth rates. Fluctuations in Health Care Sector Markup over Economy Markup slightly lead but are fairly contemporaneous with those of HCE over GDP. Increases in these two aggregate measures stimulate factor accumulation but this response is not very pronounced. Second, fluctuations in the growth rate of Health Care Sector Gross Profit over Economy Gross Profit move in tandem with that of Health Care Sector Markup over Economy Markup, but tend to have smaller values. In fact, after 1994, relative markups have an upward trend, whereas relative gross profits remain relatively flat. This suggests some kind of cost-containment as discussed in footnote 14. Overall, the present figure is consistent with the notion that growth in Health Care Sector Markup over Economy Markup is strongly linked to growth in HCE over GDP; markups seem a major driver of HCE. Finally, Figure 7 shows the evolution of the Herfindahl–Hirschman Index (HHI) that measures industry concentration; that is, a higher value of the index implies higher industry concentration for the publicly traded companies in the health care sector. Besides this index, panel (a) shows the number of publicly traded companies in the health care sector, whereas panel (b) shows Health Care Sector Markup over Economy Markup. From panel (a) we can observe that even though the number of publicly traded companies was steadily going up since the beginning of the 1978–1990 transition period, the HHI continues to increase until reaching its maximum value in the year 1989. (As discussed in Figure 6, increases in markups may stimulate further economic activity that in this case may translate into entry of firms in the medical sector.) Panel (b) shows that both the HHI and Health Care Sector Markup over Economy Markup went up during the transition period, but afterward these two measures diverge. Hence, industry concentration can only offer a partial account for the evolution of markups. Further research may be necessary (possibly at the micro level) to understand the nature of market frictions that determine the evolution of product markups. Usually, markups are tied down to changes in demand elasticities and in the regulatory environment. Figure 6. View largeDownload slide Health care sector markup over economy markup, health care sector gross profit over economy gross profit, health care sector value added over economy value added, and HCE over GDP. Source: CRSP, COMPUSTAT and authors’ computations. Figure 6. View largeDownload slide Health care sector markup over economy markup, health care sector gross profit over economy gross profit, health care sector value added over economy value added, and HCE over GDP. Source: CRSP, COMPUSTAT and authors’ computations. Figure 7. View largeDownload slide Herfindahl–Hirschman index for the health care sector. Source: CRSP, COMPUSTAT, and authors’ computations. Figure 7. View largeDownload slide Herfindahl–Hirschman index for the health care sector. Source: CRSP, COMPUSTAT, and authors’ computations. In summary, strong positive growth in the ratio of HCE over GDP can be fully backed up by corresponding growth in gross profits and markups. The connection between value added and financial variables appears to be much more tenuous, especially over the 1978–1990 transition period. For a realistic pace of technology diffusion, abnormal profits and markups help us understand that prices in the medical sector may not reflect the extra costs associated with high product quality. 3. Quality and Price Effects To evaluate the wedge between value added and HCE, we now take a different approach and adjust HCE by aggregate prices. Of course, prices may be picking up quality effects but our previous evidence on the evolution of markups and gross profits shows a price disconnect from average costs. We first document some indicators of US health care quality within the OECD countries. After filtering out price effects, we also show that US real health care consumption moves intandem with real GDP, as well as with real health care consumption in the other OECD countries. We then simulate a product quality residual over our sample of OECD countries. This approach does not change our basic findings in the previous section, but it is subject to new measurement issues as it is usually hard to disentangle price and quality effects. 3.1. Quality Our growth accounting exercises rest on a certain identifying assumption for non-observable TFP growth; see equation (2). This assumption appears to be adequate for the 1978–1990 transition period. As already discussed, it is unlikely that TFP changes can have such a sizable differential impact across economic sectors and countries over such a limited time span. We next provide some supporting evidence for our identifying assumption by studying the evolution of product quality at the aggregate level. Various international institutions such as the OECD and the World Health Organization periodically report cross-country rankings on various dimensions of health status, heath care activities, quality of care, and access to care. Generally, the United States lags behind many other OECD countries. In fact, when conditioning for income and health care expenditures the United States appears mostly as an outlier. Figure 8 shows the evolution of some of these quality proxies for health status, which will be expressed as gaps: the corresponding US value over the average value of our sample of OECD countries (excluding the United States), letting 100 be the initial value for our starting year, 1970. We consider the US gap for infant mortality, neonatal mortality, perinatal mortality, and life expectancy at birth.11 We can observe that although the life expectancy gap has remained quite stable, the gap of every other variable has increased—meaning that in fact the quality gap has evolved against the United States. These outcomes may partially be explained by other measures of health care activities and access to care such as the relatively large size of the uninsured population in the United States: Around 12% of the population has permanently been uninsured over the last three decades. Figure 8. View largeDownload slide Quality indicators over time. Source: OECD health data (June 2014). Figure 8. View largeDownload slide Quality indicators over time. Source: OECD health data (June 2014). For narrower quality indicators we have not been able to gather data over the whole sample period. Table 4 reports on the following quality measures for the year 2003: Life expectancy, breast cancer five-year observed and relative survival rates for females, cervical cancer five-year observed and relative survival rates for females, and colorectal cancer five-year observed survival rates for males and females.12 For each of these indicators the coefficient of variation (i.e., the standard deviation over the mean) is usually very small—attesting that there is very little variability in all the quality proxies across the OECD countries.13 The United States has the lowest life expectancy, the maximum survival rates for breast cancer, and is below the maximum for cervical and colorectal cancers. Overall, the assumption of similar health care quality across OECD countries seems to be supported by the data. Table 4. Quality measures in 2003. 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) Note: If there is no data for a country for a specific year but there is for an immediate previous or following year we use that number. View Large Table 4. Quality measures in 2003. 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) 2003 Variable Obs. Mean US/mean US/max Std. Dev. Std. Dev./Mean Min Max Life expectancy 11 81.82 0.97 0.93 1.97 0.024 79.60 85.30 Breast cancer five-year 18 72.66 1.07 1.00 4.28 0.059 61.11 77.72 Observed survival rate (female) Breast cancer five-year 18 84.02 1.06 1.00 3.12 0.037 78.57 89.30 Relative survival rate (female) Cervical cancer five-year 19 61.81 0.97 0.82 5.34 0.086 50.92 73.35 Observed survival rate (female) Cervical cancer five-year 18 65.81 0.98 0.84 5.03 0.077 57.28 76.69 Relative survival rate (female) Colorectal cancer five-year 18 48.25 1.11 0.89 5.52 0.114 38.28 59.80 Observed survival rate (male) Colorectal cancer five-year 19 52.95 1.06 0.96 4.24 0.080 42.79 58.23 Observed survival rate (female) Note: If there is no data for a country for a specific year but there is for an immediate previous or following year we use that number. View Large Certainly, all the previous evidence should be merely considered as informative of the health care quality level across countries. Besides, life expectancy can be affected by genetic differences and dietary habits. Similarly, survival rates are not necessarily a good proxy for the quality of life after a surgical procedure. Some indicators may proxy longevity rather than quality. Drösler, Romano, and Wei (2009) report findings on various new quality indicators regarding patient safety across OECD countries. Again, the United States always ranks about average in every category. 