Effects of occupational regulations on the cost of dental services: evidence from dental insurance claims.

Journal of Health Economics 34 (2014) 131–143

Contents lists available at ScienceDirect

Journal of Health Economics journal homepage: www.elsevier.com/locate/econbase

Effects of occupational regulations on the cost of dental services: Evidence from dental insurance claims夽,夽夽 Coady Wing a,∗ , Allison Marier b a b

Health Policy and Administration, University of Illinois at Chicago, 1603 West Taylor, Chicago, IL 60612, United States Abt Associates, Inc., 55 Wheeler Street, Cambridge, MA 02139, United States

a r t i c l e

i n f o

Article history: Received 3 February 2013 Received in revised form 29 October 2013 Accepted 10 December 2013 Available online 3 January 2014 JEL classification: I12 I18 J3 J4 K2 Keywords: Occupational licensing Healthcare costs Dental care Quasi-experimental research methods Labor market regulation Health workforce Work tasks

a b s t r a c t In the United States, occupational regulations influence the work tasks that may legally be performed by dentists and dental hygienists. Only a dentist may legally perform most dental procedures; however, a smaller list of basic procedures may be provided by either a dentist or a dental hygienist. Since dentists and hygienists possess different levels of training and skill and receive very different wages, it is plausible that these regulations could distort the optimal allocation of skills to work tasks. We present simple theoretical framework that shows different ways that such regulations might affect the way that dentists and dental hygienists are used in the production of dental services. We then use a large database of dental insurance claims to study the effects of the regulations on the prevailing prices of a set of basic dental services. Our empirical analysis exploits variation across states and over time in the list of services that may be provided by either type of worker. Our main results suggest that the task-specific occupational regulations increase prices by about 12%. We also examine the effects of related occupational regulations on the utilization of basic dental services. We find that allowing insurers to directly reimburse hygienists for their work increases one year utilization rates by 3–4 percentage points. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Occupational licensing has been an important institutional feature of the US health economy for a long time. Friedman and Kuznets (1945) argued that physician licensing regulations were a barrier to entry that pushed the wages and prices associated with medical care above efficient levels. Likewise, in one of the earliest papers in health economics, Arrow (1963) also discussed the implications of physician licensing regulations. Arrow was dismissive of

夽 We thank Morris Kleiner, Surrey Walton, Kevin Stange, Tony Lo Sasso, Bob Kaestner and seminar participants at the Upjohn Institute and the Institute for Research on Government Affairs for helpful comments and suggestions. 夽夽 Research for this article used healthcare charge data resources compiled and maintained by FAIR Health, Inc. The authors are solely responsible for the research and conclusions reflected in this article. FAIR Health, Inc. is not responsible for the research conducted, nor for the opinions expressed, in this article. ∗ Corresponding author. Tel.: +1 3123550652. E-mail addresses: [email protected] (C. Wing), [email protected] (A. Marier). 0167-6296/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jhealeco.2013.12.001

proposals to allow free entry to medical professions. But – like many economists since – he suggested that voluntary certification might be a preferable regulatory framework. He pointed out that licensure does more than simply restrict the supply of the services of the regulated profession: it also reduces the range partial substitutes for existing services, which could be very costly in the long run. Arrow went on to suggest that a system of graded licensing could reduce the pernicious effects of licensing while still maintaining a high standard of quality. He does not elaborate on the proposal, but the logic of the idea is easy to understand. Under graded regulations, people could be licensed to provide subsets of medical services. The regulations would allow overlap in the work tasks that workers with different levels of training and skill could perform. In principle this kind of arrangement could allow more substitution opportunities. It might also make health service product and labor markets more competitive. And it seems likely that graded regulations would make it easier for workers to allocate their time to tasks that make the most productive use of their skills. Changes like these could affect the cost and quality of health services available in the market in ways that might

132

C. Wing, A. Marier / Journal of Health Economics 34 (2014) 131–143

mitigate the distortionary effects of conventional all-or-nothing licensing. In the 50 years since Arrow’s paper, occupational regulations have become pervasive in the United States. In the 1950s licensing regulations covered 5% of workers. By 2006, it was 29% (Kleiner and Krueger, 2008). The health sector is particularly heavily regulated. Members of the conventional health service occupations – physicians, nurses, and dentists – are universally required to hold licenses. And over 76% of non-physician health workers also require a license (Kleiner and Park, 2010). One interesting development is that the graded licensing arrangements that Arrow envisioned are becoming more common. In the dental sector, state governments have expanded the legal scope of practice afforded to dental hygienists. It is still true in every state that dentists and hygienists are required to hold licenses, and that only a dentist may legally perform most dental procedures. But in recent years, licensed dental hygienists have gained the authority to perform a smaller list of basic procedures. The content of the list of services that may be provided by either a dentist or a hygienist varies across states and over time. In some cases, allowing hygienists to perform a service may open the possibility of hygienist-led firms. However, the regulations usually restrict what hygienists are allowed to do with and without the direct supervision of a dentist, which suggests that the overlapping regulatory framework is likely to matter most to firms that employ both dentists and hygienists. Simple economic theory suggests that increasing the independent scope of practice of hygienists should put downward pressure on the prevailing price of dental services that can be produced using hygienist labor. The price effect is plausible whether the regulations are framed as a barrier to the entry of hygienist-led dental service firms, or as a restriction on the production function of firms that combine hygienist and dentist labor inputs to produce dental services. Although the end result is similar, the production function framework is more revealing about the ways that scope of practice regulations might affect market outcomes in the health sector. For instance, regulations might represent monitoring requirements that function as an implicit tax on the use of hygienists. Another possibility is that task limitations are a type of factor de-augmenting technology, which lowers the productivity of hygienists. More broadly, scope of practice regulations may alter the elasticity of substitution between hygienists and dentist in the production process. In each case, the regulations bind when at least some firms are forced to adopt a more dentist intensive production process then they would use in the absence of regulation. The upshot is that scope of practice restrictions – either entry barriers or production constraints – could lead to higher equilibrium prices relative to an unregulated or less regulated environment. In this paper, we study the effects of licensing regulations on the transaction prices of seven basic dental services: prophylaxis, fluoride treatment, local anesthesia, nitrous oxide, sealant application, amalgam restoration, and X-rays. These services are regulated differently across states and we use the variation across service categories, states, and time periods to estimate the effect of the licensing regulations on prevailing prices. We estimated service prices using a large database of private dental insurance claims, and we linked the price data with regulatory information in order to estimate the effects of the regulations using generalized difference-in-differences (DD) and triple differences (DDD) regressions. We found that regulations that constrain the practice authority of hygienists increase the price of basic services by about 12% relative to a counterfactual market in which both dentists and dental hygienists are legally allowed to provide the service.

The main price results stand up to a variety of sensitivity analyses that probe key assumptions related to the method of aggregating prices from the claims data, the correlation structure of the error distributions that provides the basis for statistical inference, and the potential for spillovers across geographical areas. We also found evidence that regulatory changes that give more freedom to hygienists led to increases in the utilization of basic dental services. In particular, we found that when insurers are allowed to directly reimburse hygienists for their services, the proportion of people who utilize dental services within the past year increased by 3–4 percentage points. Our main purpose in choosing to focus on seven specific dental services is that comparing prices in these markets can help us identify the causal effects of licensing regulations on equilibrium prices. The bulk of the paper is devoted to making such comparisons and to ruling out alternative interpretations of our basic findings. However, the dental services that we examine also are important for at least three substantive reasons. First, these seven services may represent the “regulatory margin” for state governments considering changes in the regulatory status of dental hygienists. That is, several state governments have shown a willingness to grant hygienists the authority to perform these seven tasks. But there is little evidence that states are willing to grant them authority to perform very many other services, perhaps because other services are considered too complicated or risky. Our study of the seven regulated services is informative about policies that are feasible options in various states. Second, most dental care is basic dental care. The seven services we study represent a huge fraction of the dental services consumed in the United States each year. In the large database of dental insurance claims that we used in our main analysis, there are over 770 million dental claims spread almost evenly across three years (2005–2007). In each year, almost 40% of the insurance claims involved one of the seven dental services on our list. The social costs of a regulation that increases the price of these services by 12% are likely to be very large simply because the services are widely consumed. There are also wide disparities in dental health and utilization of basic dental services in the US (Mouradian et al., 2000). Regulations that limit the supply of dental services and that generate higher service prices may exacerbate these problems. And in a more general sense, occupational regulations that increase prices run counter to efforts to control rising health care costs in the United States. Third, the graded regulations in place for the provision of dental services could be a useful model for other parts of the health economy. It seems unlikely that many licensed occupations will converted to certified or free entry occupations. But it is possible that expanding the scope of practice of occupational groups with different overall skill levels represents a way to reduce the distortionary effects of licensing regulations. In practice, it may be easier for legislatures to expand the scope of practice of a lower skill occupation than to de-license incumbent occupations. Although scope of practice regulations are relatively common (especially in the health sector), the issue has not received much attention in the economics literature. Our paper offers evidence that graded licensing arrangements can help reduce the distortionary effects of conventional single-occupation licensing regimes.

