Medical Workforce

4. Labor Demand Topics

The productivity of the medical workforce appears to vary substantially between individual providers and between organizational forms. Several studies have shown that productivity varies across physicians within markets in the US (Epstein et al., 2010; Phelps, 2000; Welch et al., 1994), Canada (Roos et al., 1986), and Norway (Grytten and Sorensen, 2003). Furthermore, physicians in group practices appear to be more productive than solo practitioners (Reinhardt, 1972; Brown, 1988), although small groups are more productive than large groups (Gaynor and Gertler, 1990; Gaynor and Pauly, 1990).

These differences in labor productivity are persistent, which indicates that rents for labor do not get arbitraged away as they would in a perfectly competitive market. This raises the question of what labor demand factors may explain these persistent productivity differences. After first reviewing traditional empirical studies of factor demand, this section explores a number of possible explanations for the variation in labor productivity, including government regulation, differences in reimbursement incentives, politics, the effect of incentives to manage people within an organization, human resources management and motivated agents.

4.1. Factor Demand: Own- and Cross-price Elasticity Estimates

In this section we review empirical estimates of own- and cross-price elasticities in various health care production settings. These estimates shed light on the nature and flexibility of the production function, how wages in many occupations are likely to respond to shifts in labor supply, and the likely impact of shortages and surpluses. As an example, consider the recent growth in non-physician clinicians (NPCs). There were 70,000 physician assistants in the US in 2007 and 140,000 nurse practitioners in 2004, versus essentially zero of each in 1970 (Wilson, 2008). A recent report on the physician workforce predicts that the number of NPCs will grow by 60 percent in the US between 2005 and 2020, and each NPC will provide 40 percent of the work currently provided by a physician (Health Resources and Services Administration, 2006). The implicit assumptions that NPCs and physicians are substitutes and NPCs face few entry barriers reduce the projected shortage of physicians that would otherwise occur.

Early studies of US physician practices agreed that physicians were not using nonphysician staff efficiently, but disagreed regarding whether they were using too few (Reinhardt, 1972, 1975) or too many (Brown, 1988). Reinhardt (1972) estimates a production function using survey data that contain three output measures for a physician practice (number of office visits; total visits, including hospital and home visits; and annual billings) and factor inputs, including a physician’s hours worked. He concludes that if physicians doubled their use of registered nurses, technicians, and office aides (grouped together as “physician aides”), they could have boosted output by 25 percent. Brown (1988) also estimated a production function using more recent data than that used by Reinhardt. Acknowledging that physician’s choose their hours worked, Brown used a physician’s age, an indicator for an urban practice, and an indicator for having a capitated practice to instrument for physician hours worked. By comparing the marginal product-to-wage ratio between physicians and other inputs, Brown concludes that physicians are overutilizing aides but underutilizing licensed practical nurses (LPNs).

Other authors used cost regressions to examine labor demand in physician practices (Gillis et al., 1991; Pope and Burge, 1995; Escarce, 1996). Escarce and Pauly (1998) comment that the (important) opportunity cost of physician’s time had been omitted from these regressions. US physicians usually own their practices and hire inputs to maximize their income (or some other objective function), so researchers do not observe a physician’s wage. They develop a model where physicians jointly choose their labor supply and other inputs to maximize a utility function consisting of leisure and net income. Escarce and Pauly find that physicians are substitutes with all other inputs combined, and that as their output grows they use non-physician inputs disproportionately relative to their own labor.

Thornton and Eakin (1997) developed the same insight regarding the interdependence of physicians’ labor supply and production decisions. In the first stage physicians choose non-physician labor inputs for every possible value of their own labor input; in the second stage physicians choose the quantity of their own hours that maximizes utility. They find that the demands for assistants (i.e. all non-physician employees) and medical supplies are inelastic, with own-price elasticities of 20.26 and 20.05, respectively. Using a similar approach with 1998 US data, Gunning and Sickles (2011) also find input demand to be inelastic. Specifically, the own-price elasticity estimates for non-physician staff and capital are 20.22 and 20.24, respectively. These results are consistent with a low marginal rate of substitution; the physician services production function is fairly rigid.

