National and international tests of the new drug cost offset theory.
Health care costs continue to rise in the United States and throughout the world. The rising medical care costs often mean that individuals and societies must forego other important goods and services as private wallets and fiscal budgets are squeezed. (1) Most analysts point to new medical technologies as the culprit behind rising health care costs (e.g., Newhouse 1992; CBO 2008), but not all new medical technologies are created equal. According to the new drug cost offset theory, new prescription drugs may substitute for more expensive surgical procedures and thereby keep people out of expensive hospitals and other costly medical facilities. Overall health care costs decline as a result.
However, empirical findings regarding the new drug cost offset theory have been mixed. In addition, very few studies have tested this theory at the macroeconomic level to determine if system-wide savings actually accrue. As a result, this article provides some empirical analyses on the relationship between new drug introductions and health care spending both within the United States and within an international context. The next section of this article provides background information on previous studies examining the new drug cost offset theory. The new analysis is taken up in the third section, which explores the relationship between pharmaceutical innovation and various types of health care spending in the United States. This U.S. study is followed by an analysis of the impact of new drug discovery on health care spending within an international context. The final section offers a summary of the article. Both sets of empirical results suggest that new drugs slow the growth of overall health care costs and thereby provide support for the new drug cost offset theory.
2. Previous Studies on the New Drug Cost Offset Theory
Numerous national and international studies suggest that new prescription drugs provide sizeable social benefits. Among the benefits, empirical studies find that new drugs extend the quantity and quality of lives and improve the productivity of the workforce (Lichtenberg 2002, 2004, 2005a, b; Frech and Miller 2004; Cremieux et al. 2005a, b; Shaw, Horrace, and Vogel 2005; Hsieh et al. 2007). However, compared to the effect of new drugs on health outcomes, testing of the new drug cost offset theory has received limited attention in the literature and only at the microlevel.
Among the few studies, Lichtenberg (1996) analyzed whether new drugs keep people out of U.S. hospitals. He showed that the number of hospital beds declined most rapidly for those diagnoses with the greatest increase in the total number of drugs prescribed and the greatest change in the distribution of newer drugs. He went on to compute that a $1 increase in pharmaceutical expenditure is related to a $3.65 reduction in hospital care expenditure but also associated with a $1.54 increase on ambulatory care expenditures.
Lichtenberg (2001) used 1996 data for the United States to examine if newer drugs tend to lower all types of nondrug medical spending. His results suggest that replacing an older drug with a newer one would increase drug costs by $18 but lower hospital spending by $44 and all types of nondrug spending by $71 on a per capita basis. Lichtenberg (2002) updated his earlier study and found that a reduction in the age of drugs utilized reduces nondrug expenditures 7.2 times as much as it increases drug expenditures among the entire population. Most of the cost reduction is due to lower hospital spending.
Zhang and Soumerai (2007) argue that Lichtenberg's (2001) study contained a number of flaws. However, after correcting for many of them, they still provide evidence for a new drug offset effect. Miller, Moeller, and Stafford (2005/2006) find no evidence of a drug offset effect for new cardiovascular drugs, but Lichtenberg (2006) argues that their article contains a number of serious mistakes. Duggan (2005) determines that the introduction of second generation antipsychotic drugs did not result in a reduction in nondrug health care costs among Medicaid recipients. But Soumerai et al. (1994) determine that limits on Medicaid coverage for prescription drugs can lead to an increased use of acute mental health care among those with chronic mental illnesses.
Finally, Cremieux, Ouellette, and Petit (2007), using a production function approach and a panel data set of Canadian provinces, find that increases in drug spending can be more than offset by decreases in medical care spending without affecting population health. Like Soumerai et al. (1994), however, their study does not actually test the new drug offset effect because spending on all prescription drugs, both old and new, are captured in their measure of drug spending.
Thus, given the relatively limited supply of research on this topic, the mixed results, and the lack of any macrofocused studies, this article uses time-series data for the United States over the period from 1960 to 2007 and panel data for some member countries of the OECD over the last few decades to analyze if a new drug cost offset effect exists at the aggregate level.