2. Background 2.1. Economic models of licensing Friedman and Kuznets (1945) and Friedman (1962) are early economic studies of licensing. In that early work, physician


licensing is framed as a barrier to entry that restricts supply and increases wages and other prices relative to a counterfactual competitive market. Friedman (1962) interprets licensure requirements as pernicious regulations that primarily function to secure economic rents for members of the licensed group. The supply restriction or entry barrier model is an intuitive and useful starting point for the analysis of occupational licensing policy. But the approach is limited in that it does not incorporate concerns about provider quality and consumer information. There are two other models of occupational licensing that do incorporate information asymmetries. The first approach explores an adverse selection rationale for licensing. Akerlof (1970) showed that markets in which sellers have more information about the quality of a good than buyers can result in an equilibrium in which only low quality goods are sold in the market. Leland (1979) applied a similar framework to the problem of entry into specific occupations and showed that licensing regulations can improve welfare in some situations. In Leland’s model, problems arise because consumers cannot differentiate between high and low quality service providers. Without differential pricing of quality, the information asymmetry drives out the high quality workers (who have better opportunities in other sectors) and only low quality services are available. Licenses raise the equilibrium quality of the workers in an occupation and that can increase welfare by increasing the extent of the market. Shapiro (1986) takes a different approach by modeling the role of licenses in markets where service quality is unobservable but depends on time varying human capital investments made by the service provider. In this case, moral hazard and not adverse selection is the main source of market failure: service providers may underinvest in human capital. Shapiro models licensing regulations as a restriction on the inputs to the production function and shows that licensing restrictions can be welfare enhancing in some situations. He also shows that licenses provide the largest welfare gains to consumers who place high value on high quality services. Supply restrictions, adverse selection, and moral hazard are the dominant theoretical approaches to analyzing licensing in economics. But all three of these approaches are somewhat divorced from the “consumer protection” rhetoric that is used to justify licensing regulations in more popular debates (Goldstein, 2012; Klein, 2012). Concerns about information and service quality clearly are part of the popular discussion. However, the main issue in popular debates usually is not that a ‘lemons problem’ or ‘moral hazard problem’ is reducing total surplus by limiting the extent of the market. Put differently, most people do not seem to justify licensing out of a concern that dental markets will collapse because people cannot tell dentists from quacks, or because dentists will not make sufficient investments in dental-specific skills. Instead, support for licensing reflects the worry that consumers will occasionally receive low quality care and that the low quality care will have long term and perhaps irreversible consequences. One explanation for the support for licensing regulations comes from evidence that people tend to attach greater value to changes in the risk of losses than to changes in the risk of gains (Kahneman and Tversky, 1979). Most of the regulations we consider in this paper limit the independent scope of practice of dental hygienists and may reduce the amount of competition faced by dentists in providing a particular set of services. Because they are product specific, these regulations do not seem to be intended to combat an adverse selection problem in which high quality workers are discouraged from becoming dentists or dental hygienists. Scope of practice regulations also do not aim to improve effort levels among dental care providers as suggested by a moral hazard model because they are not linked to any specific educational requirement: you do not have to obtain special

133

training in order to perform the specified service (Shapiro, 1986). In some ways, it may be more appropriate to think of licensing regulations as a class of safety regulations like those that restrict the way people can access and use toxic substances or those that impose bans or design requirements on “unsafe” products (Kniesner and Leeth, 2004; Viscusi, 1984; Peltzman, 1975). In this respect, one important concern is whether there really is a difference in the safety records or relevant competencies of the two occupational groups, and whether the occupational regulations actually improve the quality of services received by consumers. 2.2. Service quality and safety Intuitive arguments about the way that licensing regulations ensure that people receive high quality services may be untrue in practice. For example, Kleiner and Kudrle (2000) note that in information based models like the ones presented by Leland (1979) and Shapiro (1986), the effects of licensing regulations on prices and on the range of service qualities available in the market can have off-setting effects on the average quality of services actually consumed in a market. Extreme forms of adverse selection and moral hazard can mean that only low quality services are available in equilibrium. In these cases, regulations can increase average quality consumed because higher quality services are available under the licensing regime. But licenses can also increase prices, which may drive some consumers to reduce their consumption of health services. If lower health service utilization is viewed as an undesirable outcome then licensing may be worse than an unregulated alternative in some situations. Kleiner and Kudrle (2000) argue that licensing regulations may not be an effective strategy for improving the average quality of care received by consumers. But they do not provide any direct evidence that hygienists and dentists provide services of comparable quality and safety. To this end, Wetterhall et al. (2010) evaluated the Alaska Dental Health Aide Initiative (ADHAI), which was designed to increase access to basic dental services by allowing lower skill aides to provide a range of basic dental procedures. The evaluation found that the quality of the services provided by the dental aides was indistinguishable from the quality of the services provided by licensed dentists. Quality effects were negligible for the services covered by the ADHAI, which are mostly the same types of basic dental procedures that we are concerned with in this paper.1 Nevertheless, measuring the quality of health services is a difficult task in part because the definition of high quality care can change over time. For example, in recent years the scientific community has learned more about negative long term effects of cumulative radiation exposure. Health care providers and patients may already be revising their views of the appropriate use of imaging technologies and information changes like these may have implications for design of occupational licensing regulations. 2.3. Benefits of dental services Most people place value on the resolution and prevention of painful dental conditions. However, there is evidence that dental

1 In a similar study in the context of medical care, Mundinger et al. (2000) randomly assigned patients without a primary care provider to receive care from either a physician or a nurse practitioner. After six months there was no difference in overall health status or patient satisfaction and no difference in health service utilization between the two groups. These two evaluations suggest that (within the parameters defined by specific programs) safety and service quality are not major concerns with a movement toward greater use of lower skill health providers. Of course, this does not mean that a much larger expansion of practice rights would be similarly safe.

134


care may have more subtle benefits as well. For example, Glied and Neidell (2010) find that exogenous exposure to fluoridated water during childhood increases the adult earnings of women by about 4%, with the strongest effects for women from low SES families. The mechanism is apparently cosmetic: people with exogenously better looking teeth tend to earn more in the labor market. Hammermesh and Biddle (1994) study the wage effects of more general measures of physical beauty; they find a substantial wage penalty for below average looks and a smaller wage premium for above average looks. These studies suggest that expenditures on even purely cosmetic dentistry may have considerable value to a person’s economic wellbeing. There is also research suggesting that periodontitis (a chronic inflammatory dental disease) is related to cardiovascular disease. The evidence from this literature is complicated. Some studies support a causal relationship between periodontitis and cardiovascular disease, and others suggest that the connection is not causal. Humphry et al. (2008) report a systematic review and meta-analysis of studies of the link between periodontal disease and coronary heart disease. The best studies involve prospective designs. Dental health is measured at baseline for a cohort of people who do not have any existing cardiovascular illnesses. Then follow up data are collected 5–21 years after baseline and researchers assess whether people who had more periodontal risk factors at baseline were more likely to develop coronary heart disease after adjusting for other baseline risk factors. After examining the literature, Humphry et al. conclude that the risk of coronary artery disease at follow up is about 1.24–1.34 times higher among people with markers of periodontal disease at baseline. In a consensus report published in both the American Journal of Cardiology and the Journal of Periodontology, Friedewald et al. (2009) discuss the conceptual mechanisms, which mainly involve the effects of chronic inflammation, that link the two health conditions. They review evidence from key individual studies and from several meta-analyses, and conclude that the available evidence suggests that untreated periodontitis does causally increase the risk of cardiovascular disease. Cardiovascular disease is the leading cause of death for men and women in the United States and was responsible for 25% of all ˜ et al., 2011). One estimate puts the total deaths in 2008 (Minino direct medical costs of cardiovascular disease at $272.5 billion in 2010 alone, and these costs are expected to grow (Heidenreich et al., 2011). The value of ancillary effects of dental care that reduce the burden of cardiovascular disease could be substantial. Recent work by Choi (2011) finds that Medicaid expansions – which lower the effective price of dental care to low income consumers – increase dental service utilization rates. It is plausible that other interventions that generate lower effective prices could also change utilization rates, perhaps in ways that would reduce disparities in dental health and capture ancillary benefits related to cardiovascular health and labor market outcomes.