3.2. The US Medical Care Price Gap and the Stability of Real HCE over Total Consumption in the United States Besides product quality, health care prices could be driving the US medical expenditure gap. A measure of the consumer price index—medical care (CPIMC) is available for all countries in our sample. These prices along with their data sources are formally explained in the Appendix. Measurement issues concerning the CPIMC will be deferred to the next section. We define the US medical care price gap as the US ratio CPIMC/CPI over the average ratio CPIMC/CPI of the other OECD countries in our sample. Figure 9 plots the US medical care expenditure gap against the price gap; both ratios are normalized to 1 in 1977. Observe that the US medical care price gap mimics quite well the US medical care expenditure gap for the 1978–1990 period in which both increased by about 35%. Then, there is a mild disconnect: The relative price of medical care appears to increase faster than HCE. More precisely, between 1993 and 2007 the US medical care price gap goes from 1.35 to 1.50, whereas the US medical care expenditure gap appears quite flat.14 Indeed, it follows from these definitions that if the US medical expenditure gap grows less than the price gap, then the ratio of real HCE over real GDP will go down in the United States as compared with the other OECD countries. As a matter of fact, the 1990s witnessed a shift to managed care leading to a relative decline of hospital care [Figure 2(b)], and to a decrease of indemnity health insurance plans in favor of HMOs, PPOs, and POSs. Cutler, McClellan, and Newhouse (2000) suggest that for certain treatments and procedures managed care organizations like HMOs may have lowered costs by about 30%–40% as a result of price declines of medical services and treatment intensities. Figure 9. View largeDownload slide The US medical care expenditure gap versus the price gap 1970–2007. Source: Health Care Expenditures are from the OECD Health Data (June 2014). Health Care Price data are taken from each country's official statistics and authors’ computations. Figure 9. View largeDownload slide The US medical care expenditure gap versus the price gap 1970–2007. Source: Health Care Expenditures are from the OECD Health Data (June 2014). Health Care Price data are taken from each country's official statistics and authors’ computations. To go further into the evolution of real consumption in health care in the United States, in Figure 10 we adjust HCE under various price measures. For the 1970–2007 sample period, HCE per capita at constant consumption prices in the United States has increased by 300%. For the same time period, HCE per capita at constant medical care prices (real HCE) in the United States has increased less than 100%. Moreover, there is no significant growth when real HCE per capita is adjusted by real total consumption (TC), but we observe some slight growth in the share of value added. Hence, Figure 10 reports: (i) HCE per capita at constant TC prices; (ii) Real HCE per capita: HCE per capita at constant medical care prices; (iii) Real HCE over real TC; and (iv) US value added in the health care sector over total value added in the economy. Figure 10. View largeDownload slide The evolution of US health care expenditures 1970–2007. Source: US Department of Health and Human Services—centers for Medicare and Medicaid services and the Bureau of Labor Statistics. Figure 10. View largeDownload slide The evolution of US health care expenditures 1970–2007. Source: US Department of Health and Human Services—centers for Medicare and Medicaid services and the Bureau of Labor Statistics. 3.3. A Cross-Country Analysis of The US Health Care Expenditure Gap as Implied by Aggregate Prices We now present a more systematic analysis of real health care consumption across countries. In a separate appendix we detail a simple model for health care demand embedding several sources of technological progress and health care quality. The main purpose of our exercise is to filter out price effects and common trends in technology and product quality under the simplifying assumption of a constant elasticity of substitution between the health care good and the remaining aggregate good in the economy. More precisely, let us assume that qh is health care expenditures (HCE), c is total consumption (TC), q is the relative price of medical care, and |$\widehat {A}$| is a residual term picking up product quality. Then, let $$\frac{\partial {\rm ln}(h_{t}/c_{t})}{\partial {\rm ln}(q_{t})}=\frac{1}{\rho -1},$$ (7) where −∞ < ρ < 1. After appending the residual term, we get: |$\text{ln}(q_{t}h_{t}/c_{t})={\rm ln}(\widehat {A}_{t})+(\rho /(\rho -1))\text{ln}(q_{t})$|⁠. This is therefore our basic equation to track the evolution of the US medical expenditure gap: $$q_{t}h_{t}/c_{t}=\widehat {A}_{t}q_{t}^{\frac{\rho }{\rho -1}}.$$ (8) In our data analysis, qh/c corresponds to the share of health care expenditures in total consumption for the United States and OECD countries, whereas q picks the evolution of the ratio of the medical care price index CPIMC over the CPI for each country. Under the identifying assumption of equal growth rates for the residuals, $$g_{\widehat{A}}^{ {\mathit US}}=g_{\widehat{A}}^{ {\mathit OECD}},$$ (9) we have the following expression for growth in the US medical care expenditure gap based on incremental inflation in the medical sector: $$\left(\frac{\rho }{\rho -1}\right)\left(g_{q}^{ {\mathit US}}-g_{q}^{ {\mathit OECD}}\right).$$ (10) Ringel et al. (2000) report estimates for the price-elasticity of demand for health care consumption of around −0.17. Since the health care good is a relatively small fraction of total consumption, we then get that ρ is approximately equal to −5. Studies consistently find inelastic demands when the share paid by the consumers is relatively low. These point estimates for the elasticity are around our calibrated value (cf., op. cit.). It is worth noticing that fractional payment policies for medical expenses (e.g., some types of coinsurance and out-of-pocket expenditures) will not affect the elasticity of demand. For comparison purposes, we also report computations for ρ = −3 and for ρ = −1, which imply elasticities of demand for health care consumption of around −0.25 and −0.50, respectively. Figure 11 reproduces these computations for the US medical care expenditure gap as implied by (10) as well as the empirical counterpart or actual gap. Observe that for ρ = −5 the simulated US medical care expenditure gap would account for about 107% of the actual gap. That is, over the sample period the implied US health care expenditure gap goes from 1.14 to 1.64 whereas in the data it goes from 1.14 to 1.53. For ρ = −3 the simulated residual would account for about 103% of the actual one, and for ρ = −1 the simulated residual would account for about 92% of the actual one. Because of the evolution of the medical expenditure gap in the last part of the sample, the breaking point is ρ = −2, where the simulated residual growth between 1970 and 2007 is almost the same as the actual one. This is consistent with a price elasticity of demand approximately equal to −0.33. For more inelastic demands, the observable growth in relative prices will overestimate the US medical expenditure gap. Figure 11. View largeDownload slide The US health care expenditure gap as implied by the relative price. Source: Health care expenditures are from the OECD health data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. Figure 11. View largeDownload slide The US health care expenditure gap as implied by the relative price. Source: Health care expenditures are from the OECD health data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. 