3. Examining the New Drug Cost Offset Theory in the United States
The conceptual model that motivates both empirical analyses begins by assuming a one-period model during which a representative consumer, given her exogenous tastes and preferences (influences), I, derives utility from consuming units of some composite good, x, and units of health services, s, that flow from her stock of health capital, h. Stated mathematically,
U = U(s, x; I). (1)
Utility is assumed to increase at a decreasing rate with respect to both health services and the composite good. It is further assumed that health services can be produced with varying +combinations of prescription drugs, q, and medical services, m, such as office visits or inpatient days, conditioned on the representative consumer's initial endowment of health capital, [h.sub.0], and the current state of medical technology, T. Thus, for ease of exposition, we ignore a set of other health care "goods" and "bads" such as exercise, diet, alcohol, and tobacco use, etc., and the consumer's time involved in producing these health care activities. A production function for units of health services can thus be written as
s = s(q, m; [h.sub.0], T), (2)
where s is assumed to be concave with respect to both q and m.
Substituting Equation 2 into Equation 1 allows utility to be expressed in terms of the production function for health services and the composite good. It is assumed that expenditures on the two inputs to produce health services, and spending on the composite good, fully exhaust the consumer's income net of taxes and insurance premiums, y. The consumer's optimization problem is, therefore, to select the amounts of prescription drugs and medical services and the composite good that maximize her utility subject to the constraints of net income and the out-of-pocket prices for drugs, [P.sub.q], medical services, [P.sub.m], and the composite good, [P.sub.x]. Stated more formally, (2)
Maximize U = U[(s(q, m; [h.sub.0], T),x; I] s.t. [P.sub.q]q+ [P.sub.m]m + [P.sub.x]x = y. (3)
Defining x as the numeraire, it follows from the utility maximization process that the representative consumer's quantity demanded of prescription drugs and medical services can be derived as a function of the relative out-of-pocket drug price, relative out-of pocket price of medical services, and her real net income, conditioned upon the parameters in the model. For example, the demand function for prescription drugs can be expressed in general terms as
q = q ([P.sub.q] / [P.sub.x], [P.sub.m] / [P.sub.x], y / [P.sub.x]; I, [h.sub.0], T). (4)
A similar demand equation can be written for medical services.
Horizontally aggregating across all consumers and assuming the demand function can be written in log-log form for estimation purposes, a more specific version of Equation 4 can be written as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)
where [Q.sub.t], identifies aggregate quantity demanded at time t. In addition, the percentage of people aged 65 and older (OLD) and quantity demanded in the previous period ([Q.sub.t-1]) stand for I, life expectancy at birth (LE) represents [h.sub.0], and the number of new molecular entities (NME) replace T. Given the log-log specification, the slope coefficients in Equation 5 can be interpreted as elasticities for nearly all of the variables. According to the law of demand, the quantity demanded of prescription drugs should be inversely related to its relative out-of-pocket price ([[beta].sub.1] < 0). The relationship between the quantity demanded of prescription drugs and the relative out-of-pocket price of medical services depends on whether the two types of medical care are substitutes ([[beta].sub.2] > 0) or complements ([[beta].sub.2] < 0) in consumption. Finally, if prescription drugs can be classified as a normal good, income should have a direct effect on the quantity demanded of prescription drugs ([[beta].sub.3] > 0). An inverse relationship holds if prescription drugs are inferior goods. Similar hypothesis can be drawn for a comparable medical services demand equation.
Estimated versions of Equation 5 for both prescription drugs and medical services can be used to test the new drug cost offset theory. The general idea is that new drugs represent a technological change in the way medical care is produced. Because the innovation is embodied in a new drug, medical care utilization becomes biased towards that newer drug. As a result, a positive coefficient estimate is expected on the measure of new drug innovation in the prescription drug demand equation. Moreover, depending upon the precise degree of substitutability between drugs and nondrug methods of medical care production, nonpharmaceutical medical care utilization may decline. A negative coefficient on the number of new drugs in the medical services equation is a necessary condition to support the new drug offset effect. The sufficient condition is that the cost of higher drug utilization resulting from new drugs must be less than the cost savings associated with the potentially lower use of medical services.