2.4. Dental service occupations Table 1 provides basic information on the three main occupations involved in the provision of dental services: dentists, dental hygienists, and dental assistants. The wages of dentists are 2.3 times higher than the wages of hygienists and 4.7 times higher than the wages of dental assistants. Across the country, there are about 2 hygienists and 3 dental assistants for each dentist (BLS, 2011). Total expenditures on dental services were about $108.4 billion in 2011 (CMS, 2011). Annual salaries for the three groups come to about $37 billion, which is 35% of total dental expenditures. The simple job descriptions reported in Table 1 are from the O*Net, a database of information about occupations that is maintained by the Bureau of Labor Statistics. The job descriptions are crude, but they imply overlap between the typical tasks performed by dentists and hygienists. The main areas of overlap seem to occur for tasks that are likely relatively high skill activities for hygienists and relatively low skill activities for dentists. In particular, both jobs involve performing basic examinations and providing preventive care. The actual degree of overlap and the ease with which dentists and hygienists can perform these tasks may depend partly on how licensing regulations constrain the activities of dental hygienists. 2.5. Dental occupational regulations In this paper, we focus on a set of seven dental procedures: amalgam restoration, sealants, fluoride treatment, X-rays, cleaning,

Fig. 1. The figure shows the proportion of states in the US that do not allow hygienists to independently perform three particular types of dental services. The services in this graph – Amalgam Restoration, Local Anesthisia, and Nitrous Oxide – are relatively highly regulated in the sense that few states allow hygienists to perform these tasks. The lines trend downwards somewhat over time, indicating that hygienists have aquired greater practice authority. Most of the variation comes from across-state variation in service specific regulations.

Table 1 Basic characteristics of key occupational groups involved in the production of dental services. Occupation

Average hourly wage

Education

Basic Job Description

90,950

77.76

Doctoral or Professional Degree

Dental hygienist

184,110

33.54

Associate Degree or Bachelor’s Degree

Dental assistant

296,810

16.70

Some College (no degree)

Examine, diagnose, and treat diseases, injuries and malfromations of teeth and gums. May treat diseases of nerve, pulp, and other dental tissues affecting hygiene and retention of teeth. May fit dental appliances of provide preventive care. Clean teeth and examine oral areas, head, and neck for signs of disease. May educate patients on oral hygiene, take and develop X-rays, apply fluoride or sealants Assist dentist, set up equipment, prepare patients for treatment, keep records.

Dentist

Employment

Notes: Employment and average wage estimates are from the May 2011 Occupational Employment Statistics report. Educational information and simple job descriptions are from the O*Net database, which is maintained by the Bureau of Labor Statistics.


135

3. Conceptual framework

Fig. 2. The figure shows the proportion of states in the US that do not allow hygienists to independently perform three particular types of dental services. The services in this graph – prophylaxis (cleaning), fluoride treatment, X-rays, and sealant application – relatively lightly regulated in the sense that most states do allow hygienists to perform these tasks. The lines trend downwards from 2001 to 2003, but have been constant since then. Overall, most of the variation comes from across state variation in service specific regulations.

administration of nitrous oxide, and administration of local anesthesia. Dentists can legally perform these procedures throughout the country. Hygienists can independently perform some subset of these services in most states, but the precise list of procedures that hygienists are legally allowed to perform varies across states and over time. Figs. 1 and 2 show the variation in procedure specific regulations from 2001 to 2007. Fig. 1 shows three procedures that hygienists usually are not allowed to perform. Close to 80% of states limit the right to perform amalgam restoration (fillings) to licensed dentists, so the overlap between the two professions with respect to this task is constrained in most states. Likewise, the majority of states do not allow hygienists to administer local anesthesia and only slightly more allow hygienists to administer Nitrous Oxide. In both cases, there is a downward trend: hygienists have been acquiring greater authority to provide anesthetic drugs. Fig. 2 shows the same type of plot for procedures that hygienists usually are allowed to perform. Almost all states allow hygienists to provide teeth cleaning (prophylaxis), fluoride treatment, and X-rays. And only 18% of states do not allow hygienists to perform sealant applications. For these lower skill tasks, hygienists have made substantial in-roads, but the regulatory trend has been flat since 2003. The consequences of scope of practice regulations may not be apparent to a person receiving dental services: across the country, people are apt to receive care from a mix of hygienists and dentists. One of the basic points of our paper is that scope of practice regulations may affect input allocation decisions on the margin and that these distortions may affect the prevailing price of dental services. The wage and employment statistics in Table 1 suggest that at prevailing wages, it would be considerably cheaper to employ hygienists to perform basic services than to employ dentists. However, it is also likely that some of the costs associated with complying with the regulations are borne by the workers themselves. Recent work by Kleiner and Park (2010), for example, finds that the ability to practice independently increases the wages of hygienists by about 10% and decreases the wages of dentists by about 16%. A key goal of the present paper is to gain a better understanding of how the occupational regulations affect outcomes that fall outside the labor market.

The canonical theoretical models of occupational licensing are concerned with explicit barriers to entering a profession – licensing exams, moral standing requirements, and restrictions on the number of schools that can provide specific credentials – and with how these barriers are likely to affect wages, prices, and service quality in the presence of various kinds of market failures. The scope of practice regulations we consider in this paper restrict the work tasks that can be performed by people with different levels of training and skill. These laws are conceptually distinct from explicit occupational entry barriers, although they may have very similar effects on various economic outcomes. To guide our quasiexperimental empirical analysis, we worked with a simple model of the production of dental services. In this section, we sketch the basic elements of the model in three parts. We start with a baseline in which there are no occupational regulations. Then we consider how occupational regulations can alter the production function when they are modeled as monitoring costs, factor specific reductions in productivity, and reductions in ease with which firms can substitute between different types of workers. 3.1. Production without occupational regulation Suppose that Y is a quality constant unit of some dental service that can be produced by combining the inputs of dental hygienists (H) and dentists (D) using constant elasticity of substitution 1/ı

(CES) technology so that Y = [H ı + Dı ] . Assume that the short run supply of both types of labor is inelastic, and that quantities of capital equipment, other types of labor, and other factors of production are fixed. The profit function for the production of Y 1/ı

is b = pY [H ı + Dı ] − wH H − wD D, where pY is the price of the health service and wH and wD are the wages of hygienists and dentists. Firms solve first order conditions and choose a mix of labor inputs that equates the rate of technical substitution with relative wages. Rearranging one further step gives the factor ratio 1/(ı−1) , which is a useful benchmark for examining H/D = (wH /wD ) how occupational regulations may affect the way that firms employ the services of hygienists and dentists in the production of dental services. 3.2. Implicit input taxes Scope of practice regulations may raise the costs of using hygienists to produce dental services by imposing supervision or monitoring requirements. Let s be the cost of complying with the regulations expressed as a fraction of hygienist wages. The new profit function is s = pY [H ı + Dı ] leads to a new factor ratio H/D =

1/ı

− (1 + s)wH H − wD D, which

(1+s)wH wD

1/(ı−1)

. Framed in this

way, regulations that impose administrative, supervision, or convenience costs lead firms to adopt a more dentist intensive production process than they would in the absence of these costs. In other words, for any level of production the costs of complying with the regulations shift production away from the original (unregulated) location on the isoquant. 3.3. Factor de-augmenting technology The framework above shows how the regulations can affect market outcomes through compliance costs that privilege the use of dentists. But the scope of practice regulations could also have a direct effect on the productive efficiency of hygienists. One way to formalize the idea is to frame the regulations as

136


factor de-augmenting technology. Let > 1 be a regulation induced technology parameter that reduces the relative efficiency of hygienists.2 With the regulation in place, the profit function is = pY [H ı + Dı ]

1/ı

becomes H/D = 1/

− (1 + s)wH H − wD D and the factor ratio

(1+s)wH wD

1/(ı−1)

. Since > 1 the regulation

moves production even further away from the original point on the isoquant, leading to a more dentist intensive production process. 3.4. Substitution constraints Scope of practice regulations could also alter the substitutability of the hygienists and dentists in the production process. In the CES framework used here, the parameter ı describes the ease with which firms can substitute between hygienists and dentists. When ı = 1 the firm faces linear technology in which hygienists and dentists are perfect substitutes. In contrast, when ı = −∞ the firm faces Leontief technology in which isoquants are L-shaped and hygienists and dentists are perfect complements that must be used in fixed proportions. Intermediate values of ı allow for isoquants with varying degrees of curvature. By limiting the tasks that hygienists can perform and the conditions under which they can perform them, scope of practice regulations may make it harder for firms to substitute between the two types of workers. Let > 0 be a regulation induced reduction in substitutability. The new profit function is = pY [H ı− + Dı− ]

1/ı−

and the implied factor ratio is H/D = 1/

− (1 + s)wH H − wD D

(1+s)wH wD

1/(ı−−1)

. The

shift to a production process with a lower elasticity of substitution leads to an even more dentist intensive production process. More broadly, reductions in substitutability could amplify the effects of variation in the relative supply of dentists and hygienists that may occur over time or across different geographical areas. 3.5. Implications of the model The stylized model above is useful for thinking about the different ways that occupational regulations may constrain production in settings involving heterogeneous labor inputs. We outlined three mechanisms through which regulations could alter the behavior of firms – monitoring costs, factor de-augmenting technology, and substitution restrictions. All three of these mechanisms imply that the regulations will generate a more dentist intensive production process that is technically inefficient relative to benchmark model where hygienists and dentists are allowed to work in an unrestricted way. Adopting a less efficient input allocation provides one explanation for a supply restriction that could lead to higher prices. Fig. 3 shows a very simple supply and demand relationship. The line S(p) shows a hypothetical supply curve when both dentists and hygienists are eligible providers. The line S(r) shows the supply curve when the scope of practice regulations lead to a restriction in supply. With a fixed demand curve, the equilibrium price is higher when the regulations are in place and the quantity of dental services consumed is lower. The price effect in the graph is P(r) − P(p), which is the difference in equilibrium prices that prevail under the alternative regulations. In theory, the magnitude of the effect will vary with the size of the supply restriction

2 The positive value of may seem counterintuitive given that the idea of the parameter is to reduce the productivity hygienists. But the technology is expressed relative to dentists workers. In a model where the two types of workers are perfect substitutes (ı = 1) then H = D so that > 1 hygienists are needed to match the work of a single dentist.