There are several empirical challenges when estimating hospital cost functions. First, hospitals produce many different outputs (e.g. emergency room visits, clinic visits, surgical admissions, medical admissions) that cannot be accurately captured by a single variable. Second, patient severity, and therefore cost, is likely to differ substantially between hospitals. Third, although physicians are an important input in hospital production, often there is no observed physician wage (especially in the US where most physicians are paid by health insurers rather than hospitals). Fourth, the quality of labor inputs may differ across hospitals within the broad categories used in regressions, such as nurses or technicians. Vita (1990) estimates a hospital cost function using a flexible functional form (the translog function). He allows for five unique outputs, observes input prices for five inputs, and includes a case mix variable to control for differing patient severity. The demand for nurses (estimated own-price elasticity of 20.34), medical supplies (20.17), and non-physician medical practitioners and technicians (20.75) is inelastic, while the demand for managers and supervisors (21.94) and auxiliary staff (21.27) is elastic. Not surprisingly, non-clinical personnel appear to be more discretionary than clinical personnel.

Gaynor et al. 2011 develop a new output index model to estimate a cost function for 320 California hospitals in 2003. Unlike earlier studies that included counts of five (or so) types of outputs as regressors, Gaynor et al. include detailed data on the diagnoses and services received by 3.5 million patients in their cost function estimation.

The factor inputs with the lowest own-price elasticities (in absolute value) are registered nurses (20.17), medical supplies and equipment (20.35), and clerical staff (20.42). As with Vita (1990), they find that management is the most price elastic input (21.06).

Other studies provide insights into the flexibility of the health production function. As discussed above, Gunning and Sickles (2011) incorporate a physician’s labor supply choice with their factor input decisions to estimate a multi-product cost function. They estimate a cross-price elasticity between non-physician staff and capital (measured by office rent) of 0.45, indicating that these two inputs are substitutes. Doyle et al. (2010) exploit a situation where patients were randomly assigned to clinical teams at hospitals with different quality rankings. Although patient outcomes were not significantly different between the two hospitals, patients at the lower-ranked institution stayed in the hospital longer, experienced more expensive stays, and received more diagnostic tests. This is consistent with the substitution of time (and hospital labor generally, such as nursing time) and diagnostic tests for skill/judgment. Wanchek (2009) finds that dental hygienists are substitutes for dentists and dental assistants, and Anderson et al. (2000) find that homeopaths (i.e. “alternative medicine” providers) and physicians are substitutes. Although various inputs are substitutes, the conclusion remains that the health production function does not respond substantially to differences/changes in input prices. However, grouping different quality inputs into broad categories would attenuate regression coefficients toward zero, so that the true elasticities may be larger in absolute value.

4.2. Reimbursement, Politics, and Government Regulation

In traditional labor demand analyses, firms are assumed to choose labor inputs to either maximize profits or minimize costs conditional on chosen output level. In many countries hospitals are run by the state, either at local or national level, and physicians are employed by the state. If hospitals and physicians deviate from standard profit-maximizing behavior in these less competitive settings, this may provide one explanation for the variation in labor productivity and population health across geographical areas. Furthermore, the way a provider is reimbursed may affect their labor demand. Fee-for-service reimbursement pays practices according to the services they provide to patients. However, services are sometimes defined according to what labor inputs are used. An office visit with a primary care physician, for example, may be reimbursed at a higher rate than a visit with a nurse practitioner even if both have the same effect on the patient’s health, or a visit that involves capital (e.g. MRI exam) may trigger a higher payment than one that does not. Reimbursement can subsidize or tax certain labor inputs unless the rates are set to generate equal profit margins across all possible services. In this section we examine four factors that may affect the demand for labor by hospitals and physician practices: reimbursement methods, the degree of competition, government wage regulation, and politics.

4.2.1. Reimbursement Methods

Schoen et al. (2009) surveyed primary care physicians in 11 countries and asked them, among other things, about the health production function they use. Only 2 and 6 percent of physicians in Sweden and the United States, respectively, indicated that they could benefit financially from hiring additional non-physician clinicians to their practice. In Italy and the Netherlands, by contrast, 44 and 60 percent of physicians responded that they could benefit financially from such hiring. In France, which has a fee-for-service physician reimbursement system, only 11 percent of respondents indicated that they use non-physician staff to manage chronic care, versus 98 percent in the UK where there are strong pay-for-performance incentives. Jacobson et al. (1998) conducted interviews at nine health maintenance organizations and multi-specialty clinics in the US. They found that organizations with a relatively large managed care population, and presumably a large percentage of revenue being generated via capitation, gave nurse practitioners and physician assistants a considerable amount of clinic autonomy. These two studies show that physician practices in countries or organizations where capitation is important have relatively strong incentives to hire physician substitutes and define their responsibilities broadly.