The demands for both prescription drugs and medical services are estimated using national data for the United States over the period from 1960 to 2007. (3) Expenditure data come from the health accounts at the Centers for Medicare and Medicaid Services (CMS). Price data come from the Bureau of Labor Statistics (BLS) and income data are obtained from the Bureau of Economic Analysis. In the regression analysis, the aggregate quantity indices, Q and M, are calculated by dividing each of the two nominal expenditures by their corresponding price indices from the BLS (i.e., prescription drug or medical services consumer price index).
The (nominal) out-of-pocket prices for prescription drugs and medical services are determined by multiplying the ratio of consumer spending to total spending for each type of medical care by the respective consumer price index. These measures are then divided by the general consumer price index to arrive at the relative or real out-of-pocket price for both prescription drugs and medical services for each year. Real net income per capita, Y, is measured by personal disposable income per person less health insurance premiums per capita after dividing by the general consumer price index. Premium data come from the health accounts at CMS.
Data for the percentage of the population that is OLD come from the Bureau of the Census, whereas information on LE is found at the Department of Health and Human Services. Finally, the number of NME approved each year by the Federal Food and Drug Administration (FDA) are found in Lichtenberg (2004) for the years 1960 to 2000 and at the FDA website for the remaining years. The final estimation equation for prescription drugs (and a similar one for medical services) takes the following form:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (6)
Interestingly, Equation 6 captures what Newhouse (1992) believes are the four most important determinants of health care spending growth: changes in out-of-pocket prices, income, ageing, and medical technology.
Notice in Equation 6 that all of the variables, other than the number of new drugs, in the equation are first differenced for estimation purposes. (4) This transformation is made for two reasons. First, the transformation converts these variables into rates of growth. Hence, we can observe how the number of new drugs affects the growth rather than level of the different types of medical care. Because of compounding over time, potentially "bending the curve" represents a more significant achievement than a one-shot reduction in the level of medical care spending. Second, unlike the number of new drugs which tends to cycle over time, medical care expenditures and GDP have generally been found to contain unit roots which result in spurious correlations because of nonstationary trending variables (Bloomquist and Carter 1997; MacDonald and Hopkins 2002). First differencing of the data serves as a common method of correcting for unit roots in the data. (5)
The explanatory variable of key interest in Equation 6 is the number of NME. The sign and magnitude of the coefficient estimate on the number of new drugs captures how new drug innovation affects the growth rate of a particular type of medical care. It is expected that the growth of drug expenditures speeds up with a greater number of newer and more expensive drugs on the market ([[beta].sub.7] > 0). However, if the drug offset effect prevails, the growth of medical services should slow such that [[beta].sub.7] < 0 in the estimated demand equation for medical services. Equation 6 is also estimated for the growth of total health care utilization in the United States. The coefficient estimate on the number of new drugs in that demand equation directly captures the overall marginal net savings, if any, from an additional new drug.
Since theory offers no advice, various lag lengths for the number of NME were tried in addition to a contemporaneous value. Experimentation shows that a one-period contemporaneous measure of the number of new drugs works best in terms of statistical significance with the results remaining fairly robust with respect to specification. While Lichtenberg (2005b, 2009) finds that new drugs influence the level of drug utilization for many years into the future, one should recall that Equation 6 predicts the annual or short-run growth rate rather than the level of the various types of medical care. Certainly, a spike in the number of new drugs may affect the level of medical care utilization several years into the future. Growth in medical care spending, however, may suddenly speed up or slow down in response to an abrupt change in medical technology. In addition, it should be pointed out that the estimated coefficient on the lagged measure of the dependent variable in Equation 6 can be used to generate the long-run change in medical care growth that results from a change in the number of NME, assuming a new long-run equilibrium is finally reached. (6) The long-run change should prove to be much larger than the short-run effect.
Both demand equations are estimated jointly by the seemingly unrelated regression technique. Symmetry of the cross-price term is imposed as a constraint during estimation. The Q-statistics and Breusch-Godfrey tests showed no evidence of serial correlation after first differencing the data. The empirical results are reported in the second and third columns of Table 1.