3.4. Medical Expenditure Gaps by Country To have a better sense of these computations, we now replicate the US medical care expenditure gap for every single country i using ρ = −2. In fact, to adjust directly for actual growth in each individual gap, we consider the evolution of the following growth components for each country i: $$\left(\frac{\rho }{\rho -1}\right) \left(g_{q}^{ {\mathit US}}-g_{q}^{i}\right)-g_{HC}^{ {\mathit US}}+g_{HC}^{i},$$ (11) where |$g_{HC}^{i}$| is the is the growth rate of HCE over Total Consumption for country i. We can interpret expression (11) as the actual evolution of the ratios between the residuals |$\widehat {A}{}_{t}^{i}$| and |$\widehat {A}{}_{t}^{ {\mathit US}}$|⁠. Figure 12 shows the evolution of these ratios for the residuals for each country i in the sample. Except for France and Spain, all these ratios for the individual residuals are fairly stable. Hence, the United States displays a residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| quite in line with the corresponding average of the other OECD countries. Roughly speaking, there are not noticeable differences in the US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| between the 1978 and 1990 transition period and the rest of the sample. For reasons already discussed, the US residual declines after 1992, which suggests either a relative loss in US real health care consumption or an upward bias in the US price index. Figure 12. View largeDownload slide Evolution of the residuals by country as compared to the United States. Source: Health care expenditures are from the OECD Health Data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. Figure 12. View largeDownload slide Evolution of the residuals by country as compared to the United States. Source: Health care expenditures are from the OECD Health Data (June 2014). Health care price data are taken from each country's official statistics and authors’ computations. The French National Health Care System initial program was created in 1928 but was not comprehensive (Rodwin and Sandier 1993). France expanded its public health insurance programs at various stages, and it became universal for all its citizens and residents in 2000 (Rodwin 2003). Spain shows sharp trend breaks in 1986 and 2003 corresponding to two major medical reforms. The General Health Law of 1986 recognized the right to health care services for all citizens and foreign residents in Spain, and the Law of Cohesion and Quality of 2003 modernized and broadened the scope of the previous law. In conclusion, three important facts emerge from this study: (i) The US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| behaves quite similarly throughout all three time subperiods considered in our sample. This seems to be in accord with the aforementioned stability of real health care expenditures over real GDP observed in the US economy and the growth patterns of labor and capital in the health care sector. (ii) The pace of the US residual |$\widehat {A}{}_{t}^{ {\mathit US}}$| does not differ much from the average pace of the OECD residual |$\widehat {A}{}_{t}^{ {\mathit OECD}}$|⁠. And (iii) sharp breaks in the residuals are usually associated with medical reforms and regulations. Therefore, this growth accounting study appears to indicate that unobservables affecting the residuals like technology and product quality change would not be a major driving force for the pronounced increase in the US medical care expenditures gap over the 1978–1990 transition period. 4. Discussion of US Medical Care Prices As shown in Figure 9, aggregate prices seem to track well the evolution of the US medical expenditure gap. In this section, we discuss some of the limitations of aggregate health care prices. We also compare the evolution of our US health care price index with independent measures of cost and sources of inflation in the medical sector. As illustrated in Figure 13, health care prices in the United States are around 60%–80% higher than in other OECD countries. These differences seem quite commensurate with the size of the US medical care expenditure gap. Figure 13. View largeDownload slide International health care prices over US health care prices 2011. Source: International Federation of Health Plans. 2011 Comparative price report: medical and hospital fees by country. http://ifhp.com/documents/IFHP_PriceReport2010ComparativePriceReport29112010.pdf. Figure 13. View largeDownload slide International health care prices over US health care prices 2011. Source: International Federation of Health Plans. 2011 Comparative price report: medical and hospital fees by country. http://ifhp.com/documents/IFHP_PriceReport2010ComparativePriceReport29112010.pdf. The CPI-medical care (CPIMC) is published by the US Bureau of Labor Statistics (BLS) since 1935. The PCE-health care (PCEHC) is another major price index with data from 1929 and published since 1994 by the US Bureau of Economic Analysis (BEA). The evolution of these two measures is quite similar over the 1978–1990 transition period. As stated in Fixler and Jaditz (2002), there are three main differences between the BLS's CPIMC and the BEA's PCEHC: “First, the two indexes use different formulas. The CPI is a Laspeyres index, whereas the BEA product is a Fisher Ideal index. Second, the two indexes have different underlying concepts. The BLS product measures the prices paid by (urban) consumers, whereas the BEA product measures the prices of final consumption goods, wherever they are purchased. Finally, differences in how the detailed components are implemented lead to differences in how prices are measured and the weights attached to specific series.” After adjusting for aggregate inflation in the economy, Hospital and Related Services has experienced the highest overall growth rate (3.76% annually for the CPIMC over CPI since 1978). Prescription Drugs shows the lowest growth rate before the transition period starts, and it grows at 3% annually for the CPIMC over CPI after 1978. Professional Services appreciates at a slower pace (1.35% annually for the CPIMC over CPI) since 1978. After 1992, for all these categories the PCEHC grows at a slower pace than the CPIMC. In summary, the highest increases relative to average inflation are observed in Hospital and Related Services and within the 1978–1990 transition period. 4.1. Measurement Issues Inflation in the health care sector has been greater than average inflation in the economy in almost every year. As explained in the Boskin Commission Report (1996), the CPIMC and its various subcategories had an estimated annual upward bias of 3% between 1970 and 1995 when adjusting for quality. During the same time period, the overall CPI has an estimated annual upward bias of 0.6%, leading to an estimated yearly bias in the growth rate of the CPIMC over CPI of 2.4%. Technology may impact HCE in various ways. To control for the effects of technological progress our macro-level difference-in-difference approach conditions upon a group of OECD countries with similar health care quality and comparable methodologies in data collection. For many countries public provision of health care is the norm. Accordingly, medical care expenditures are estimated as in any other public good—including the cost of intermediate inputs and labor but not capital rents. Medical care prices in some countries do not depart much from the overall consumer price index. We should also point out that average inflation in the OECD was much greater than in the United States in the 1970s, but inflation has been much lower and more stable in the remaining part of the sample for all these countries. In fact, for both the United States and the OECD countries the CPIMC over the CPI did not grow at all in the 1970s. The US relative price of medical care started to grow in the 1978–1990 transition period. Therefore, the US medical care price gap is not influenced by unequal cross-country inflation trends; when inflation trends diverge between the United States and the OECD, there are no deviations in the relative price of medical care. A major component of the CPIMC is Hospital and Related Services. There is evidence that this component is upward biased because list prices are affected by discounts. Indeed, this bias has been the norm under the expansion of HMOs and PPOs since managed care organizations are able to bargain on hospital's list prices (Dranove et al. 2008). This possible upward bias may have been rather small during the 1978–1990 transition period since HMOs and PPOs just became important in the late 1980s. In 1997 the BLS started to implement corrective measures to the CPIMC-hospital and related services to remove this sampling error (Cardenas 1996). In a recent study, Koechlin, Lorenzoni, and Schreyer (2010) circumvent this upward bias by comparing what they call hospital quasi-prices over several OECD countries. These quasi-prices are defined as negotiated or administrative prices or tariffs on various hospital services items. Still, these authors find that US hospital services are 60% above the average cost of these other countries. Berndt et al. (2000) discuss another source of overstatement of the CPIMC-hospital and related services component: The switch from inpatient procedures to outpatient procedures—mainly because of the cost-containment efforts supported by managed care organizations. The switch from inpatient to outpatient treatments may have increased the average complexity and cost of medical procedures for both types. Indeed, very complex procedures were still left as inpatient care, and those shifted to outpatient care were more complex