Two economic studies formally examine how reimbursement affects factor demand. Acemoglu and Finkelstein (2008) study the effect of a 1983 Medicare hospital reimbursement rule (the prospective payment, or DRG, system) in the US that increased the price of labor relative to capital. As expected, they find that the capitalto-labor ratio subsequently increased as the policy spurred the adoption of (expensive) medical technologies. Moreover, hospitals responded by increasing their nurse skill mix, presumably because skilled labor is a complement of capital/technology. The same payment policy also provided teaching hospitals with an extra $70,000 for each medical resident they hired. Because the reimbursement formula is based on a hospital’s resident-to-bed ratio, this subsidizes residents and taxes beds. Nicholson and Song (2001) find that hospitals responded to these incentives by hiring more residents, but did not close beds. In fact, the subsidy can explain about 40 percent of the increase in residents between 1984 and 1991.

The studies above indicate that the prevailing reimbursement system has a strong influence on providers’ factor demand. Two studies comparing Canadian and US hospitals, however, come to an opposite conclusion. Hospitals in Quebec are highly regulated: their budgets are determined by prior experiences and political decisions rather than from market forces, and they do not compete directly with other hospitals. US hospitals, by contrast, are less regulated and face more competition. However, Bilodeau et al. (2000) find that Canadian and US hospitals have similar cost functions—both consistent with short-run cost minimizing behavior. This result is consistent with earlier analysis by Haber et al. (1992). They found that labor shares in US and Canadian hospitals in 1985 were very similar despite different environments and incentives. For example, registered nurses represented 45.6 percent of hospital labor in both countries.

4.2.2. Input Wage Regulation

In many sectors of the economy wages are mandated to be very similar across different geographic labor markets. In this case, if the competitive outside wage is higher than the regulated wage, the wage regulation essentially acts as a wage ceiling for medical labor. There are many studies of labor quantity restrictions (e.g. firing costs) and labor price floors (e.g. minimum wages), but this type of wage regulation has been much less studied in economics. Yet it is likely to be important in the health care sector because of the large role for the public sector as a direct employer of labor. Within Europe, such centralized wage regulation occurs in the UK, France, and Spain.

Propper and Van Reenen (2010) advance a simple two-sector, two-region equilibrium model of an occupational labor market to capture the salient features of pay regulation in the nursing labor market and use this to explore how this affects productivity. Their set-up is the following. They consider an economy with two sectors, j 5 {1, 2}, where sector 1 is the “skill-sensitive sector,” which they assume to be the public sector provider of health care. Sector 2 is less sensitive to skill (they propose the nursing home sector, which is much more low-tech than the hospital sector but employs some individuals who are trained as nurses). There are two regions, r 5{L, H}, where L is the low-price region and H is the high-price region. Prices are Pr and PH . PL (e.g. because of ground rental prices due to land scarcity). Region H (L) is the “high (low) outside wage” region because nominal wages are higher in the unregulated sector 2. There are two skill types, s 5{S, U}, where S is skilled and U is unskilled, and with nominal wages Wjs.

They assume that consumers in region r need to be serviced by hospitals located in region r and that producer prices and consumption prices are equal within a region. Unskilled workers’ wages are fixed in the world market and there is an infinitely elastic supply of such workers. The basic premise is that nurses will respond to the wages they are offered and are geographically mobile though there may be some fixed costs of moving which add frictions. Three predictions emerge from this model.