The first item to notice in Table 1 is that the explanatory variables explain over 50% of the variation in the growth rate of the two types of health care. In addition, the results reveal that the demand for prescription drugs, while price-inelastic, is more elastic than the demand for medical services with respect to its own relative out-of-pocket price. Moreover, the estimated -0.324 own-price elasticity of demand for prescription drugs falls in the middle of the range of estimates found by previous microlevel studies (Coulson and Stuart 1995). The own-price elasticity of medical services, although negative as theory suggests, is not statistically different from zero, perhaps suggesting that the aggregate demand for medical services may be perfectly inelastic.
Another result worth highlighting is that the coefficient estimates on the growth of net personal income per capita are both positive and statistically significant. That means that both prescription drugs and medical services are normal goods. According to the results, a 10% increase in the growth of personal income per capita is associated with a 2.3% rise in the short-run growth rate of medical services. The empirical findings also suggest that neither complementarity nor substitutability holds between drugs and medical services given the statistically insignificant estimate on the cross-price term. While Santerre and Vernon (2006) find some substitutability in demand between drugs and medical care, they do not estimate both demand equations jointly nor specify the number of new drugs in the estimation equation. Thus, these new results indicate the findings of Santerre and Vernon regarding the substitutability between prescription drugs and medical care should not be generalized. However, we learn next that some substitutability does hold between new drugs and nondrug methods of producing health.
Of more importance to the research at hand are the signs and magnitudes of the coefficient estimates on the number of new drugs in the various demand equations. As the new drug cost offset theory suggests, the coefficient estimate on the number of new drugs is positive and statistically significant in the prescription drug demand equation. Its coefficient estimate can be interpreted as meaning that an additional new drug speeds up the growth of drug costs by roughly 0.136 percentage points at the margin. When evaluated for the year 2007 with drug costs per capita of $753, an additional new drug raises spending on drugs in the short run by $1.02 on a per capita basis or $309 million in the aggregate.
However, offsetting the higher drugs costs are the savings on medical services as evidenced by the negative coefficient estimate on the number of new drugs in that demand equation. If 2007 figures are used, the savings on medical services amount to $6.62 on a per person basis. In addition, we can use the coefficient estimate on the number of new drugs in column four (Table 1) to calculate the net total personal health care cost savings, which include spending on prescription drugs and medical devices in addition to medical services. Calculations show that the net overall savings amount to $5.91 per capita if 2007 figures are used. In fact, according to calculations, one additional new drug produces short-run aggregate net medical costs savings of approximately $1.8 billion.
Long-run marginal effects are even greater. Based upon the coefficient estimate on the lagged quantity demanded terms (see footnote 6), the long run marginal effect of a new drug on real total personal health care costs can be computed as -0.183. That means the long-run net medical costs savings from an additional drug amount to $11.38 per person or $3.4 billion in the aggregate if 2007 figures are used. These results also indicate that an additional new drug reduces nondrug expenditures six times as much as it increases drug expenditures in the long run. This estimate comes fairly close to those of Lichtenberg (2002) who finds a reduction in the age of drugs utilized produces nondrug to drug expenditure ratios of 7.2 for the entire population and 8.3 for the Medicare population.
These empirical findings are certainly plausible and both statistically and economically significant. They show that new-drug innovation increases drug costs, as the new-drug cost offset theory predicts, but that these greater drug costs are offset by savings in other sectors of the health economy. On average, a new drug appears to pay for itself by offering savings on other medical care services. It will be interesting to see if the results for the OECD sample corroborate this important finding.
4. Examining the New Drug Cost Offset Theory in an International Context
Most countries do not post on a website the necessary data for new drug approvals like the FDA does in the United States. As a result, an extensive search of the literature was conducted to find any scholarly studies that provide information on the number of new molecular entity or new active substance approvals in various OECD countries. This search yielded two studies concerning new drug approvals by Jefferys et al. (1998) and Lundkvist, Jonsson, and Rehnberg (2005). (7) In addition, information on new drug approvals was found online for two other countries. Finally, we contacted by E-mail either the National Ministry of Health or the National Pharmaceutical Society of each OECD country with a webpage in English. That strategy yielded information on new drug approvals for three additional countries. Consequently, the seven OECD countries used in the empirical analysis, along with their years of data, are Belgium (1986-2004), Canada (1985-2004), Finland (1984-2004), Germany (1995-2004), Japan (1982-2004), Sweden (1987-2000), and the UK (1971-2004). (8)
The rest of the data on medical care expenditures and income were collected from OECD (2006). Unfortunately, the OECD data contains numerous empty cells in various expenditure categories. For example, data on private health care spending are unavailable for inpatient, outpatient, and pharmaceutical expenditures in the UK even though it is thought that private expenditures comprise about 10% of all medical care spending in that country. Also, some countries apparently report information on pharmaceutical expenditures in outpatient settings but not inpatient settings. Finally, it is unclear how countries like Japan separate pharmaceutical expenditures from physician costs when drugs are both prescribed and dispensed by physicians.