than the average procedure of this category. Since outpatient and inpatient procedures enter separately and with fixed weights in the index calculation, the possible cost reduction of the shift to outpatient care may have not been reflected in the CPIMC-hospital and related services. This upward bias may have been exacerbated after 1987 with the expansion of managed care organizations. The BLS did not take corrective measures until 1998 (Berndt et al. 2000). Again, this bias should be rather small for the 1978–1990 transition period. A pioneering work by Griliches and Cockburn (1994) found that US pharmaceutical price indexes were upward biased because of a shift from branded to generic drugs. Danzon and Chao (2000) found that the large differences between US prices and those of other countries reported in many studies might be smaller than previously thought. After 1995, the BLS implemented corrective measures for the construction of its pharmaceutical price indexes, including generic drugs (Berndt et al. 2000). It seems that neglecting generic drugs in the price index should have a very limited impact during the 1978–1990 transition period. The market for generic drugs expanded after the Waxman–Hatch Act of 1984 (Berndt and Aitken 2010). The generic drugs’ share of total dispensed prescription drugs in the US retail market represented just 18% in 1984. It increased to 36% by 1994, then 56% by 2004, and 74% by 2009. At the same time, the revenue share of generic drugs has been relatively small. For example, in 2004 generic drugs were 56% of dispensed drugs but its revenue share was only 17%. The revenue share of generic drugs rose to 24% by 2009 (Berndt and Aitken 2010). Hence, during the 1978–1990 transition period the upward bias would seem rather limited. Of course, generic drugs are also dispensed in every OECD country. Danzon and Furukawa (2011) found that generic drugs are more expensive in the United States than in most of the OECD countries right after the branded drug's patent expires (around 75% of the branded drug price). Three years after patent expiration, prices of generic drugs decline more sharply in the United States, and after eight years of patent expiration generic drugs are cheaper in the United States than in all other OECD countries except the United Kingdom. Again, these patterns may further reduce the possible upward CPIMC bias in the 1978–1990 transition period. 4.2. US Medical Prices, Markups, and Health Insurance Premiums We now study the evolution of independent price measures beyond the CPI and other sources of inflation supporting high prices in the US medical care sector. Price increases for medical products and services have come along with hikes in markups and health insurance premiums. Table 5 reports the evolution of several cost measures in the health care sector. Table 5. Independent price measures beyond the CPI and other sources of inflation in the US medical sector. 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 Source: US Bureau of Labor Statistics, US Bureau of Economic Analysis, US Department of Health and Human Services – Centers for Medicare and Medicaid Services, The 2003 GAO Report on Medical Malpractice Insurance, Goldman and McGlynn (2005), and Cohen et al. (2009). Physicians’ compensation data come from authors’ calculations using data from the American Medical Association and the Medical Group Management Association. View Large Table 5. Independent price measures beyond the CPI and other sources of inflation in the US medical sector. 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 1975 1980 1985 1990 1995 2000 2005 Markup in the health care sector over markup in the economy 97 100 107 138 124 156 168 Average physician compensation over CPI N/A 100 103 124 124 125 139 Cost per inpatient day over CPI 61 100 154 192 245 238 N/A Insurance-premium malpractice index over CPI 110 100 254 287 278 291 552 Private health insurance premium per enrollee over CPI 73 100 137 206 241 284 377 Cost of medicare per enrollee over CPI 77 100 136 156 204 208 258 HCE per capita over CPI 84 100 127 162 187 208 256 CPIMC over CPI 99 100 115 137 155 162 178 Source: US Bureau of Labor Statistics, US Bureau of Economic Analysis, US Department of Health and Human Services – Centers for Medicare and Medicaid Services, The 2003 GAO Report on Medical Malpractice Insurance, Goldman and McGlynn (2005), and Cohen et al. (2009). Physicians’ compensation data come from authors’ calculations using data from the American Medical Association and the Medical Group Management Association. View Large 4.2.1. Aggregate Markup in the Health Care Sector over Aggregate Markup in the Economy A weighted average of Total Revenues over Total Variable Cost of publicly traded companies in the health care sector relative to the remaining publicly traded companies in the US stock market. (Construction of this variable is explained in Section 2.3; source: COMPUSTAT.) 4.2.2. Average Physician Compensation over CPI A measure of the evolution of the average physician's salary in the United States (sources: The American Medical Association for the years 1978–2000 and the Medical Group Management Association for the years 1994–2005). 4.2.3. Cost per Inpatient Day over CPI An independent measure of health care inflation reported by the American Hospital Association (sources: Goldman and McGlynn 2005, and the BLS). 4.2.4. Insurance-Premium Malpractice Index over CPI A measure of malpractice insurance costs for the providers of medical services. It is the average change in the cost of malpractice insurance for three physicians’ specialties: General Medicine, General Surgery, and Obstetrics/Gynecology. Data from 1976 to 1986 are from Danzon (1991). Data from 1987 to 1990 are from Harrington, Danzon, and Epstein (2008). Data from 1991 to 2007 are from authors’ computations from insurance costs for the specialties (Medical Liability Monitor Reports) over 51 states. 4.2.5. Private Health Insurance Premium per Enrollee over CPI The cost of health insurance for a privately insured person (sources: Cohen et al. 2009, the US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.6. Cost of Medicare per Enrollee over CPI Government's expenses per publicly insured elderly person (sources: The US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.7. HCE per Capita over CPI Real HCE per US inhabitant (sources: US Department of Health and Human Services – Centers for Medicare and Medicaid Services, and the BLS). 4.2.8. CPIMC over CPI A measure of excess inflation in the US medical care sector with respect to the average inflation in the US economy (source: The BLS). Observe from this table that Health Care Sector Markup over Economy Markup moves remarkably close to the CPIMC over CPI. The CPIMC displays a modest growth in comparison with other measures of medical costs, except for Average Physician Compensation that grows much less. These price measures grow much faster over the 1980–1990 period. The Cost per Inpatient Day over the CPI shows a steep cumulative increase of about 70% over the 1978–1990 transition period, whereas the CPIMC over CPI presents a cumulative increase of around 35% over the same period. Malpractice insurance premiums almost triple during the 1978–1990 transition period. The health economics literature (e.g., Harrington et al. 2008) usually refers to the 1980s as the “crisis” of medical malpractice because of the frequency and severity of claims and the dramatic increase in the cost of malpractice insurance. A main argument against using malpractice as an explanatory variable is that these costs are rather low: About 1.25% of HCE expenditures.15 Private Health Insurance per Enrollee has grown faster than HCE per Capita over the entire sample period. This increment came along with decreasing out-of-pocket expenditures—albeit out-of-pocket expenditures have still grown faster than nominal income per capita. Private Health Insurance per Enrollee over CPI more than doubles over the 1978–1990 transition period. Changes in Private Health Insurance per Enrollee over CPI are contemporaneously correlated with changes in CPIMC over CPI (a correlation coefficient of 0.43). The Cost of Medicare per Enrollee and HCE per Capita have moved together over the sample period. Hence, costs in the private and public sectors may be driven by the same set of common factors. Since the distribution of the population over the categories of uninsured, privately, and publicly insured remained quite stable over the last three decades, it appears that aging and the size of the elderly population cannot possibly account for the observed increase in the US medical care expenditure gap over our sample period. To sum up, except for aggregate markups and average physician compensation, related measures of medical care costs show much steeper increases than the CPIMC over both the entire sample period and the 1978–1990 transition period. On these grounds, the CPIMC seems a conservative measure of health care costs. We should remark that these measures of medical care costs have been obtained from unrelated data sources. 