First, if the pay regulation causes wages to fall, nurses are likely to geographically migrate to where wages are relatively higher. The hospitals in areas to which they migrate will benefit in performance through having access to higher-quality human capital. The regulated wage will make the high-price region less attractive to nurses relative to the low-price region. Consequently, some skilled nurses move to the lowprice region, and relative productivity in the high-price region deteriorates. Second, even if they remain in the same region, nurses can move into sectors where their pay is not regulated. This will tend to cause productivity to rise in the unregulated sector as skilled workers switch toward this sector. A third result is that for some plausible values of the regulated wage, we may observe “convexity” in the sense that the deterioration in the quality of public health care in the high outside wage region is greater than the (possible) increase in quality in the low outside wage region. Thus the regulation can decrease aggregate productivity in the public health care system.

They test the predictions of this model using data from the English NHS where wages are regulated centrally. As a measure of hospital quality, they use deaths following emergency admissions for heart attacks (acute myocardial infarction, AMI). The unskilled sector is taken to be the nursing home sector. They find all three predictions of the model are supported. The quality of care in hospitals located in high outside wage areas is lower, the quality of care in nursing homes in high wage areas is higher and, finally, there is non-convexity in the quality effect. The negative effect of regulation on hospital quality is much stronger in the high-cost areas (where regulated wages are much lower than the outside wage) than the positive effect in the low-cost areas (where regulated wages are higher than the outside wage). Thus, the aggregate effect of the pay regulation is to increase aggregate death rates and to strongly reduce social welfare. Essentially any gains from wage regulation in terms of keeping down the wage bill in high-cost areas are more than offset by the extra deaths which arise from low quality.

Nurse shortages are a perennial problem of many nursing labor markets, as discussed in section 3.11. Propper and Van Reenen (2010) do not directly explore the implications of their model for nurse shortages, although a wage shock in the high outside wage region of their model will lead nurses to move to the low-wage region and/or into the unskilled sector, so leaving hospitals in the high-wage region with a nurse shortage. They may fill these with less skilled temporary nurses and this may be one explanation of decreased productivity in high-wage areas. Some evidence to support this is presented in Hall et al. (2008).

Elliott et al. (2007) explicitly focus on this issue and examine the vacancy rate for nurses in the UK NHS in areas of different outside wages. Their analysis uses a large number of small areas but is cross-sectional in nature. They find that a reduction in the wage gap in a local area results in an increase in the long-term vacancy rate for National Health Service (NHS) nurses. The competitiveness of nursing pay has a strong effect on the ability of the NHS to attract and retain nurses. While they do not undertake a full costbenefit analysis, they, like Propper and Van Reenen (2010), conclude that changing relative pay between qualified nurses in different areas of Britain is a mechanism for affecting health authority vacancy rates, and that measures to introduce a greater responsiveness of nurses’ pay to local labor market conditions are required. As discussed in section 3, nurse wage elasticities are consistent with estimates from the general female labor supply literature.

4.2.3. Politics

Clarke and Milicent (2008) also explore the interface between government actions and labor demand in an analysis of employment in French hospitals. French public hospitals are funded and run by local government. Clarke and Milicent argue that left wing local administrations, who are committed to keeping down unemployment, will use hospital employment as a way of dealing with local unemployment, thereby increasing their chances of remaining in office. They test this on a sample of French hospitals and find that hospital employment is consistently higher in public hospitals than not-for-profits or private hospitals. Public hospital employment is positively correlated with the local unemployment rate, whereas no relationship is found in nonpublic hospitals. This is consistent with public hospitals providing employment in depressed areas. They find the relationship between public hospital employment and local unemployment is stronger the more left wing the local municipality. This latter result holds especially when electoral races are tight, consistent with a concern for re-election.

4.3. The Organizational Form of Firms: Why Do Physicians Form Groups?

There are large documented differences in the productivity of physician firms. Physicians in group practices appear to be more productive than solo practitioners (Reinhardt, 1972; Brown, 1988), and small groups are more productive than large groups (Gaynor and Gertler, 1990; Gaynor and Pauly, 1990). There have also been changes in the organization of physician firms. For example, in the US in 1980 the majority of physicians were solo practitioners whereas in 2006 only 32 percent of primary care physicians practiced alone (DeFelice and Bradford, 1997; Bodenheimer and Pham, 2010).

To date, much of the research on partnerships in health care has been motivated by theoretical work in industrial organization and the theory of the firm, which examined the issue of production in teams and the associated potential for moral hazard. It therefore adopts the “design” perspective stresses that different organizational forms are the optimal response to external or internal constraints, so that single and group practices will be optimal in different circumstances.