Thus, for the sake of consistency, only the net impact of an additional new drug on total medical care spending can be examined empirically. Consequently, an OECD version of the equation that appears in column four of Table 1 is estimated. The only differences are that total rather than personal health care expenditures, gross domestic product instead of net personal income, and time and country fixed effects are specified in the estimation equation. (9) In addition, price data are unavailable, and data on the out-of-pocket fraction are relatively sparse, so the relative out-of-pocket price could not be specified. However, the absence of a price term may matter little given its statistical insignificance for the United States. Also, the presence of public health insurance in these countries greatly insulates the consumer from price considerations. Finally, experimentation shows that a one-year lag in the number of new drug approvals provides the best fit in terms of statistical significance, perhaps illustrating that new drugs disperse more slowly in OECD countries than in the United States.
The regression results for the OECD sample of observations are reported in Table 2. (10) Notice the general similarity of the results shown in the fourth column of Table 1. One similarity is the positive and statistically significant intercept term. The intercept term represents the growth of medical care spending that cannot be explained by the right-hand side variables such as income, ageing, new drugs, and buying habits, as captured by the lagged quantity demanded variable. They likely capture the effect of cost-enhancing medical care technologies such as new medical devices and surgical procedures. The intercept terms are not statistically different from one another (t = 0.83) and suggest that real medical care costs grow by about 3-4% per year as a result of omitted factors such as new nonpharmaceutical medical technologies.
A second similarity is the positive and statistically significant coefficient estimates on the lagged expenditure term. Both estimates are less than one, which is necessary for long-run stability. Another similarity is the positive and statistically significant coefficient on the measures of national income growth, which essentially capture the effect of the business cycle on medical care spending growth. The more remarkable similarity is the coefficient estimates on the number of new drugs. More specifically, the coefficient estimate on the number of new drug approvals equals -0.065 for the OECD sample and -0.095 for the U.S. sample of observations. These estimates are not statistically different from one another (t = 1.07). This similarity attests to the robustness of the results for the United States and offers some additional evidence in support of the new drug cost offset effect.
While the literature on the relationship between new drugs and health outcomes is relatively voluminous, very little research exists on the impact of new drugs on overall medical care costs, particularly on a macrolevel and in an international setting. This is unfortunate because the new drug cost offset theory argues that new drugs pay for themselves through a reduction in nondrug medical care spending. It would be beneficial to know if the new drug offset effect holds universally and in the aggregate.
This article studies the new drug cost offset theory by empirically examining the impact of new drugs on the growth rate of various types of medical care at the national level in the United States and on the growth of total medical care spending in some OECD countries. The results for the United States suggest that the typical new drug speeds up the growth of pharmaceutical spending but slows the growth rate of spending on medical services at the margin. Overall, the typical new drug lowers the growth of overall health care spending per capita by 0.095 percentage points in the short run and 0.183 percentage points in the long run. In the United States, the average new drug is found to produce $1.8 billion of medical care cost savings in the short run and $3.4 billion of savings in the long run.
The results for seven other OECD countries are very similar in terms of the short-run marginal impact of an additional drug. The empirical findings indicate that the typical drug reduces the growth of total medical care spending by 0.065 percentage points in the short run and by 0.087 percentage points in the long run. Although the health economy is more intensively regulated in these countries, calculations suggest the overall costs savings from an additional drug still amount to millions of dollars.