Fig. 3. A simple depiction of supply and demand for services under alternative licensing policies.

and will also depend on the price elasticity of the demand and supply curves. Since these characteristics could vary across markets for different dental services in different geographical locations and different time periods we should not expect the effect to be constant. 4. Quasi-experimental design 4.1. Notation With the theoretical framework in mind, we take a quasiexperimental approach to the empirical analysis. To clarify the identification strategy and various potential sources of bias we adopt a potential outcomes notation that is common in the treatment effects literature. To this end, let Y(1)pst and Y(0)pst be a pair of potential outcomes that represent the equilibrium prices that would prevail in the market for product p in state s at time t under alternative regulatory regimes. Y(1)pst is the equilibrium price in the market if regulations require that only dentists may provide the service. Y(0)pst is the equilibrium price in the same market if the regulations permit either a dentist or a dental hygienist to perform the service. The market specific effect of the regulation is ˇpst = Y(1)pst − Y(0)pst . The market level effect of the regulation is not directly observable because one of the two potential outcomes will always be counterfactual. Let Rpst = 1 if the market is subject to a ‘dentist only’ regulation. Rpst = 0 if both occupations are allowed perform the service. Then the observed price is Ypst = Y(1)pst Rpst + Y(0)pst (1 − Rpst ), which we can rewrite as Ypst = Y(0)pst + ˇpst Rpst . To make progress on the identification problem, we follow the quasi-experimental literature by focusing on the average effect across those markets that are actually regulated: E[ˇpst |Rpst = 1]. 4.2. Key sources of bias If regulations were randomly assigned across a well-defined population of dental markets then a simple comparison of mean prices among regulated and partially deregulated markets would provide a credible estimate of the average effect of the regulations. In practice, regulations are not randomly assigned: some dental services may be more likely to be regulated than others. Likewise, some geographical areas may be more open to partial deregulation


than others. And regulatory changes may also be more common in some time periods than others. If the differences in regulations that are in place across markets is systematically related to other factors that affect prices in those markets then simple comparisons of mean prices in groups of regulated and partially deregulated markets will yield biased estimates of the effect of the regulations. These selection problems make it likely that E[Y(0)pst |Rpst = 1] = / E[Y(0)pst |Rpst = 0] so that the regulated markets would have different average prices than the deregulated markets even in the absence of the regulations. The three most important sources of selection bias involve differences in service complexity, differences in the demand for high quality dental services, and secular trends in dental service prices and regulations. 4.2.1. Service complexity A study based on comparisons of the prices of regulated and unregulated dental services could be misleading if the regulated services tend to be more complicated and difficult to perform than the partially deregulated dental services. For example, it may be more common for states to partially deregulate the provision of prophylaxis (teeth cleaning) than amalgam restoration (fillings) because amalgam restoration may require more skill to perform than prophylaxis. Differences in service complexity across regulated and partially deregulated services could explain observed price differences even in the absence of a regulatory treatment effect. 4.2.2. Demand for dental services Another way to study the effects of the regulations would be to compare prices for the same dental service across states with different regulations. This method would avoid bias from service complexity. However cross-state comparisons could lead to biased estimates if there are differences in levels of demand for high quality or safe dental services across the states. Higher demand could drive up prices and could also provide support for occupational licensing regulations that are intended to promote high quality dental care. This relationship between demand for dental services and the demand for regulation could lead to higher prices in regulated states even in the absence of a regulatory effect. Cross state differences in the supply of dentists and hygienists could lead to similar types of bias. 4.2.3. Secular trends One final straightforward way to study the price effects of regulations is to compare prices for the same goods (perhaps even in the same state) before and after changes in relevant occupational regulations. Here the problem is that regulatory changes might be confounded with secular trends in the prices of dental services. 4.3. Difference in differences To cope with the sources of bias described above we estimate generalized difference in difference (DD) regression models (Wooldridge, 2005; Blundell and Dias, 2009; Bertrand et al., 2004). To understand how the standard model applies to our study, suppose that p is a vector of characteristics associated with product p, including measures of the complexity of providing the service. Similarly, let s and t be vectors of characteristics associated with state s and time period t, respectively. Suppose that if we could perfectly observe the information contained in these three vectors, it would be credible to estimate the effect of the regulation using the prices observed in regulated and partially deregulated markets that were matched on levels of p , s , and t . The identifying assumption in that scenario is that E[Y(0)pst |Rpst = 1, p , s , t ] = E[Y(0)pst |Rpst = 0, p , s , t ], which simply means that the distribution of prices that

137

would prevail in the absence of regulation is mean independent of actual regulatory status after conditioning on the values of p , s , and t . In practice, we do not have sufficient data on the values of the product, state, and year characteristics to implement the matched comparison design in a convincing way. However, the panel structure of our data allows us to account for these unobserved characteristics by including a full set of product, state, and year fixed effects in a regression of prices on the regulation indicator. The specific regression equation that we estimate is: Ypst = ˇRpst + p + s + t + εpst In the model, εpst is an error term, ˇ = E[ˇpst |Rpst = 1], and p , s , and t are “differenced” out of the model using vectors of product, state, and time period dummy variables. In other words, the assumptions supporting the model are sufficient to identify the average effect of the regulations across markets that actually are regulated (Blundell and Dias, 2009).3 4.4. Triple differenced models The DD model provides unbiased estimates of the effect of the regulation under the assumption that the deregulated potential outcome prices are mean independent of the presence of the regulations, conditional on product, state and, year fixed effects. The key remaining threats to the internal validity of the design stem from unobservable factors that could affect both prices and regulations and that are not differenced away by the product, state, and year fixed effects. In our data the regulations vary across service categories, states, and years. This gives us some latitude to implement even more general research designs that can account for unobserved confounders that vary at the state × year, year × product, and product × state level. Some of these higher order types of confounders are more plausible than others. For instance, a state × year confounder requires a state specific and period specific event that could explain both a change in dental service prices and a change in dental regulations. A year × product confounder could arise if technological changes make it easier or cheaper to perform a particular dental service and these changes encouraged deregulation in the associated dental markets. A state × product confounder that somehow generates different prices and regulations for a single dental service in a particular state is more difficult to imagine. Nevertheless the data underlying our analysis technically permit estimation of a generalized version of a triple differenced (DDD) regression model: Ypst = ˇRpst + p + s + t + ıps + ıst + ıpt + εpst , Here ıps , ıst , and ıpt represent vectors of variables that vary at the product × state, state × time, and product × time levels. The interpretation of the effect parameter ˇ is the same as it is in the simpler DD model but here the identification strategy is robust to a more elaborate class of unobserved shocks that may be correlated with both regulations and prices. In essence, the model expands the conditioning set of variables that are in place to establish the conditional mean independence condition.

3 Under the additional assumption that E[Y(1)pst |Rpst = 1, p , s , t ] = E[Y(1)pst |Rpst = 0, p , s , t ] the same effect parameter can be interpreted as the overall average treatment effect of the regulations. Intuitively the addition of this second assumption rules out the possibility that regulated markets are selected with reference to their idiosyncratic response to the regulations. Blundell and Dias (2009) provide a helpful review of the identification conditions considered here.