A significant barrier to well-functioning principalagent relationships between the firm and its constituent members is the difficulty associated with monitoring effort. Holmstrom (1982) showed that where effort is difficult to monitor and output is rewarded at the team level, the larger the team the greater the potential for free riding and shirking by any individual member. This work points to the importance of firm size and homogeneity for efficiency. Groups that are relatively homogeneous will have constituent members whose objective functions do not vary significantly from the “average” member’s objective function. If the group’s welfare function is a simple amalgam of its members’ objectives, then each member’s goals will stand in less serious contrast to those of the firm, which will mitigate (although not completely alleviate) the problems associated with shirking. On the other hand, more heterogeneous firms will have members whose objective function is significantly different from the firm’s; members will then have greater incentive to take actions that are in their interest and not the group’s as a whole.

If shirking is a problem, then firm size may be important. While on one hand economies of scale may encourage larger firms, it is also likely that as the firm grows monitoring of effort by workers becomes more difficult. Smaller groups may have an easier time monitoring and may be able to do so without an explicit and expensive contract (or at least use contracts which are less comprehensive and rely more upon social or peer pressure). In larger groups, it is more likely that shirking problems would require more complete contracts as well as formal monitoring and punishment schemes. In the limit, in a solo practice there is by definition no distinction between the objective function of the physician and the firm—they are one and the same and shirking is not a relevant concept.

Physician firms, because of their relatively small size and the availability of financial and labor input information, have been used to test these ideas (see also the reviews in Scott (2000) and DeFelice and Bradford (1997)). Gaynor and Gertler (1995), for example, examine the relationship between the degree of risk sharing, compensation, and effort among primary care physicians in the US in medical group practice. They specify a model where demand is uncertain, and where physicians choose effort to maximize utility in response to the incentives in the firm’s compensation structure. The utility maximizing level of effort is where the marginal revenue product of effort is equal to its marginal disutility. They derive comparative static results of the effect of changes in internal compensation on the number of patients seen (defined as effort). They also examine the effect of risk aversion on choice of compensation structure. The empirical results found that a stronger link between compensation and productivity leads to more office visits per week (effort), and that the greater the risk aversion of physicians the less strongly the compensation structure is related to productivity.

Encinosa et al. (2007) explore why compensation may not always be linked to productivity in medical groups, using the sociological concept of “group norms” incorporated into an economic framework of risk sharing and multi-tasking. Group norms are defined as the social interactions resulting from comparisons of effort and pay within groups. They demonstrate that group income and effort norms make small groups more likely to adopt equal sharing rules than large groups, and that risk aversion and multi-tasking make equal sharing more likely in large groups. They find evidence that group norms do influence choice of compensation method, in addition to the usual factors analyzed in principalagent models (risk aversion and multi-tasking).

In the same contracting theme, Gaynor (1989) raises the issue of intra-firm competition, i.e. competition between members of the practice. In Gaynor’s model, physicians’ revenue depends in part on their individual productivity. Physicians who produce more visits receive a larger share of the firm’s net revenue. Physicians may attract patients from two sources: they may bring them in from outside the firm or they may attract them away from other partners within the firm. So the incentives of the individual partner may be incompatible with the interests of the group.

Under the design perspective, each firm will choose whatever organizational form is optimal. The question is therefore not whether partnerships or solo practices are more or less efficient, but under what conditions is one form preferable to another. This also has the implication that testing which form in general is more efficient makes little sense because there will be some conditions under which one form is optimal and another set of conditions under which another form is optimal. On the other hand, empirical analyses that relate organizational form to the extent of risk sharing or risk to optimal size of a partnership do have value.

A more recent literature has emphasized the matching of patients, physicians, and firms as a reason for the existence of firms. Theoretical work in industrial organization and organizational economics has established that one reason firms exist is to promote specialization by matching opportunities, consumers in the case of professional services, to workers who have a particular comparative advantage (Garicano and Santos, 2004; Garicano and Hubbard, 2007, 2009). This matching encourages workers to specialize because the returns to specialization are greater when workers have more opportunities to use their specialized skills.