Before concluding, we should point out three potential shortcomings associated with the aggregate approach taken in this article. First, the aggregate approach implicitly assumes that the quality and productivity of new drugs are homogeneous across all disease/therapeutic groups (Hsieh et al. 2007). This is not the case, however, and unfortunately time-series data on the number of new drugs in different disease/therapeutic categories are unavailable for the United States and the sample of OECD countries. Second, the aggregate approach focuses on the average effect of pharmaceutical innovation on health care costs. Although that perspective offers some useful information on whether pharmaceutical innovations as a whole pay for themselves, it does not tell us if a particular new drug pays for itself.
Finally, the results show that the number of new drugs is associated with a reduction in the growth of overall medical care spending after controlling for prices, income, initial endowment of health, percent of the population that is elderly, and previous changes in health care spending (as well as fixed effects for the set of OECD observations). But other important variables may be missing from the estimated equations and thereby biasing the results. One such factor may be the number of new medical devices and surgical and diagnostic procedures introduced each year. These medical interventions also represent important sources of technological advances in medical care practice, but longitudinal data are unavailable for either sample. However, as long as medical device and surgical/diagnostic innovations are not correlated with new drug innovations, the results will remain unbiased. Given that the decision to undertake new drug discovery was probably made about 12 to 15 years earlier and that new drug development often represents a random process of experimentation, there is little reason to suspect that they are correlated. Nevertheless, future studies may want to improve upon this article in these three respects.
Bloomquist, A. G., and R. A. L. Carter. 1997. Is health care really a luxury? Journal of Health Economics 16:207-29.
Congressional Budget Office (CBO). 2008. Technological change and the growth of health care spending. Washington, DC: Congressional Budget Office.
Coulson. N. E., and B. C. Stuart. 1995. Insurance choice and the demand for prescription drugs. Southern Economic Journal 61:1146-57.
Cremieux, Pierre-Yves, Marie-Claude Meilleur, and Pierre Ouellette, et al. 2005a. Public and private pharmaceutical spending as determinants of health outcomes in Canada. Health Economies 14:107-16.
Cremieux, Pierre-Yves, Pierre Ouellette, and Marie-Claude Meilleur, et al. 2005b. Pharmaceutical spending and health outcomes in the United States. In Investing in health: The social and economic benefits of health care innovation, edited by Irina Farquhar, Kent Summers, and Alan Sorkin. Oxford: Elsevier Science, pp. 59-76.
Cremieux, Pierre-Yves, Pierre Ouellette, and Patrick Petit. 2007. Do drugs reduce utilization of other healthcare resources? Pharmacoecouomics 25:209-21.
Cutler, David M., and Mark McClellan, 2001. Is technological change in medicine worth it? Health Affairs 20:11-29.
Duggan, Mark. 2005. Do new prescription drugs pay for themselves? The case of second generation antipsychotics. Journal of Health Economies 24:1-31.
Frech, H. E., and Richard D. Miller, Jr. 2004. The effects of pharmaceutical consumption and obesity on the quality of life in the Organization of Economic Cooperation and Development (OECD) countries. Pharmacoeconomics 22:25-36.
Greene, William. 2004. Distinguishing between heterogeneity and inefficiency: Stochastic frontier analysis of the World Health Organization's panel data on national health care systems. Health Economics 13:959-80.
Greene, William. 2005. Fixed and random effects in stochastic frontier models. Journal of Productivity 23:7-32.
Grubaugh, Stephen G., and Rexford E. Santerre. 1994. Comparing the performance of health care systems: An alternative approach. Southern Eeonomic Journal 60:1030-42.
Hsich, Chee-Ruey, Kuang-Ta Vance Lo, Yichen Hong, and Ya-Chen Shih. 2007. Pharmaceutical innovation and health: Evidence from Taiwan. In Pharmaceutical innovation: Incentives, competition, and cost-benefit analysis in international perspective, edited by Frank A. Sloan and Chee-Ruey Hsieh. Cambridge: Cambridge University Press, pp. 242-62.
Jefferys, David B., Diane Leakey, John A. Lewis, Sandra Payne, and Michael D. Rawlins. 1998. New active substances authorized in the United Kingdom between 1972 and 1994. British Journal of Clinical Pharmacology 45:151-6.
Lichtenberg, Frank R. 1996. Do (more and better) drugs keep people out of hospitals? American Econamic Review 86:384-8.