4.3. Microeconomic Evidence of International Health Care Prices Several studies and international institutions have reported marked cross-country differences in health care prices. Figure 13 presents costs of several health care items in five OECD countries as of year 2011. These costs are reported as fractions of the corresponding US cost. Switzerland exhibits the second highest costs: Around 65% of US dollar costs. For the remaining countries in the sample, the unitary prices observed are around one fifth of US costs. In these comparative studies it becomes quite difficult to control for quality across countries. There is, however, a large international market for prescription drugs, and the cross-country variability in wholesale drug prices has been well documented. An early study by Jacoby and Hefner (1971) reported prices for twenty drugs in nine countries. The study confirms a great variation from country to country for a single product by the same manufacturer. Some on-patent drugs were even three times cheaper outside the United States. Using 121 drugs to compare US prices with those of Canada and 76 drugs to compare with the United Kingdom, two GAO reports (GAO 1992, 1994) found that manufacturers prices in the United States were on average 32% higher than in Canada and 60% higher than in the United Kingdom. Later research has expanded the range of sample products to provide accurate price measures. Danzon and Furukawa (2003) find that manufacturer drug prices in eight representative countries are usually between 20% and 40% lower than in the United States. Although on-patent brands may be almost 50% cheaper in some of these countries, generic drugs are usually cheaper right after patent expiration, but later on they become relatively more expensive than in the United States (Danzon and Furukawa 2011). A related study by the US Department of Commerce (2004) on patented prescription drugs reports price indexes that could be 50% lower than their US counterparts (op. cit., p. 38). Hence, even for tradable products of a given quality, medical prices in the United States are much higher. 5. Concluding Remarks In this paper we present a macro-level difference-in-difference approach to study the evolution of US health care expenditures. Since 1980 the United States features the highest health care expenditure share of all the OECD countries. With the approaching retirement of the baby-boom generation, it is feared that the US medical care expenditure gap may take up a new expansionary turn. Thus, managing health care expenditure growth has become a topic of national concern, and a tall order for balanced economic growth. To guide this discussion, we examine a sample of eleven OECD countries with similar income and quality data over the sample period 1970–2007. We define the US medical care expenditure gap as the US medical expenditure share in nominal GDP over the average share of the other OECD countries. We distinguish the following time periods: (i) the 1970–1977 period: The US medical care expenditure gap hovered around 1.25, (ii) the 1978–1990 period: The US medical care expenditure gap increased steadily from 1.23 to 1.68, and (iii) the 1991–2007 period: The US medical care expenditure gap stabilized around 1.70. Hence, the 1978–1990 period stands out as a transition episode with the sharpest increase in the US medical care expenditure gap. One major difficulty in uncovering these growth patterns for US medical expenditures appears to be the high degree of uniformity in the various National Health Expenditure Accounts’ categories by source and use of funds. These regular trends prevail over the 1978–1990 transition period. When adjusting for inflation in the medical sector, we obtain that the ratio of real health care expenditures over real GDP is quite flat over the entire sample period, but we observe some slight growth in the share of US value added. We first perform a growth accounting exercise based on value added to assess the contribution of observable production factors. This exercise focuses on the supply side of the economy, and hence it has the advantage of circumventing obvious issues related to price and quality data as well as the various degrees and forms of government intervention across countries. As is well known, growth accounting methods become most powerful under a stable influence of technological progress across economies, and a similar composition and productivity of labor and capital over time. The ratio of physician to non-physician health care workers has been quite stable in our sample of countries for the periods in which data is available. Also, in spite of global growth in markups, capital, and labor shares, trends have evolved quite similarly within our sample of countries. For the 1970–2007 sample period, labor in the health care sector has grown faster in the OECD, whereas capital in the health care sector has grown faster in the United States. Overall, our growth accounting exercise based on value added indicates that the US medical care expenditure gap should have gone down in the period 1970–2007, whereas it went up by about 30% in the data. Therefore, the divergent patterns of growth of health care expenditures between the United States and the OECD cannot be explained by factor accumulation. We then provide some supporting evidence to account for the wedge between growth in value added and US health care expenditures. Various international institutions such as the OECD and the World Health Organization periodically report cross-country rankings on various dimensions of health status, health care activities, quality of care, and access to care. In many of these rankings the US lags behind many other OECD countries. Hence, the assumption of similar growth in technology and product quality across countries seems quite plausible at the aggregate level. Moreover, it is unclear that technology could account for the sharp break in the trend of the US medical gap over the 1978–1990 transition period—just spanning over a single decade. Indeed, US medical care expenditures are characterized by large corporate markups, gross profits, and stock market returns. Periods of strong positive growth in the ratio of HCE over GDP are backed by contemporaneous and commensurate changes in gross profits and markups. When we filter out price effects we find that the growth of the product quality residual in the US falls into the OECD average. Whereas medical prices in some OECD countries followed the general pace of inflation, US medical inflation has considerably outpaced general inflation. The health economics literature has identified various factors that account for the higher prices of prescription drugs in the United States, which may also extend to the observed high prices of hospital and other medical services. Health insurance premiums present much higher price increases than the CPI-medical care. Increasing charges in hospital and related services appear to be a main driving force in the US health care sector over the 1978–1990 transition period. Finally, our methods impose severe limitations on the form of the health care production function. A statistical-based approach could also identify main drivers of health care expenditure patterns. In fact, in our supporting exercises we have reported several estimates of price elasticities for health care demand, and one would need to check whether or not these elasticities are statistically significant across countries. In unreported results available upon request, we have run panel data regressions for the US medical care expenditure gap with respect to each individual country as the dependent variable and the corresponding US gaps for the following four independent variables: The medical care price, GDP per capita, population 65 years old and over, and life expectancy. In this strongly balanced panel with ten cross-sectional observations and 38 time-series observations, we find that only the US medical care price gap and the US GDP per capita gap are statistically significant. For our sample period, however, the US GDP per capita gap went down by about 10%, and hence it works in the opposite direction. As discussed previously when considering gaps by country, this income gap may pick up a catching-up effect as a result of converging GDP per capita trends in the OECD. Appendix A: A Simple Model of Health Care Expenditures Let us just consider an endowment economy with a representative agent. The economy can also be reinterpreted as a social planning problem for the optimal allocation between a variety of medical services and all the other goods. A key parameter in the model is the price elasticity for health care demand. We should note that even if the health care good is subsidized the price elasticity of demand will not be changed when the representative agent has to pay a constant fraction of medical expenditures. At every time t = 0, 1, …, the economy receives yt units of an aggregate commodity that can be transformed into two types of consumption goods: A composite consumption good c and a variety of health care consumption goods ms for s ∈ [0, σ(at)], where at denotes the technology level at time t and [0, σ(at)] is the mass of available varieties. Preferences are represented by a CES utility function. All health care varieties mts at time t enter symmetrically into a utility aggregator |$M(t)=\big[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\big]^{\frac{1}{\gamma }}$|⁠. The representative agent solves the following budget-constrained maximization problem: \begin{eqnarray*} &&{\max _{c,h,m_{s}} \quad \sum _{t=0}^{\infty }\beta ^{t}\left[\!