Matching by firms may be particularly important in economic sectors with humancapital intensive production. In professional services markets such as law, medicine, automobile repair, consulting and financial advising, asymmetric information is likely to inhibit the amount of matching in the market. Professionals in these fields acquire considerable human capital. Without this training, consumers may not know the specific nature of their problems nor, once obtaining a diagnosis, are they able to identify the most appropriate professional to address that problem. Likewise, adverse selection due to asymmetric information between professionals also inhibits matching. Professionals who diagnose problems have incentives to refer only the least profitable consumers to their peers, which in turn deters other professionals from accepting these referrals (Garicano and Santos, 2004).

Epstein et al. (2010) test this idea in the context of medical firms. They examine how one type of professional services firm, obstetrics practices, coordinates workers and matches them with consumers. They test two hypotheses from the models of firms as coordinators of human capital. First, do firms overcome asymmetric information and institutional barriers to achieve higher levels of specialization and coordination than occur in the market? Second, does firm specialization improve productivity, especially in ways that benefit consumers?

In their first set of analyses they examine the extent to which physicians specialize in certain medical problems, and whether the amount of specialization differs between solo physicians and those in groups. In the second set of analyses they consider the implications of specialization on aspects of productivity that benefit consumers. To do the latter they exploit the random pairing of workers and jobs that results from obstetricians’ weekend call schedules to overcome endogeneity in the matching of worker and consumer and use this random pairing to develop unbiased measures of each physician’s skills in performing Cesarean section deliveries and vaginal deliveries. They consider whether firm coordination increases productivity by matching patients to workers based on workers’ absolute and comparative advantages.

Their results provide support for both hypotheses. Relative to solo physicians, physicians in group practices are more likely to specialize in treating patients with high-risk health conditions. Furthermore, those in groups who treat such patients treat substantially more of them than solo physicians. They find this despite also observing very similar total caseloads and frequencies of high-risk conditions among group and solo physicians overall. As a result, high-risk patients in group practices are more likely to match with an appropriate specialist than are patients of solo practitioners. On balance, this suggests that obstetrics markets themselves provide some coordination, but firms are able to improve on this.

Huckman and Pisano (2006) also examine matching of firms and workers. In their case, they examine freelancers and ask the question of whether they are more productive when matched with a particular firm. In many settings, firms rely on independent contractors, or freelancers, for the provision of certain services. The benefits of such relationships for both firms and workers are often understood in terms of increased flexibility. Less understood is the impact of freelancing on individual performance. While it is often presumed that the performance of freelancers is largely portable across organizations, it is also possible that a given worker’s performance may vary across organizations if he or she develops firm-specific skills and knowledge over time. They examine this issue empirically by considering the performance of cardiac surgeons, many of whom perform operations at multiple hospitals within narrow periods of time. Using patient mortality as an outcome measure, they find that the quality of a surgeon’s performance at a given hospital improves significantly with increases in his or her recent procedure volume at that hospital, but does not significantly improve with increases in his or her volume at other hospitals. These findings suggest that surgeon performance is not fully portable across hospitals (i.e. some portion of performance is firm specific).

The fastest growing physician specialty in the US is hospitalists, who provide general medical care in the hospital setting only, rather than splitting time between the outpatient clinic/office and the hospital. Meltzer (2001) and Meltzer and Chung (2010) present theories to explain this trend. They argue that the benefits of having a single physician who provides coordinated care in both the clinic and hospital settings used to dominate a model where some physicians would specialize in each sector, and there would be coordination costs when “handing off” a patient between sectors. However, as the transportation costs of switching from one sector to the other increased (e.g. driving across town), the model of specialized human capital became preferred. Using a similar argument, Cebul et al. (2008) suggest that the hospitalist movement may be a method to lower the cost of coordination in an increasingly fragmented health care system. All three of these papers provide support for the use of firms versus solo practitioners.