Lichtenberg, Frank R. 2001. Are the benefits of newer drugs worth their costs? Evidence from the 1996 MEPS. Health Affairs 20:241-51.
Lichtenberg, Frank R. 2002. Benefits and costs of newer drugs: An update. NBER Working Paper No. 8996.
Lichtenberg, Frank R. 2004. Sources of U.S. longevity increase, 1960 to 2001. The Quarterly Review of Economics and Finance 44:369-89.
Lichtenberg, Frank R. 2005a. Availability of new drugs and Americans' ability to work. Journal of Occupational & Environmental Medicine 47:373-80.
Lichtenberg, Frank R. 2005b. The impact of new drug launches on longevity: Evidence from longitudinal disease-level data from 52 countries, 1982 to 2001. International Journal of Health Care Finance and Economies 5:47 73.
Lichtenberg, Frank R. 2006. Letter to the editor: On "New cardiovascular drugs: patterns of use and association with non-drug health expenditures." Inquiry 43:80-1.
Lichtenberg, Frank R. 2007. Effect of new drugs on overall health spending: Frank Lichtenberg responds. Health Affairs 26:887-90.
Lichtenberg, Frank R. 2009. The effect of new cancer approvals on the life expectancy of American cancer patients. Economies of Innovation and Technology 18(5):407-28.
Lundkvist, Jonas, Bengt Jonsson, and Clas Rehnberg. 2005. Cost-effectiveness of new drugs: A systematic review of published evidence for new chemical entity drugs introduced on the Swedish market 1987-2000. International Journal of Technological Assessment 21:187-93.
MacDonald, G., and S. Hopkins. 2002. Unit root properties of OECD health care expenditure and GDP data. Health Economics 11:371-6.
Miller, G. Edward, John F. Moeller, and Randall S. Stafford. 2005/2006. New cardiovascular drugs: Patterns of use and association with non-drug health expenditures. Inquiry 42:397-412.
Newhouse, Joseph P. 1992. Medical care costs: How much welfare loss? Journal of Economic Perspectives 6:3-22.
Organization for Economic Cooperation and Development (OECD). 2006. OECD Health Data 2006. Paris: OECD.
Santerre, Rexford E., and John A. Vernon. 2006. Assessing consumer gains from a drug price control policy in the U.S. Southern Economic Journal 73:233-45.
Shaw, James W., William C. Horrace, and Ronald J. Vogel. 2005. The determinants of life expectancy: An analysis of OECD data. Southern Economic Journal 71:768-83.
Soumerai, Stephen B., Thomas J. McLaughlin, Dennis Ross-Degnan, Christina S. Casteris, and Paola Bollini. 1994. Effects of limiting Medicaid drug-reimbursement benefits on the use of psychotropic agents and acute mental health services by patients with schizophrenia. The New England Journal of Medicine 331:650-5.
Studenmund, A. H. 2006. Using econometrics. Boston: Pearson/Addison Wesley.
Zhang, Yuting, and Stephen B. Soumerai. 2007. Do newer prescription drugs pay for themselves? A reassessment of the evidence. Health Affairs 26:880-90.
(1) High and rising medical care costs do not necessarily reflect inefficiency. The costs must be compared to the benefits associated with the medical care before drawing conclusions about inefficiency. Cutler and McClellan (2001) show that the benefits of five medical technologies, which explain a large proportion of the change in medical care spending over time in the United States, outweigh their corresponding costs.
(2) It is assumed that the insurance purchase decision has already been made. Santerre and Vernon (2006) construct a similar model.
(3) Consider that the national data reflects millions of choices made each year over more than 40 years.
(4) The number of new drugs is not first differenced because new drugs already represent the change in knowledge from one year to the next. Also, unit root tests show that the number of new molecular entities does not contain a unit root.
(5) First differencing of the data does not eliminate the intercept term if a time trend is initially specified in the levels equation.
(6) The long-run marginal effect of an additional new drug can be expressed as [[beta].sub.7]/( 1 - [[beta].sub.5]). As a result, the final estimation equations represent dynamic growth models. A dynamic model results in more degrees of freedom and avoids the multicollinearity that often occurs when a finite distributed lag model is used (see Studenmund 2006).