\lambda c_{t}^{\rho }+(1-\lambda )\left(\phi (a_{t})^{\frac{1}{\rho }}\left[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\right]^{\frac{1}{\gamma }}\right)^{\rho }\!\right]^{\frac{1}{\rho }},}\\ &&\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\text{subj. to:} \quad\,\,\, c_{t}+q_{t}h_{t}=y_{t},\\ &&\,\,\,\, \int _{0}^{\sigma (a_{t})}m_{ts}ds=a_{t}h_{t}, \\ &&\,\,\,\, 0<\beta <1,0<\lambda <1,0<\gamma <1,-\infty <\rho <1, \end{eqnarray*} where qtht represents nominal health care expenditures, qt is the relative price, and ht represents real expenditures. Parameter λ is called the consumption share parameter. Parameter ρ determines the degree of substitution between the composite consumption good c and the health care utility aggregator |$M_{t}=[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds]^{\frac{1}{\gamma }}$|⁠. Parameter γ determines the degree of substitution of the health care varieties ms. Function ϕ(at) allows for shifts in the expenditure share as a result of technological change. An increase in the technology level at may change the composition of expenditures through the following three channels: (i) the price effect: An increase in at lowers the relative price of health care varieties qt/at in terms of the numeraire good. This effect is present in economic growth models of embodied technological change (e.g., Greenwood, Hercowitz, and Krusell 1997). (ii) the productivity effect: An increase in at results in higher productivity because it expands the mass of available varieties [0, σ(at)] to allow for a more efficient production of health care utility |$M(t)=\big[\int _{0}^{\sigma (a_{t})}m_{ts}^{\gamma }ds\big]^{\frac{1}{\gamma }}$|⁠. That is, for 0 < γ < 1, the same utility level Mt can be obtained under lower spending. This effect is present in economic growth models with a continuum of product varieties (cf., Romer 1990). And (iii) the expenditure effect: An increase in at may shift the consumption expenditure share because technological change may expand the domain of application of health care varieties. This effect is reflected in function ϕ(at), and allows for an increase in the health expenditure share under a unitary income elasticity as documented in several studies (Chernew and Newhouse 2012). Note that for an inelastic demand (i.e., ρ < 0) both the price and productivity effects (i)–(ii) may lead to a decrease in the health care expenditure share under an increase in a. Hence, function ϕ(a) builds in some further flexibility to model the effects of a change in at on health care expenditures qtht. There are some other well-known models in which technological change may generate nonlinear Engel curves (e.g., Becker, Philipson, and Soares 2005; Hall and Jones 2007). Although this rich form for the utility function contemplates various channels for the influence of technological change, we will again collect all of these effects as part of the nonobservable residual. Optimality Conditions. The representative agent assumes that the relative price qt and the level of technological change at are exogenously given. Over the optimal solution, consumption must be constant across medical varieties, that is, mts = mt for all s. Then, from the first-order conditions of the agent's optimization problem we obtain the optimal ratio of average consumption of health care varieties mt over the composite consumption of all other goods ct, \begin{equation*} \frac{m_{t}}{c_{t}}=\left(\frac{1}{q_{t}}\right)^{\frac{1}{1-\rho }}\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}\left(a_{t}\phi (a_{t})\sigma (a_{t})^{\frac{\rho -\gamma }{\gamma }}\right)^{\frac{1}{1-\rho }}. \end{equation*} Now, multiplying both terms by relative price qt, we can express the ratio of health care expenditures over total expenditures in non-health care goods as follows: \begin{equation*} \frac{q_{t}h_{t}}{c_{t}}=\left(\frac{1}{q_{t}}\right)^{\frac{\rho }{1-\rho }}\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}a_{t}^{\frac{\rho }{1-\rho }}\phi (a_{t})^{\frac{1}{1-\rho }}\sigma (a_{t})^{\frac{\rho (1-\gamma )}{\gamma (1-\rho )}}. \end{equation*} These equation provides an expression for the evolution of health care expenditures relative to non-health care expenditures as a function of the relative price between the two goods and a residual term |$\widehat {A}{}_{t}$| collecting the underlying quality effects, \begin{equation*} \widehat {A}_{t}=\left(\frac{1-\lambda }{\lambda }\right)^{\frac{1}{1-\rho }}a_{t}^{\frac{\rho }{1-\rho }}\phi (a_{t})^{\frac{1}{1-\rho }}\sigma (a_{t})^{\frac{\rho (1-\gamma )}{\gamma (1-\rho )}}. \end{equation*} Appendix B: Main Definitions Data Sources Health Care Expenditures (OECD Data). “The expenditure on activities that—through application of medical, paramedical, and nursing knowledge and technology—has the goals of: Promoting health and preventing disease; Curing illness and reducing premature mortality; Caring for persons affected by chronic illness who require nursing care; Caring for persons with health-related impairments, disability, and handicaps who require nursing care; Assisting patients to die with dignity; Providing and administering public health; Providing and administering health programs, health insurance and other funding arrangements. With this boundary, general public safety measures such as technical standards monitoring and road safety are not considered as part of expenditure on health. Activities such as food and hygiene control and health research and development are not included in total health expenditure.” (OECD Health Data 2012 Definitions, Sources and Methods; available at http://stats.oecd.org) Medical Care Prices. Medical care prices from 1970 to 1977 for the eleven OECD countries in the sample are from Gillion, Schieber, and Poullier (1985). Remaining data come from the following sources: Australia. Data from 1989 to 2007 corresponds to the health component of the CPI (available from the Australian Bureau of Statistics at http://www.abs.gov.au). Data from 1978 to 1988 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). Canada. Data corresponds to the health and personal care price index component of the CPI (available from Statistics Canada at http://www76.statcan.gc.ca). Denmark. Data from 1996 to 2007 corresponds to the CPI-health price, available at the FRED database (Federal Reserve Bank of St. Louis). Data from 1978 to 1995 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). Finland. Data from 1996 to 2007 corresponds to the CPI-health price, available at the FRED database (Federal Reserve Bank of St. Louis). Data from 1978 to 1995 corresponds to the health price index published by the OECD (available at http://stats.oecd.org/). France. The CPI-medical care is the union of the following three price indexes: (i) the health services up to year 1992 (and then discontinued), (ii) the medical services and health care expenditures up to year 1998 (and then discontinued), and (iii) the health services from 1998 to 2007 (available from the French National Institute for Statistics and Economic Studies at http://www.bdm.insee.fr). Germany. Data up to 1983 is from Gillion et al. (1985). Data from 1991 until 2007 corresponds to the health component of the CPI (available from the German Federal Statistics Office at https://www.destatis.de/). Data for the missing period 1984–1990 has been interpolated using data from Schieber et al. (1994). Ireland. Data corresponds to the health subcategory of the CPI, which is available from the Irish Central Statistics Office at http://www.cso.ie/. Japan. Data corresponds to the medical care item of the CPI [available from the Japanese Statistics