138


5. Data 5.1. Insurance claims data The core of our analysis is based on a large database of private insurance claims maintained by FAIR Health, Inc, which is a non-profit firm that provides independent estimates of the distribution of charges for health services across the United States. The database consists of de-identified individual insurance claims provided by a large set of contributing insurers and third party administrators who operate in markets across the United States. Each contributing company agrees to submit, at regular intervals, a complete and unadulterated dataset of the insurance claims it processed over some defined period of time.4 We use three years of the dental insurance claims database, which consists of insurance claims for approximately 250 million dental procedures each year.5 For each claim, the database contains information on the Current Dental Terminology (CDT® ) code6,7 associated with the service, the zipcode of the location where the service was provided, the amount of money billed by the provider, and the amount of money reimbursed or “allowed” by the insurance company for the service. We used the procedure codes to identify claims for services affected by the seven service specific regulations that are the topic of this paper.8 These 21 codes accounted for 39% of the approximately 750 million claims in the full dental insurance claims database. As mentioned above, the claims data contain two variables that could be used to study prices: the billed amount and the allowed amount. The billed amount is the provider’s list price for the service. The allowed amount is the amount of money that the insurance company reimbursed the provider. In many cases, insurance companies may negotiate discounted prices by offering membership in a network of preferred providers. In this paper, we use the allowed amount information to measure prices because it seems to provide the closest approximation to the transaction price of the service. We removed claims with missing or corrupted information for the billed or allowed charge variable, and we also removed claims with very large prices that were likely data transcription errors – the charge has extra zeros or slipped decimal places. These steps affected a very small fraction of the cases. Because the occupational regulations that are the focus of this paper vary at the state level we aggregated the claims data to

4 A contributor is an organization that has made a contractual agreement to submit claims to the database; the structure of the insurance industry in the United States means that these companies may be affiliated with a larger parent company and so it may or may not be reasonable to think of each contributor as an independent company. 5 There were 157 contributing companies in 2005, 137 in 2006, and 125 in 2007. Changes in the number of contributing firms may reflect the consolidation of contributions made by multiple subsidiary companies but it is difficult to be certain because the identity of the companies is masked in the database for confidentiality reasons. 6 Current Dental Terminology (CDT® ) codes are developed and maintained by the American Dental Association. Each CDT® code is accompanied by a brief description of the procedure; CDT® codes tend to be grouped by procedure. 7 CDT® is a registered trademark of the American Dental Association. 8 Some of the covered procedures categories involve multiple CDT® codes. For instance, patients may be billed for complete series X-rays, single film X-rays, panoramix X-rays, etc. We were interested in CDT® codes for services the following classes of services: (1) prophylaxis, (2) fluoride treatment, (3) local anesthesia, (4) nitrous oxide, (5) sealant application, (6) amalgam restoration, and (7) X-rays. We compiled a list of CDT® codes that fell into each of these seven categories and then selected these claims from the overall database. Specifically, we extracted claims for the following 21 CDT® codes: D0240 D0250 D0260 D0272 D0274 D0340 D0350 D1110 D1120 D1203 D1204 D1351 D2140 D2150 D2160 D2161 D9210 D9211 D9212 D9215 D9230.

Table 2 Average transaction prices by service type and year across all state markets. Dental service

Statistic

2005

2006

2007

Prophylaxis (cleaning)

Mean SD

52.16 12.74

54.2 12.96

57.06 13.9

Fluoride treatment

Mean SD

24.85 5.76

25.9 5.8

24.97 4.52

Amalgam restoration

Mean SD

108.94 31.94

113.33 34.23

121.15 36.21

X-ray

Mean SD

34.57 19.03

35.86 20.27

37.97 21.12

Sealant application

Mean SD

34.09 5.42

35.86 5.73

37.97 6.16

Anesthesia

Mean SD

33.81 70.92

31.3 32.34

30.8 23.04

Nitrous oxide

Mean SD

35.51 8.23

37.65 9.29

40.77 11.56

Note: We estimated the median allowed charge from insurance claims within state × year × service code cells. The table reports the average and standard deviation of these cell medians across all of the states in the sample. Each cell receives equal weight in the calculation.

the state × procedure × year level (Amemiya, 1978; Angrist and Pischke, 2009; Bertrand et al., 2004). The main outcome variable we use in our sample is the median allowed charge for a specific procedure code within a given state and year. We use the state median as the estimate of a typical price because price distributions are sometimes skewed and the median provides a more robust measure of a typical price. Analysis based on mean prices in state × year × procedure code cells produces nearly identical results. Table 2 reports across state averages of the median price in each of the seven product categories for the years 2005–2007. These are averages calculated from the aggregated data set and they give equal weight to each state. Across states, the average price received by providers is about $113 for amalgam restoration (fillings), $55 for prophylaxis (teeth cleaning), $25 for fluoride treatment, and between $30 and $40 for X-rays, Sealant Restoration, Local Anesthesia, and Nitrous Oxide. 5.2. Dental regulations The regulations we use to define a market as regulated or partially deregulated vary across states, years, and service categories. Some of the service categories are more likely to be regulated than others; in fact, there is much more variation in regulatory status across products and states than there is over time. We have regulatory data for the years 2001–2007 but we only have insurance claims data from 2005 to 2007. As discussed earlier in the paper, Figs. 1 and 2 plot the proportion of states in which only a dentist may perform a particular dental service over time. At least two important points are clear in Figs. 1 and 2. First, there is substantial variation in the regulatory status of specific dental services across states and there is also substantial variation in the regulatory status of different types of dental services. Second, there is some variation in the regulatory status of the different products over time but not much of this variation occurs during the 2005–2007 window for which we have insurance claims data. Variation across states and product categories is the main source of identifying variation in our analysis. 5.3. Additional data sources Our main analysis is based on the insurance claims and regulatory data. In order to assess the robustness of the results

C. Wing, A. Marier / Journal of Health Economics 34 (2014) 131–143 Table 3 Effects of licensing regulations on the cell median prices. Statistic

1

2

3

4

5

Regulation effect SE p

6.73 2.74 0.01

6.76 2.46 0.01

6.8 2.51 0.01

6.8 2.53 0.01

2.67 1.94 0.08

No Yes Yes Yes No No No

Yes Yes Yes Yes No No No

Yes Yes Yes Yes Yes No No

Yes Yes Yes Yes Yes Yes No

Yes Yes Yes Yes Yes Yes Yes

0.643 2979

0.643 2979

0.645 2979

0.651 2979

0.703 2979

Covariates Product fixed effects State fixed effects Year fixed effects State × year fixed effects Product × year fixed effects Product × state fixed effects R squared N

Note: Standard errors are calculated using a Huber–White robust variance matrix that allows for clustering at the state level.

Table 4 Effects of licensing regulations on the log of cell median prices. Statistic

1

2

3

4

5


0.12 0.03 p < .01

0.12 0.03 p < .01

0.12 0.03 p < .01

0.12 0.03 p < .01

0.07 0.03 p < .01






0.645 2979

0.646 2979

0.650 2979

0.651 2979

0.695 2979

Covariates Product fixed effects State fixed effects Year fixed effects State × year fixed effects Product × year fixed effects Product × state fixed effects R squared N


with respect to key assumptions and to examine some additional research questions, we also made use of two additional sources of information. We used a commercial database provided by zipcodes.com to construct measures of the demographic, economic, and geographic information about each zipcode in the United States. We used the latitude and longitude of the centroid of each zipcode area to calculate the distance of each zipcode in each state to various state borders, which allowed us to study the effects geographical spillovers on the effects of the regulations.9 We also used data from several waves of the Behavioral Risk Factor Surveillance Surveys (BRFSS) to examine the effects of the regulations on the utilization of dental services. The BRFSS data and the utilization analysis are explained in more detail later in the paper. 6. Main results

139

the regression implies that the regulations increased prices by $6.73. The log-linear model implies that the regulations increase prices by about 12%, which is consistent with the level price effect in dollar terms. A price effect of 12% is substantial given that these basic services represent almost 40% of all of the dental insurance claims filed in a given year. However, a price effect of 12% does not seem implausible given the large wage differential between dentists and hygienists: Table 1 shows that dentists earn about $78 per hour and hygienists earn only $34 per hour. We estimated the standard error of the effect estimates using a Huber–White robust variance matrix that allowed for clustering at the state level. In both the level and log-linear models the effects were large enough to reject the null hypothesis that the regulations have no effect on prices at conventional levels of statistical significance (p < .05). Column 2 reports estimates from the models that add state level time varying covariates to the model; the estimated effects of the regulation are virtually identical. To account for the possibility of state specific temporal shocks that affect both regulations and prices, column 3 reports estimates of regulatory effects using models that include state × year fixed effects. Adding these state × year specific effects does not change the results in a meaningful way. Column 4 reports results that allowed for state × year fixed effects and also product × year fixed effects, which are meant to capture any product specific technological changes that might explain a change in prices. Again the effects are nearly identical whether the outcome is in level form or log form. Finally, Column 5 reports estimates from the fully saturated DDD model, which adds state × year, product × year, and state × product fixed effects to the standard DD model. In the level outcome model, the effect estimate falls from $6.80 to $2.67 and is no longer statistically significant. In the log-linear model the effect falls from 12% to 7% and remains statistically significant. Adding state × product fixed effects to the regression model dramatically reduces the amount of regulatory variation that can be used to estimate the model parameters. Figs. 1 and 2 show that much of the variation in regulations comes from the fact that the states regulate a different list of products and nearly all of the variation in the regulation variable is removed by the state × product fixed effects. This is the main reason why the DDD specification yields a smaller and noisier effect estimate. Given the lack of regulatory variation in the full DDD specification it is sensible to ask whether a state × product specific fixed effect is an important threat to the validity of our analysis. We find it difficult to come up with a plausible explanation for product specific technological changes that occur only in a specific state and that would affect both the regulations and prices in the state. Accordingly, we interpret the results in Tables 2 and 3 as evidence that the occupational regulations increase prices by about 12%. 7. Further analysis 7.1. Statistical inference

Table 3 reports estimates of the effects of hygienist scope of practice regulations on the price of basic dental services from five regression models. In each model, the dependent variable is the median price in the state × year × procedure category cell. Table 4 reports results from the same specifications but the dependent variable is the log of the median price in the state × product × year cell. Column 1 reports estimates from the generalized DD regression, which included a full set of state fixed effects, year fixed effects, and product category fixed effects. In the level effect model,

9 Professor Thomas Holmes of the University of Minnesota generously provides the state border coordinates data on his website Holmes (2013).