(7) However, data from the Jefferys et al. (1998) study only covered the years 1972 to 1994. New drug approval data for the period 1971 to 2004 were provided by the Medicines and Healthcare Products Regulatory Agency in the UK.
(8) Numerous studies, such as Grubaugh and Santerre (1994) and Greene (2004, 2005), have analyzed international differences in health care system performance.
(9) Total spending includes personal spending plus research, construction, and public health expenditures. All values are expressed in U.S. dollars based on purchasing power parity.
(10) Missing data resulted in four observations that could not be used in the empirical test.
Rexford E. Santerre, Department of Finance, School of Business, University of Connecticut, 2100 Hillside Road, Unit 1041, Storrs, CT 06269, USA: E-mail email@example.com.
I thank Peg Hewitt, research librarian at the Tufts Center for the Study of New Drugs, for helping me locate some online sources with information on new drug approvals. I also thank Richard Goldfinch of the Medicines and Healthcare Products Regulatory Agency in the UK, Nadia Tamminen of the Pharma Industry in Finland, Leila Malkonen of the National Agency for Medicines in Finland, Dominique Leyh of the Federal Agency for Medicinal Products and Health Products in Belgium, and Martin Bernard of Health Canada for graciously responding to my E-mail inquiries and providing me with the necessary data. I also thank the anonymous referees of this journal for their helpful advice on improving this article and the comments of those who participated in the 2008 Graduate Economics Alumni Reunion at the University of Connecticut.
Received August 2009; accepted April 2010.
Table 1. Multiple Regression Results for Medical Care Growth the United States, 1960-2007 Growth of Medical Growth of Drugs Services Per Capita Per Capita Variable Estimated Coefficient (Absolute Value of t-Statistic) Constant 3.684 *** (4.10) -2.341 (1.35) Growth of real out-of pocket own price -0.035 (0.41) -0.324 *** (3.77) Growth of real out-of pocket cross price -0.012 (0.25) -0.012 (0.25) Growth of real net income 0.228 * (1.99) 0.547 *** (3.04) Growth of lagged quantity demanded per capita in the prior year 0.378 *** (3.26) 0.548 *** (4.86) Growth of life expectancy at birth 0.155 (0.21) -5.197 *** (3.98) Growth of percent elderly -0.493 (1.46) 0.878 (1.25) Number of new molecular entity approvals -0.123 *** (4.11) 0.136 *** (2.69) Adjusted [R.sup.2] 0.559 0.596 Growth of Personal Health Care Per Capita Variable Estimated Coefficient (Absolute Value of t-Statistic) Constant 3.186 *** (3.60) Growth of real out-of pocket own price -0.069 (0.93) Growth of real out-of pocket cross price -- Growth of real net income 0.241 *** (2.32) Growth of lagged quantity demanded per capita in the prior year 0.482 *** (4.36) Growth of life expectancy at birth -0.308 (0.45) Growth of percent elderly -0.525 * (1.85) Number of new molecular entity approvals -0.095 *** (3.30) Adjusted [R.sup.2] 0.650 *** Statistically significant at the 1%n level. ** Statistically significant at the 5% level. * Statistically significant at the 10% level. Table 2. Multiple Regression Results for Total Per Capita Health Care Expenditure Growth in Seven OECD Countries (N = 128) Coefficient Estimate (Absolute Value of the Variable t-Statistic) Constant 3.823 *** (2.88) Growth of GDP per capita 0.477 *** (3.00) Lagged expenditure growth 0.255 ** (2.29) Growth of life expectancy at birth 0.091 (0.07) Growth of percent elderly -0.142 (0.41) Number of new molecular entity approvals in the prior year -0.065 ** (2.28) Time and country fixed effects included Adjusted [R.sup.2] 0.306 *** Statistically significant at the 1% level. ** Statistically significant at the 5% level.
|Printer friendly Cite/link Email Feedback|
|Comment:||National and international tests of the new drug cost offset theory.|
|Author:||Santerre, Rexford E.|
|Publication:||Southern Economic Journal|
|Date:||Apr 1, 2011|
|Previous Article:||Trade and tax reforms in a cash-in-advance economy.|
|Next Article:||Market and welfare effects of mandatory country-of-origin labeling in the U.S. specialty crops sector: an application to fresh market apples.|