Bureau, Director-General for Policy Planning (Statistical Standards) and Statistical Research and Training Institute at http://www.stat.go.jp/]. Spain. The CPI-medical care is constructed as the union of several price indexes. From 1977 to 1992 we used the CPI-medicine (IPC-medicina). For the period 1993–2001 we used the average change in five subcategories of the CPI: medical, dental, and non-hospital paramedical services (servicios médicos, dentales y paramédicos no hospitalarios), drugs and other pharmaceutical products (medicamentos y otros productos farmacéuticos), machines, therapeutic material and its repairs (aparatos y material terapéutico y sus reparaciones), hospital care (cuidados en hospitales y similares), and medical insurance (seguros médicos). From 2002 to 2007 we used the average change in the three available subcategories: Drugs, pharmaceutical products and therapeutic material (medicamentos, otros productos farmacéuticos y material terapéutico), medical, dental, and non-hospital paramedical services (servicios médicos, dentales y paramédicos no hospitalarios), and hospital services (servicios hospitalarios). Available from the Instituto Nacional de Estadística at http://www.ine.es/. United Kingdom. Data up to 1983 is taken from Gillion et al. (1985). Data from 1988 until 2007 corresponds to the health component of the CPI (available from the Office for National Statistics at http://www.ons.gov.uk/). Data from the missing period 1984–1987 has been interpolated using data from Schieber et al. (1994). United States. Data corresponds to the CPI-medical care (available from the Bureau of Labor Statistics at http://www.bls.gov/). Acknowledgments We would like to thank an editor and four anonymous reviewers for their constructive comments. Notes The editor in charge of this paper was Dirk Krueger. Footnotes 1 As discussed in what follows, changes in relative prices and public policies can account for some of these expenditure patterns. 2 As documented in Cutler and Ly (2011), three major factors may account for the high level of HCE in the United States: (i) Higher cost of production factors, (ii) Greater share of administrative expenses, and (iii) More generous provision of medical services. As our main goal is to assess the evolution of the US medical care expenditure gap, we shall be concerned with all sources of factor accumulation in the health care sector, and so we focus on growth in value added rather than on actual levels of all these economic variables. 3 The economics literature usually estimates technological change from price data. According to the Boskin Commission Report (1996) the US consumer price index does not take into account a 3% yearly increase in the quality of health care goods. Unfortunately, we were not able to find data on consumption quality for the other OECD countries. 4 We calculate average worker compensation as the ratio of compensation of employees (NIPA Table 1) over full-time and part-time employees (NIPA Table 6.4). These NIPA tables are available at http://www.bea.gov/iTable. 5 Health care investment as percentage of HCE data are available from OECD Health Care Statistics (http://stats.oecd.org/). This item comprises gross capital formation of domestic health care provider institutions (excluding retail sale and other providers of medical goods). Health care investment could be privately or publicly financed. 6 Between 1970 and 2007, for the value-weighted portfolio of all the publicly traded companies in the United States (excluding ADRs and REITs), the yearly gross return was 11.2%, which translates into a yearly rate of return of approximately 8% after adjusting for inflation (data available from the Center for Research in Security Prices). 7 In unreported simulations available upon request, we estimate αH for each country in our sample using the World Input-Output Database tables (http://http://www.wiod.org) for the year 2000. We found that the capital income share αH in the health care sector for Canada, Germany, Japan, and the United States are quite similar (0.23, 0.21, 0.26, and 0.23 respectively), whereas those of Australia, Spain, and the United Kingdom are significantly lower (0.12, 0.16, and 0.10 respectively). Allowing αH to vary across countries does not change our qualitative results. If anything, it strengthens our conclusion that factor accumulation cannot account for the evolution of the US medical care expenditure gap. 8 See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/det_49_ind_port.html. The list of companies used to construct the health care portfolio and SIC codes is also available from the authors upon request. 9 A firm's market capitalization is the total number of shares outstanding times the price. 10 The return of the market weighted portfolio also comes from Kenneth French database. See http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/Data_Library/f-f_factors.html. 11 Infant mortality is the yearly number of deaths of children under 1 year of age per 1,000 live births. Neonatal mortality is the yearly number of deaths of children under 28 d of age per 1,000 live births. Perinatal mortality is the yearly number of deaths of children within 1 week of birth (early neonatal deaths) plus fetal deaths of minimum gestation period of 28 weeks or minimum fetal weight of 1,000 grams per 1,000 births (source: http://stats.oecd.org/). 12 Except for life expectancy, the set of countries used to construct the quality estimates in Table 4 is different from the eleven countries in the original comparison group. Since we could not find data for some of the countries in our original sample, we decided to use all the OECD countries with available data. Five-year observed survival rates are defined as the number of people diagnosed with cancer (age 15–99) within a certain period surviving five years after diagnosis, over the number of people diagnosed with the same cancer (age 15–99) within a certain period. Five-year relative survival rates are defined as the observed rate of people diagnosed with cancer (age 15–99) surviving five years after diagnosis, over an expected survival rate of a comparable group from the general population (see http://stats.oecd.org/). 13 Although the value of the standard deviation over the mean for life expectancy is 0.024 for our sample of OECD countries, this same ratio for all countries in the world in 2003 is about 0.145, around six times larger (data available at http://data.un.org/). There is no general data about survival rates for different types of cancer across countries. The American Cancer Society (2008) publishes some statistics for a few developing countries, and there is much more variability in survival rates than in the OECD. For example, the breast cancer 5-year relative survival rates in 2008 for China, India, Thailand, and Uganda are 0.82, 0.52, 0.63, and 0.46, respectively. For the sample of 18 countries reported in Table 4 the minimum is 0.79 and the maximum is 0.89. 14 This is in line with our analysis of financial variables in Section 2.3, where after the 1978–1990 transition period we observe that relative markups between the US health care sector and the rest of the economy have an upward trend whereas relative gross profits remain quite flat. Thus, the evidence suggests that in relative terms, real medical consumption may have declined in the United States because of cost containment whereas prices in the health care sector were still outpacing those of the rest of the economy. 