Other ideas as to why group practices arise include smoothing production and reputation. Physicians supply services that clearly are not amenable to inventory storage. For some specialties, most notably obstetrics and gynecology, production is a lumpy process, with deliveries occurring throughout the day. It may be easier to share such duties within a firm than to write contracts between independent contractors (DeFelice and Bradford, 1997). Getzen (1984) examines the role of reputation. Reputation is an important component of the market for physician services; simultaneously, information used to form reputations may be very costly for consumers to obtain. Getzen argues that larger firms, by concentrating physicians, may be able to take advantage of “reputational economies of scale,” and so support higher fees within consumer search models. In summary, there is no single message from the literature as to the optimal form of organization for physicians. In general, the literature on physician organization has been dominated by a focus on the organizational form as a response to risk sharing and the need to induce effort when it is not fully contractible. This fits more generally with a focus in health economics on responses to financial incentives and the design of optimal contracts. But there has been considerably less work focused on why changes in the organizational form of physician firms have come about and on the role of marketlevel factors, for example competition. In addition, issues in the organization of physicians and their related productivity may benefit from the “technology” perspective on firm organization. This is discussed in more detail in section 4.4 below, but briefly, this perspective does not stress the optimality of any single organizational form, but stresses that frictions and the impact of regulation may leave certain forms existing while they are not the most efficient. For example, thinking of the organization as a form of technology may allow a diffusion perspective, which might help explain why large differences in productivity between different physician firms persist over time and why solo practices continue to exist in many markets where group practices have been shown to have advantages. In addition, the stress in the technology perspective on responses to changes in external factors, for example product market competition or regulation, provides another perspective to explore in order to explain why some forms of organization of medical labor are more productive than others.

In the final two sections, we highlight two areas in labor economics which are receiving increasing attention outside health economics. Both seem to us to be promising areas for research in the economics of labor in health care. The first is human resource management, the second motivated agents.

4.4. Human Resource Management in Health Care

Traditionally, labor economics has focused on the labor market rather than looking inside the “black box” of firms. This has changed dramatically in the last two decades and Human Resource Management (HRM) is a growing field in labor economics. The hallmark of this work is to use standard economic tools applied to the special circumstances of managing employees within companies. HRM covers a wide range of activities. Bloom and Van Reenen (2007) include remuneration systems (e.g. individuals or group incentive/contingent pay), the system of appraisal, promotion, and career advancement, the distribution of decision rights (autonomy/decentralization) between managers and workers, job design (e.g. flexibility of working, job rotation), teamwork (e.g. who works with whom), and information provision. Several of these topics as they arise in the health care field are covered in this volume, although not necessarily labeled as HRM. Since we see this as a promising avenue for future research in health care, we sketch here recent economic approaches to HRM and discuss applications to date in health economics.

Bloom and Van Reenen (2007) argue that, in thinking about the reasons for variations in HRM and productivity, a contrast can be drawn between two possible approaches. The first, which is the now classic approach of Personnel Economics, they label the “design” approach. The view here is that the HRM practices we observe are chosen by a profit-maximizing firm; they are explicit strategic choices of the firm, and variations in HRM reflect variations in the firm’s environment. This perspective is situated within the broader fields of the economics of contracts (see Bolton and Dewatripont, 2005, for an overview) and the economics of organizations (see Gibbons and Roberts, 2008). The key feature of the design approach is that the HRM practices we observe are chosen by firms to maximize profits in an environment that departs from perfectly competitive spot markets. This approach puts the reason for heterogeneity in the adoption of different practices as mainly due to the different environments firms face rather than inefficiencies at the firm level. For example, the technology of an industry will determine why certain labor remuneration practices are adopted and others not. So, for example, changes in technology that allow better monitoring of employee output will allow the introduction of performance-related pay where previously it was not possible. With this perspective, the growth of pay-for-performance (P4P) in health care can be seen as a response to widespread development of better health care performance measurement systems (for a review of recent development in performance measurement in health care see Smith et al. (2009)) and one that should improve productivity in health care.