Several authors have shown that nominal standard error estimates often are too small in DD regressions because the effects of clustering and serial correlation are amplified by aggregate level regressors (Moulton, 1990; Bertrand et al., 2004; Donald and Lang, 2007; Conley and Taber, 2011). Tables 3 and 4 report standard errors that were estimated using a Huber–White robust variance matrix that allowed for clustering at the state level. Technically, these standard errors are consistent provided that there are a sufficiently large number of clusters and so the question is whether 51 clusters are enough to rely on the asymptotic approximation. An alternative interpretation suggests that the important source of clustering in our work occurs at the product level. If product

140


level clustering is central, then with only 7 product groups the asymptotic approximation is not a convincing basis for statistical inference. To explore the robustness of our results to assumptions about the structure of the error distribution, we implemented a randomization test based on placebo laws. In the most basic implementation of the idea, we randomly selected a set of state × product × year cells, defined them as “pseudo regulated markets”, estimated the regression model using the placebo regulations instead of the real regulations, and stored the estimated coefficient on the pseudo regulation. We repeated this process 2000 times to build a distribution of placebo effects. On average, these placebo laws should have no effect on prices because they are simply randomly chosen cells. However, some of the placebos will lead to large effects by chance. We can understand the likelihood that the effect we estimated was observed by chance without appealing to the asymptotic sampling distribution of a given estimator by comparing the regulatory effect produced by the actual regulations to the empirical distribution of placebo regulation effects. A version of this test led to inferences with the desired statistical properties in Bertrand et al. (2004). Rosenbaum (2002, 2009) also argues that randomization tests have good properties in a variety of non-experimental settings. The test does not require assumptions about clustering and serial correlation; instead it hinges on the assumption that the regulation variable is independent of the potential outcomes conditional on the covariates included in the model. Since we were already imposing this assumption as part of the identification strategy, the placebo laws test seems to only weaken the assumptions underlying our analysis. One conceptual problem that arises with implementing a placebo test is that the true “law generating process” is unknown. To assess the implications of alternative law generating processes on the distribution of placebo outcomes, we computed randomization tests that constrained the placebo laws to arise within the same states as the real laws, within the same product groups as the real laws, and within the same years of the real laws. Note that in each case the placebos are permutations of the existing laws so that in the full sample the total number of laws stays fixed and the identification of the model does not fail for any particular arrangement of laws. Fig. 4 plots kernel density estimates of the placebo law effects. The distributions vary somewhat across the four different law generating processes; however the means are approximately zero, which implies that the effect estimator is unbiased. The vertical line on the far right of Fig. 4 represents the effect estimate – $6.80 – that we actually observed in our data. The observed estimate of the regulatory effect falls in the extreme tail of the distribution of placebo effects, which suggests it is unlikely that the effect was observed due to chance. Based on the placebo law distribution, the probability of observing an effect of $6.80 when the true effect is 0 is well below .05. Thus, the placebo laws tests cast doubt on the possibility that our conclusions are invalid because of biased standard errors. 7.2. Geographical spillovers A literal interpretation of the panel data fixed effects models that we have estimated treats the state × product × year cells as strict compartments that do not interfere with each other in any way. The formal assumption is sometimes called the Stable Unit Treatment Value Assumption (SUTVA) and it is used to rule out spillover effects, general equilibrium effects, and peer effects. In a strict sense, this assumption is quite unrealistic in the current application because there are a variety of ways that regulations applied to one market could lead to price effects in other markets. One of

Fig. 4. The graph shows a kernel density plot of the distribution of 2000 placebo estimates of the effects of the regulation on cell level prices. The different lines show the distributions that arise from different law generating procedures. Each regression allowed for state, year, and product fixed effects. The estimate observed in the actual data is indicated by the vertical red line at about $6.80. The graph shows that an estimate as large as $6.80 is extremely unlikely to have arisen by chance. The placebo law distribution provides a way to conduct statistical inference under less restrictive assumptions than standard methods such as cluster-robust standard errors.

the most obvious scenarios involves cross-border shopping or the threat of cross-border shopping for dental services. Consumers who live close to a border that marks a change in the regulatory environment could avoid higher prices by crossing the border to purchase dental services. In essence, these consumers will have more price elastic demand for dental services and this elasticity will limit the ability of dental providers to pass on regulatory costs to consumers. This type of dynamic should attenuate the effect of the regulation on prices. To investigate the problem, we studied the effects of regulations in metropolitan statistical areas (MSAs) that cross state borders. Specifically, we aggregated the claims data to the MSA × state × product × year cells using only claims for dental services that were performed in geographical areas that were less than 20 miles from the state border. The identifying assumption in this model is that unobservable price trends on either side of the border are captured by MSA, year, and product fixed effects and do not bias the regulation coefficient. To control for additional variation, product × MSA and year × MSA fixed effects are included, along with state level demographic control variables. The results from regressions based on the level price are in Table 5. Estimates based on log prices are in Table 6. The estimates from the level price regressions imply that the regulations increase prices by about $3 within MSAs when the model accounts for product fixed effects, MSA fixed effects, and year fixed effects. These estimates are sufficient to reject the null of no effect at conventional levels of significance. The models based on log prices produce estimates that are very similar in magnitude to the level price estimates but they are less precisely estimated and are not statistically different from zero. In both the level and log models, the effects become quantitatively and statistically insignificant when product × MSA and MSA × year fixed effects are included in the model. As previously discussed, product × MSA shocks that affect both regulations and prices seem like an implausible source of confounding. Accordingly, our preferred estimate comes from the specification in Column 3, which implies that the regulations increase price by about 3% within the same MSA. The MSA regressions provide a way of more directly controlling for unmeasured differences between regulated and unregulated

C. Wing, A. Marier / Journal of Health Economics 34 (2014) 131–143 Table 5 Effects of licensing on the price of dental services within MSA × year × service type cells. Statistic

1

2

3

4

5


3.53 1.35 0.01

3.53 1.17 0.01

3.48 1.16 0.01

0.32 1.62 0.85

0.31 1.64 0.85






0.742 3741

0.745 3483

0.747 3483

0.769 3483

0.771 3483

Covariates Product fixed effects MSA fixed effects Year fixed effects Product × year fixed effects Product × MSA fixed effects MSA × year fixed effects R squared N


Table 6 Effects of licensing regulations on log prices within MSA × year × service type cells. Statistic

1

2

3

4

5


0.027 0.022 0.23

0.033 0.021 0.14

0.033 0.022 0.14

-0.008 0.03 0.80

-0.008 0.031 0.81






0.706 3741

0.709 3483

0.710 3483

0.737 3483

0.739 3483

Covariates Product fixed effects MSA fixed effects Year fixed effects Product × year fixed effects Product × MSA fixed effects MSA × year fixed effects R squared N


geographic areas because the analysis is limited to coherent economic zones that are bisected by a state border that marks a change in licensing regulations. The smaller effects are consistent with the idea that the possibility of cross-border shopping attenuates the effects of the regulations. The finding that the regulatory effects persist even in these small areas that facilitate shopping around adds support to our state level analysis. Taken at face value, the analysis implies that the “true” effect of regulations that were applied uniformly across the country would actually be larger because they would no longer be diminished by cross-border shopping effects. 7.3. Effects on utilization Our analysis thus far provides evidence that allowing hygienists to provide more dental services reduces the prices of those services. A natural follow-up question is whether the regulations also led to an increase in the use of dental services. To examine this question, we pooled data from the 2002, 2004, and 2006 waves of the Behavioral Risk Factor and Surveillance Survey (BRFSS). In these three years, the BRFSS asked dental service utilization questions in every state. Respondents were asked: “how long has it been since you last visited a dentist or dental clinic for any reason?” And also: “how long has it been since you last had your teeth cleaned by a dentist or a dental hygienist?” We constructed four dichotomous measures of dental utilization that indicated: (i) whether the person had visited a dentist or dental clinic in the last 12 months, (ii) whether the person had visited a dentist or dental clinic in the last