15 Kessler (2011) argues that although the indemnity payments and administrative expenses of the malpractice system amount to less than 1% of health care spending, the costs of defensive medicine are likely to be far greater because neither patients nor physicians bear most of the marginal costs. References American Cancer Society ( 2008 ). Global Cancer Facts and Figures , 2nd ed . American Cancer Society . Anderson G. , Reinhardt U. , Hussey P. , Petrosyan V. ( 2003 ). “ It's the Prices, Stupid: Why the United States is So Different from other Countries .” Health Affairs , 22 , 89 – 105 . Google Scholar Crossref Search ADS PubMed Anderson G. , Frogner B. ( 2008 ). “ Health Spending in OECD Countries: Obtaining Value per Dollar .” Health Affairs , 27 , 1718 – 1727 . Google Scholar Crossref Search ADS PubMed Becker G. , Philipson T. , Soares R. ( 2005 ). “ The Quantity and Quality of Life and the Evolution of World Inequality .” American Economic Review , 95 (1) , 277 – 291 . Google Scholar Crossref Search ADS PubMed Berndt E. , Cutler D. , Frank R. , Griliches Z. , Newhouse J. , Triplett J. ( 2000 ). “ Medical Care Prices and Output .” Handbook of Health Economics , Vol. 1 , edited by Culyer A. , Newhouse J. , Elsevier, Amsterdam, pp. 119 – 180 . Berndt E. , Aitken M. ( 2010 ). “ Brand Loyalty, Generic Entry and Price Competition in Pharmaceuticals in the Quarter Century After the 1984 Waxman-Hatch Legislation .” NBER Working Paper 16431, National Bureau of Economic Research, Cambridge, MA. Boskin Commission Report ( 1996 ). “ Toward a More Accurate Measure of the Cost of Living .” Final Report to the Senate Finance Committee from the Advisory Commission to Study the Consumer Price Index , http://www.ssa.gov/history/reports/boskinrpt.html. Cardenas E. ( 1996 ). “ Revision of the CPI Hospital Services Component .” BLS Monthly Labor Review , 119 , 40 – 48 . Chandra H. Skinner J. ( 2012 ). “ Technology Growth and Expenditure Growth in Health Care .” Journal of Economic Literature , 50 , 645 – 680 . Google Scholar Crossref Search ADS Chernew M. Newhouse J. ( 2012 ). “ Health Care Spending Growth .” Handbook of Health Economics , Vol. 2 , edited by Pauly M. Mcguire T. Barros P. , Elsevier, Amsterdam, pp. 1 – 43 . Cohen R. Makuc D. Bernstein A. Bilheimer L. Powell-Griner E. ( 2009 ). “ Health Insurance Coverage Trends, 1959–2007: Estimates from the National Health Interview Survey .” National Health Statistics Reports 17 . Comin D. Hobijn B. ( 2004 ). “ Cross-Country Technology Adoption: Making the Theories Face the Facts .” Journal of Monetary Economics , 51 , 39 – 83 . Google Scholar Crossref Search ADS Comin D. Hobijn B. ( 2010 ). “ An Exploration of Technology Diffusion .” American Economic Review , 100 (5) , 2031 – 2059 . Google Scholar Crossref Search ADS Cutler D. Ly D. ( 2011 ). “ The (Paper) Work of Medicine: Understanding International Medical Costs .” Journal of Economic Perspectives , 25 (2) , 3 – 25 . Google Scholar Crossref Search ADS PubMed Cutler D. McClellan M. Newhouse J. ( 2000 ). “ How Does Managed Care Do It? ” RAND Journal of Economics , 31 , 526 – 548 . Google Scholar Crossref Search ADS PubMed Danzon P. ( 1991 ). “ Liability for Medical Malpractice .” Journal of Economic Perspectives , 5 (3) , 51 – 69 . Google Scholar Crossref Search ADS Danzon P. Chao L. ( 2000 ). “ Cross-National Price Differences for Pharmaceuticals: How Large, and Why? ” Journal of Health Economics , 19 , 159 – 195 . Google Scholar Crossref Search ADS PubMed Danzon P. Furukawa M. ( 2003 ). “ Prices and Availability of Pharmaceuticals: Evidence from Nine Countries .” Health Affairs , 22 , 521 – 536 . Google Scholar Crossref Search ADS Danzon P. Furukawa M. ( 2011 ). “ Cross-National Evidence on Generic Pharmaceuticals: Pharmacy vs. Physician-Driven Markets .” NBER Working Paper 17226 , National Bureau of Economic Research, Cambridge, MA. De Long B. Summers L. ( 1992 ). “ Equipment Investment and Economic Growth: How Strong is the Nexus? ” Brookings Papers on Economic Activity , 2 , 157 – 199 . Google Scholar Crossref Search ADS Di Matteo L. ( 2005 ). “ The Macro Determinants of Health Expenditures in the United States and Canada: Assessing the Impact of Income, Age Distribution and Time .” Health Policy , 71 , 23 – 42 . Google Scholar Crossref Search ADS PubMed Dranove D. Lindrooth R. White W. Zwanziger J. ( 2008 ). “ Is the Impact of Managed Care on Hospital Prices Decreasing? ” Journal of Health Economics , 27 , 362 – 376 . Google Scholar Crossref Search ADS PubMed Drösler S. Romano P. Wei L. ( 2009 ). “ Health Care Quality Indicators Project: Patient Safety Indicators Report 2009 .” OECD Health Working Paper 47 , Organisation for Economic Co-operation and Development, Paris, France. Dunn A. Shapiro A. ( 2011 ). “ Physician Market Power and Medical-Care Expenditures .” BEA Working Paper 84 , Bureau of Economic Analysis, Washington, DC. Finkelstein A. ( 2007 ). “ The Aggregate Effects of Health Insurance: Evidence from the Introduction of Medicare .” Quarterly Journal of Economics , 122 , 1 – 37 . Google Scholar Crossref Search ADS Fixler D. Jaditz T. ( 2002 ). “ An Examination of the Difference Between the PCE Deflator and the CPI .” BLS Working Paper 361 , Bureau of Labor Statistics, Washington, DC. Garber A. Skinner J. ( 2008 ). “ Is American Health Care Uniquely Inefficient? ” Journal of Economic Perspectives , 22 (4) , 27 – 50 . Google Scholar Crossref Search ADS PubMed Gillion C. Schieber G. Poullier J. ( 1985 ). “ Measuring Health Care, 1960–1983: Expenditure, Costs, and Performance .” OECD Social Policy Studies 2 . Goldman D. McGlynn E. ( 2005 ). “ U.S. Health Care: Facts About Cost, Access, and Quality .” RAND Corporate Publication Series CP-484/1-(1/05) . Greenwood J. Hercowitz Z. Krusell P. ( 1997 ). “ Long-Run Implications of Investment-Specific Technological Change .” American Economic Review , 87 (3) , 342 – 362 . Griliches Z. Cockburn I. ( 1994 ). “ Generics and New Goods in Pharmaceutical Price Indexes .” American Economic Review , 84 (5) , 1213 – 1232 . Hall R. Jones C. ( 2007 ). “ The Value of Life and the Rise in Health Spending .” Quarterly Journal of Economics , 122 , 39 – 72 . Google Scholar Crossref Search ADS Harrington S. Danzon P. Epstein A. ( 2008 ). “ Crises in Medical Malpractice Insurance: Evidence of Excessive Price-Cutting in the Preceding Soft Market .” Journal of Banking & Finance , 32 , 157 – 169 . Google Scholar Crossref Search ADS Jacoby E. Hefner D. ( 1971 ). “ Domestic and Foreign Prescription Drug Prices .” Social Security Bulletin , 34 , 15 – 22 . Kessler D. ( 2011 ). “ Evaluating the Medical Malpractice System and Options for Reform .” Journal of Economic Perspectives , 25 (2) , 93 – 110 . Google Scholar Crossref Search ADS PubMed Koechlin F. Lorenzoni L. Schreyer P. ( 2010 ). “ Comparing Price Levels of Hospital Services Across Countries: Results of Pilot Study .” OECD Health Working Papers 53 , Organisation for Economic Co-operation and Development, Paris, France. Koijen R. Philipson T. Uhlig H. ( 2016 ). “ Financial Health Economics .” Econometrica , 84 , 195 – 242 . Novy-Marx R. ( 2013 ). “ The Other Side of Value: The Gross Profitability Premium .” Journal of Financial Economics , 108 , 1 – 28 . Google Scholar Crossref Search ADS Ringel J. Hocek S. Vollaard B. Mahnovski S. ( 2000 ). “ The Elasticity of Demand for Health Care: A Review of the Literature and Its Application to the Military Health System .” National Defense Research Institute/RAND . Romer P. ( 1990 ). “ Endogenous Technological Change .” Journal of Political Economy , 98 , 71 – 102 . Google Scholar Crossref Search ADS Rodwin V. ( 2003 ). “ The Health Care System under French National Health Insurance: Lessons for Health Reform in the United States .” The American Journal of Public Health , 93 , 31 – 37 . Google Scholar Crossref Search ADS PubMed Rodwin V. Sandier S. ( 1993 ). “ Health Care under French National Health Insurance .” Health Affairs , 12 , 111 – 131 . Schieber G. Poullier J. Greenwald L. ( 1994 ). “ Health System Performance in OECD Countries 1980–1992 .” Health Affairs , 13 , 100 – 112 . Google Scholar Crossref Search ADS PubMed Skinner J. Fisher E. Wennberg J. ( 2005 ). “ The Efficiency of Medicare .” In Analyses in the Economics of Aging . University of Chicago Press , pp. 129 – 157 . Smith S. Newhouse J. Freeland M. ( 2009 ). “ Income, Insurance and Technology: Why Does Health Spending Outpace Economic Growth? ” Health Affairs , 28 , 1276 – 1284 . Google Scholar Crossref Search ADS PubMed US Department of Commerce ( 2004 ). Pharmaceutical Price Controls in OECD Countries . International Trade Administration , Washington, DC . US General Accounting Office ( 1992 ). Prescription Drugs: Companies Typically Charge More in the United States than in Canada . Government Printing Office , Washington, DC . US General Accounting Office ( 1994 ). Prescription Drugs: Companies Typically Charge More in the United States than in the United Kingdom . Government Printing Office , Washington, DC . US General Accounting Office ( 2003 ). Medical Malpractice Insurance: Multiple Factors Have Contributed to Increased Premium Rates . Government Printing Office , Washington, DC . © The Author(s) 2017. Published by Oxford University Press on behalf of European Economic Association. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

### Journal

Journal of the European Economic AssociationOxford University Press

Published: Feb 1, 2019

## You’re reading a free preview. Subscribe to read the entire article.

### DeepDyve is your personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month ### Explore the DeepDyve Library ### Search Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly ### Organize Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place. ### Access Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals. ### Your journals are on DeepDyve Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more. All the latest content is available, no embargo periods. DeepDyve ### Freelancer DeepDyve ### Pro Price FREE$49/month
\$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off