Bloom and Van Reenen (2007) identify a second approach as “managerial technology.” In this view some aspects of HRM could be considered as a technology or “best practice.” Management is partially like a technology, so there are distinctly good (and bad) practices that would raise (or lower) productivity. This view sees a large role for inefficiencies: firms can persist in their adoption of “bad” technology for some time. While low-productivity firms are selected out over time, there will be some stochastic element to this, so in the steady state there will always be some dispersion of productivity. This managerial technology approach raises the question of why differences in managerial quality persist. Bloom and Van Reenen (2007) put forward several reasons. One is that all technologies have some diffusion curve whereby not all firms immediately adopt them. A second is that there is imperfect competition. With imperfect competition firms can have differential efficiency and still survive in equilibrium. With perfect competition inefficient firms should be rapidly driven out of the market as the more efficient firms undercut them on price. A corollary is that an increase in competition should lead to better management practices and, as a consequence, an increase in productivity. A third is “frictions.” Costs of adjustment are ubiquitous in capital investment and have usually been found for labor, especially skilled labor (see Bond and Van Reenen, 2008, for a survey). Thus, firms facing asymmetric shocks will adjust differentially to their new conditions only slowly over time even if they all have identical adjustment cost technologies. In such an environment, low total factor productivity (TFP) firms will not immediately vanish, as there is an option value to remaining active in the sector.

Within the economics of health care, there has been little interest in managerial practices or the quality of management, although there is a large literature on the response to incentive pay and on the optimal design of incentives in hospitals and physician firms. However, in a recent paper, Bloom et al. (2010) use the methodology developed by Bloom and Van Reenen (2007) to measure management quality in UK hospitals. They find that management quality is positively associated with a range of performance measures used to assess hospital quality, including death rates, financial performance, and staff satisfaction. In addition, they investigate whether managerial quality in the health care system is a function of competition. Exploiting the fact that hospitals in marginal political seats are rarely closed in the UK, they derive an instrument for competition defined by numbers of hospitals located in an area. They find that competition appears to improve managerial quality in UK hospitals. This result accords with findings for firms in the rest of the economy (Bloom and Van Reenen, 2007).

4.5. Motivated Agents

The use of models in which agents display some type of other-regarding preferences is only recently becoming common in the economics literature (e.g. Fehr and Schmidt, 2006). Health economics is probably one exception to this. The role of doctor as an agent has been a recurring theme in health economics and is discussed in McGuire (2000). Here we do not discuss the agency role per se, but focus on recently emerging literature in economics on pro-social behavior and motivated agents. This literature has implications for the design of optimal incentives, the selection of motivated agents and its interaction with monetary rewards, and the optimal organizational form required to exploit such motivations.

In a recent review, Francois and Vlassopoulos (2008) distinguish between extrinsic motivation which stems from the standard pecuniary or other material rewards that an individual may receive from outside and intrinsic motivation which is where an individual pursues actions not because of external rewards but because the activity is valuable in its own right. Two conceptualizations of intrinsic motivation have been used in the economics literature—Impure or Action-oriented Altruism in which the individual receives a “warm glow” from the actual act of contributing to a public good and Pure or Output-oriented Altruism where the individual cares about the overall value of the public good to which he contributes but does not receive a benefit from the direct provision of the good.

Besley and Ghatak (2005) examine impure altruism and its implications for the optimal incentive contract in a moral hazard setting. Their paper studies the provision of optimal incentives in a principalagent model when some agents are driven by pro-social motivations while others have conventional pecuniary motivations. Agents are matched with principals who have “missions.” Missions can be seen as attributes of a project over and above the financial payoff that the project has. The effect of impure altruism is to lower the need for “power” in incentives. The essence of the idea is that when agents match to principals who hold similar missions to them, the agent’s identification with a firm or principal’s “mission” lowers the cost of agent effort and so they require less monetary compensation. This is akin to a compensating differential where the greater the agent’s motivation the less high powered incentive pay needs to be. One corollary is that when agents are motivated, effort will be negatively correlated with incentive pay. This is in stark contrast to the usual setting for incentive pay where incentive pay is used to increase effort. This impure altruism approach suggests that services such as health care may be cheaper to deliver when people have pro-social motivations.

Francois (2000, 2007) uses pure altruism to explain why not-for-profit firms are widespread in the provision of public services. The punch line of his model is that a government bureaucracy or non-profit firms, because they do not have a residual claimant, can obtain labor donations due to the service motivation of their employees when a private firm could not. This suggests that such organizations will produce public services more cheaply. This theory may be one explanation for the widespread presence of not-for-profits firms in health care provision, though it does not explain why for-profits have not been completely driven out of the hospital market.

Few models of altruism have been tested within a health care setting. However, they add to the arguments that in health care worker motivation may outweigh the need for high-powered incentives and that not-for-profits may be an optimal way of organizing medical practice.