141

24 months, (iii) whether the person had had his/her teeth cleaned by a dentist or hygienist in the last 12 months, and (iv) whether the person had his/her teeth cleaned by a dentist or hygienist in the last 24 months. Since these variables do not map directly to the seven service specific regulations, we were not able to estimate utilization models that exactly paralleled our price analysis. To make progress, we defined summary measures of the restrictiveness of the licensing environment for hygienists in each state and estimated models with the following form: Yist = Rst ˇ + Xist + t + ˛s + εist In the model, Yist is one of the four measures of dental utilization, Xist is a vector of individual covariates including measures of age, age squared, race-ethnicity, gender, education, household size and composition, employment status, and bracketed income categories. t and ˛s represent year and state fixed effects, and Rst is a summary measure of the regulatory status of state s in year t. We examined three summary measures of the regulatory environment. The first summary measure is the number of the seven dental procedures examined in our price analysis that a hygienist was legally allowed to independently perform in the specified state and year. The second variable is an indicator set to 1 if private insurance companies were allowed to directly reimburse hygienists for their services rather than pay for their services through a dentist. The third summary regulation variable is an indicator set to 1 if the state Medicaid program was allowed to directly reimburse hygienists for their services. These last two indicators are meant to capture a broader notion of the independent practice authority of dental hygienists. In the model, the regulatory effect – ˇ – is identified by within state changes in the underlying regulations that happen between 2002 and 2006. That is, the analysis is a generalized difference in differences design in which some states change their regulations over this time period and some do not. The strategy accounts for confounders that are fixed at the state level over time and time trends that are fixed across states within years. The final column of Table 7 reports which states actually changed each regulatory variable between 2002 and 2006 period: 12 states changed the sum of tasks index, 4 states changed the insurance reimbursement regulation, and 4 states changed the Medicaid reimbursement regulation. We estimated the models using least squares regressions with robust standard errors that allowed for clustering at the state level (Angrist and Pischke, 2009). The results are in Table 7. The first panel shows estimates of the effects of allowing hygienists to perform more tasks on the probability that a person uses dental services. The estimates are quantitatively small and are not precisely estimated. We find no evidence that increasing the number of tasks affects dental service utilization. The story is somewhat different when reimbursement restrictions are used to measure the economic independence of dental hygienists. The second panel in Table 7 shows that allowing insurance companies to directly reimburse a dental hygienist increases the probability that a person received dental care in the last 12 months by about 2.8 percentage points (p = 0.055). There is no effect on utilization in the past 2 years, which is sensible because the model is identified using within state regulatory changes that should not have affected past behaviors. The reimbursement policy increases the probability that a person has had his/her teeth cleaned in the last 12 months by 4.3 percentage points (p = 0.004). The larger effect on teeth cleaning relative to overall visits also makes sense because teeth cleanings are one of the main services provided by dental hygienists. The estimated regulatory effect on teeth cleaning in the last two years is effectively zero. The estimates based on

142


Table 7 Effects of hygienist licensing restrictions on utilization of dental services. Regulation

Statistic

Dental visit in last 12 months

Dental visit in last 24 months

Teeth cleaned in last 12 months

Teeth cleaned in last 24 months

States that changed each regulation measure 2002–2006

Number of tasks hygienists are allowed to perform

Coefficient SE p N

−0.006 0.012 0.593 91,169

0.003 0.006 0.685 91,169

−0.013 0.014 0.351 83,086

0.005 0.012 0.697 83,086

MI DC RI CT FL ME MN WV OH WY ND SC

Insurance companies may reimburse hygienists

Coefficient SE p N

0.028 0.014 0.055 91,169

0.001 0.012 0.927 91,169

0.043 0.014 0.004 83,086

−0.001 0.014 0.96 83,086

AZ MN MT NV

Medicaid may reimburse hygienists

Coefficient SE p N

0.025 0.009 0.007 91,169

0.007 0.013 0.58 91,169

0.037 0.01 0.001 83,086

0.015 0.011 0.179 83,086

MN MT NV WI

Regressions were estimated using OLS regressions. Standard errors are calculated using a Huber–White robust variance matrix that allows for clustering at the state level. State and year fixed effects were included in all models, so that regulation effects are identified using within state variation in licensing regulations. The models also controlled for Models also include controls for age, age squared, race-ethnicity, gender, education, household size and composition, employment status, and bracketed income categories.

Medicaid reimbursement policies, presented in the third panel, tell a similar story. Allowing Medicaid to reimburse hygienists directly increases one year utilization rates by about 2.5 percentage points (p = 0.007). There is no effect on two year utilization rates. One year teeth cleaning rates increase by 3.7 percentage points (p = 0.001) and there is no effect on two year teeth cleaning rates. 8. Conclusions In this paper, we studied the effects of a task based graded occupational licensing scheme that affects the way that dentists and dental hygienists are used to produce dental services. We presented a simple theoretical model that helps explain some of the ways that such regulations might affect the dental service production function. A simple implication of the model is that the regulations lead to a more dentist intensive production process, which is likely to affect the equilibrium price of dental services. We used a quasi-experimental approach to study the effects of the regulations on prevailing prices using data from dental insurance claims. Our results showed that the price of basic dental services were about 12% higher when the service could only be provided by a dentist rather than by either a dentist or a dental hygienist. These results were quite robust to key assumptions related to spillover effects and statistical inference. In further analysis, we found that utilization of dental services is 3–4 percentage points higher when hygienists can be reimbursed directly for their services, which is an important gain given well-documented disparities in dental health and access to dental care. Overall, our results are consistent with the constrained production function model presented in the paper. In most instances, the costs and benefits of licensing are difficult to empirically assess because licenses, almost by definition, make it difficult to construct reasonable comparison groups that can be used to estimate the levels of key health and economic outcomes under alternative policies. Cross-state comparisons are the most common way to proceed (Kleiner, 2000, 2006), but these methods make it difficult to separate the effects of licensing changes from state-specific trends in the demand and supply for the affected services. In a broader sense, most of the licensing literature examines the effects of licensing on wages rather than on the prices that prevail in related product markets. Our paper made some progress on both of these limitations. By focusing on service-specific regulations, we were able to compare the effects of regulations within the same state by comparing prices in different product markets that should share similar underlying demand and supply conditions. By

studying prices rather than wages, our analysis gave a different perspective on the way that licensing restrictions affects consumers. Another contribution of our paper comes from the analysis of graded licensing regulations. The bulk of the licensing literature frames policy discussions in terms of licensing, certification, and free entry options. This may be a natural statement of policy options, but there are very few examples of occupations that have been de-licensed in the United States (Kleiner, 2006). The graded licensing approach that was discussed in Arrow (1963) may offer an alternative approach that can reduce the economic disadvantages of licensing without overtly deregulating an incumbent occupational group. Graded licensing regulations that expand the scope of practice afforded to occupations with different levels of skill deserve additional research. This paper found that occupational regulation had an important effect on the cost and utilization of widely consumed dental services. We were not able to shed light on how these regulations affect the quality of services available in the dental markets or on how the regulations affect people’s exposure to the health risk of receiving care from a dangerously incompetent provider. As we argued earlier in the paper, previous research suggests that hygienists and dentists are equally competent and safe providers for the kinds of dental services we are studying in this paper. But the quality and safety effects of occupational regulation is a topic that needs additional research. It also seems to be true that the scope of practice regulations we studied in this paper are in place throughout the health care sector; however, there is not much research that documents the way these regulations affect the cost and quality of health care, or of the way that they affect the organization of care in different settings. The explicit link between job tasks and skill levels that is built into scope of practice regulations seems to fit well into the labor economics literature that is concerned with the factors shaping recent changes in the structure of wages (Acemoglu and Autor, 2011; Autor et al., 2003; Goldin and Katz, 2008). Occupational regulations have not been examined much in that literature, which has instead focused mainly on the slowdown in the supply of skills, increases in demand for skill produced (perhaps) by skill biased technological change, changes in international trade that have led to the “off-shoring” of certain types of work, and changes in labor market institutional structures such as labor unions and minimum wage levels (Autor et al., 2008). One conceptual insight from the wage structure literature that may be particularly useful for research on occupational regulation is the idea of separating the concept of the skills possessed by different workers from the


concept of the job tasks that workers perform in the economy (Acemoglu and Autor, 2011; Autor et al., 2003). This perspective seems to apply very naturally to the notion of scope of practice regulations, although it does not appear to have been considered in this way in previous work. References Acemoglu, D., Autor, D., 2011. Skills, tasks and technologies: implications for employment and earnings. In: Ashenfelter, O., Card, D. (Eds.), Handbook of Labor Economics, vol. 4, Part B. Elsevier, Amsterdam, pp. 1043–1171. Akerlof, G.A., 1970. The market for lemons: quality uncertainty and the market mechanism. The Quarterly Journal of Economics 84 (3), 488–500. Amemiya, T., 1978. A note on a random coefficients model. International Economic Review 19 (3), 793–796. Angrist, J.D., Pischke, J.S., 2009. Mostly Harmless Econometrics: An Empiricist’s Companion. Princeton University Press, Princeton. Arrow, K.J., 1963. Uncertainty and the Welfare Economics of Medical Care. The American Economic Review 53 (5), 941–973. Autor, D.H., Katz, L.F., Kearney, M.S., 2008. Trends in US wage inequality: revising the revisionists. The Review of Economics and Statistics 90 (2), 300–323. Autor, D., Levy, F., Murnane, R.J., 2003. The skill content of recent technological change: an empirical exploration. The Quarterly Journal of Economics 118 (4), 1279–1333. Bertrand, M., Duflo, E., Mullainathan, S., 2004. How much should we trust differences-in-differences estimates? The Quarterly Journal of Economics 119 (1), 249–275. Blundell, R., Dias, M.C., 2009. Alternative approaches to evaluation in empirical microeconomics. Journal of Human Resources 44 (3), 565–640. Bureau of Labor Statistics, 2011. Occupational Employment Statistics. U.S. Department of Labor www.bls.gov/oes/ Centers for Medicare and Medicaid Services, 2011. National Health Expenditure Accounts (Online) Available: https://www.cms.gov/Research-StatisticsData-and-Systems/Statistics-Trends-and-Reports/ NationalHealthExpendData/NationalHealthAccountsHistorical.html Choi, M.K., 2011. The impact of Medicaid insurance coverage on dental service use. Journal of Health Economics 30 (5), 1020–1031. Conley, T.G., Taber, C.R., 2011. Inference with “Difference in Differences” with a Small Number of Policy Changes. Review of Economics and Statistics 93 (1), 113–125. Donald, S.G., Lang, K., 2007. Inference with difference-in-differences and other panel data. The Review of Economics and Statistics 89 (2), 221–233. Friedewald, V.E., Kornman, K.S., Beck, J.D., Genco, R., Goldfine, A., Libby, P., Roberts, W.C., 2009. The American Journal of Cardiology and Journal of Periodontology Editors’ Consensus: periodontitis and atherosclerotic cardiovascular disease. Journal of Periodontology 80 (7), 1021–1032. Friedman, M., 1962. Capitalism and Freedom. University of Chicago Press, Chicago. Friedman, M., Kuznets, S., 1945. Income from Individual Professional Practice. National Bureau for Economic Research, New York. Glied, S., Neidell, M., 2010. The economic value of teeth. Journal of Human Resources 45 (2), 468–496. Goldin, C., Katz, L., 2008. The Race Between Technology and Education. Belknap, Cambridge, MA. Goldstein, Jacob, 2012. So you think you can be a hair braider? New York Times Magazine, June 17, 2012, on page MM20.

143

Hammermesh, D., Biddle, J., 1994. Beauty and the labour market. The American Economic Review 84 (5), 1174–1194. Heidenreich, A., Bellmunt, J., Bolla, M., Joniau, S., Mason, M., Matveev, V., Mottet, N., Schmid, H.P., van der Kwast, T., Wiegel, T., Zattoni, F., Zattoni, F., European Association of Urology, 2011. EAU guidelines on prostate cancer. Part 1. Screening, diagnosis, and treatment of clinically localised disease. European Urology 59 (1), 61–71. Holmes, T.J., 2013. (Online) Available: http://www.econ.umn.edu/∼holmes/ Humphry, L.L., Fu, R., Buckley, D.I., Freeman, M., Helfand, M., 2008. Periodontal Disease and Coronary Heart Disease Incidence: A Systematic Review and Metaanalysis. Journal of General Internal Medicine 23 (12), 2079–2086. Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica: Journal of the Econometric Society, 263–291. Klein, E., 2012. The Occupational Licensing Racket, Wonkblog, June 13, 2012 (Online). Available: http://www.washingtonpost.com/blogs/wonkblog/post/ the-occupational-licensing-racket-wonkbook/2012/06/13/gJQAd59dZV blog. html Kleiner, M.M., 2000. Occupational licensing. The Journal of Economic Perspectives 14 (4), 189–202. Kleiner, M.M., 2006. Licensing Occupations: Ensuring Quality or Restricting Competition? WE Upjohn Institute, Kalamazoo. Kleiner, M.M., Krueger, A.B., 2008. The Prevalence and Effects of Occupational Licensing. NBER Working Paper No. 14308. Kleiner, M.M., Park, K.W., 2010. Battles Among Licensed Occupations: Analyzing Government Regulations on Labor Market Outcomes for Dentists and Hygienists. NBER Working Paper No. 16560. Kleiner, M., Kudrle, R., 2000. Does regulation affect economic outcomes? The case of dentistry. Journal of Law and Economics 14 (4), 189–202. Kniesner, T.J., Leeth, J.D., 2004. Data mining mining data: MSHA enforcement efforts, underground coal mine safety, and new health policy implications. Journal of Risk and Uncertainty 29 (2), 83–111. Leland, H.E., 1979. Quacks, lemons, and licensing: a theory of minimum quality standards. The Journal of Political Economy 87 (6), 1328–1346. ˜ A.M., Murphy, S.L., Xu, J., Kochanek, K.D., 2011. Deaths: final data for 2008. Minino, National vital statistics reports: from the Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System 59 (10), 1. Moulton, B.R., 1990. An illustration of a pitfall in estimating the effects of aggregate variables on micro units. The Review of Economics and Statistics 72 (2), 334–338. Mouradian, W.E., Wehr, E., Crall, J.J., 2000. Disparities in children’s oral health and access to dental care. JAMA: The Journal of the American Medical Association 284 (20), 2625–2631. Mundinger, M.O., Kane, R.L., Lenz, E.R., Totten, A.M., Tsai, W.Y., Cleary, P.D., et al., 2000. Primary care outcomes in patients treated by nurse practitioners or physicians. JAMA: The Journal of the American Medical Association 283 (1), 59. Peltzman, S., 1975. The Effects of Automobile Safety Regulation. Journal of Political Economy 83 (4), 677–725. Rosenbaum, P.R., 2002. Observational Studies. Springer Verlag, New York. Rosenbaum, P.R., 2009. Design of Observational Studies. Springer Verlag, New York. Shapiro, C., 1986. Investment, moral hazard, and occupational licensing. The Review of Economic Studies 53 (5), 843. Viscusi, W.K., 1984. The Lulling Effect: The Impact of Child-Resistant Packaging on Aspirin and Analgesic Ingestions. American Economic Review 74 (2), 324–327. Wetterhall, S., Bader, J.D., Burrus, B.B., Lee, J.Y., Shugars, D.A., 2010. Evaluation of the Dental Health Aide Therapist Workforce Model in Alaska. Kellogg Foundation, Rasmuson Foundation, and Bethel Community Services Foundation, 97pp. Wooldridge, Jeffrey M., 2005. Fixed-effects and related estimators for correlated random-coefficient and treatment-effect panel data models. The Review of Economics and Statistics 87 (2), 385–390.

Dental insurance and equity of access to dental services.

Controlling the cost of dental health care insurance.

Spillover Effects of the Affordable Care Act? Exploring the Impact on Young Adult Dental Insurance Coverage.

The patient's records and the defense of dental malpractice claims.

Occupational injuries to dental personnel.

Occupational Hazards among Dental Surgeons.

Controlling the cost of dental care.

The effect of dental insurance on the use of dental care for older adults: a partial identification analysis.

Association Between Employee Dental Claims, Health Risks, Workplace Productivity, and Preventive Services Compliance.

Dental professional liability prevention: part one. Council on Insurance.

Dental professional liability prevention: part 2. Council on Insurance.

Comparing medical and dental providers of oral health services on early dental caries experience.

Ayurvedic drugs in case: Claims, evidence, regulations and ethics.

Evidence-based Update of Pediatric Dental Restorative Procedures: Dental Materials.

Evaluation of dental services in Somerset.

Cost-related dental non-attendance in older adulthood: evidence from eleven European countries and Israel.

Insurer views on reimbursement of preventive services in the dental setting: results from a qualitative study.

Public perceptions of the funding of NHS dental services.

Cost effectiveness of dental surgery procedures: a call for strengthening the evidence.

Self-rated dental health and dental insurance: modification by household income.

Cost-effectiveness of dental surgery procedures: a call for strengthening the evidence: reply.

Patterns of dental services and factors that influence dental services among 64-65-year-old regular users of dental care in Denmark.

Direct evidence of milk consumption from ancient human dental calculus.

Dental services for special and priority groups.