Occupational licensing of a credence good: the regulation of midwifery.

Jackson, John D.

1. Introduction

Occupational licensing is as old as trade. Estimates are that in the United States alone, at least 800 occupations require some form of "license to practice" (Rottenberg 1980, p. 2). Midwifery is most certainly among the oldest occupations known to Homo sapiens, and, unsurprisingly, it has been the subject of licensing regulations over the 20th century. There has been, however, a marked reemergence of the practice over the past 20 years in the United States. After nearly being driven from existence by physicians in the early part of the 20th century, the percentage of midwife attended births has risen from 0.9% of all births in 1975 to 5.95% of all births in 1995. This latter figure translates into 231,921 midwife-attended births for the year 1995. Of this figure, CNMs attended 94.3%, or 218,613, births. A number of factors account for this resurgence, including women's expression of their right to choose birth practitioners and place of birth, increased political expression of that right, and the escalating costs of traditional childbirth services by obstetricians (OBs) and hospitals (Butter and Kay 1988). In contrast, midwife-attended births account for a full 75% of all births in Europe, with far lower infant and maternal mortality rates reported (Coburn 1997).

Midwives are classified into two basic categories in this country: lay midwife and certified nurse-midwife (CNM). Lay midwives typically receive no formal educational training but are clinically trained through apprenticeships. On the other hand, a CNM "is a registered nurse with advanced training in midwifery who possesses evidence of certification by the American College of Nurse-Midwives (ACNM)" (Adams 1989, p. 1038). The practice of nurse-midwifery, as defined by the ACNM, is "the independent management of care of essentially normal newborns and women, antepartally (before birth), intrapartally (during birth), postpartally (after birth) and/or gynecologically ... within a health care system which provides for medical consultation, collaborative management, and referral" (Safriet 1992, p. 425).

The causes and effects of state regulation that determines the extent of professional independence from physicians of advanced practice nurses (APNs) has been analyzed by Dueker et al. (2000) for the same general period we employ. Advanced practice nursing, however, includes nurse practitioners, clinical nurse specialists, and nurse anesthetists as well as CNMs. Dueker et al. (2000) suggest that, for this larger category of nurse specialists, APN earnings are lower and physicians assistants earnings are higher in states where APNs have attained higher levels of professional independence (measured in part by prescriptive authority). (1) Midwifery has been included, along with other heath care professions, in interesting studies of the impact of the composition of public licensing boards on particular occupational requirements (Graddy and Nichol 1989; Graddy 1991), but (to the best of our knowledge) midwifery has not been isolated in any study of effects of regulation(s). (2) The purpose of this paper is thus t o analyze empirically the economic impact of alternative forms of regulation within the state markets for midwife services. Certified nurse-midwives are formally recognized by the American College of Obstetricians and Gynecologists (ACOG) and are now able to practice legally in all 50 states including the District of Columbia, but CNMs practice under significant and significantly different regulations that limit their scope of practice and constrain their use by women (DeVries 1985) within the 50 states. There are suggestions in the literature that the severity of regulations at the state level--a partial product of past pressure by the medical establishment (OBs in particular)--has had deleterious effects in the market for midwives' services. However, there has been (again to the best of our knowledge) no empirical support for such propositions or an analysis of the particular impact of alternative regulations. (3)

We believe that the market for midwives is particularly interesting from an economic perspective. (4) Midwifery is, to a large extent, a credence good, as much certainly as many other medical services. Such goods, it is sometimes argued (Leland 1979; Shapiro 1986), "demand" regulation on the basis of quality certification. Consumers, it is often alleged, will tend to drift to the low-price, low-quality alternative in the absence of such regulation. Imposition of some regulation in such markets may, in effect, shift the quality-adjusted demand curve rightward, improving consumer welfare and increasing the quantity supplied of such services. We label the potential quality-improving aspect of regulation the "demand-side effect."

Alternatively, mandatory occupational licensing, along with restrictive regulations supported by OBs and other medical professionals, may restrict entry, competition, and consumer choice. In short, a "supply-side" effect may be identified with restrictive regulations on CNMs that potentially reduces consumer welfare and redistributes wealth to competitors. The most important expressions of this view may be found in the work of Stigler (1971) and Peltzman (1976). In the case of CNMs, some regulations permitting certain benefits to the occupation, such as access to hospital facilities, granting prescriptive authority, or Medicare reimbursement, would put midwives on parity with OBs. These regulations, which would put midwives on a level competitive status are (generally) opposed by OBs (who wish to suppress a substitute and raise OB price), would shift the supply of CNM services rightward. Alternatively, regulations that limit the scope of midwives' activities would shift the supply curve of such services leftw ard, restricting supply and transferring income from CNMs and consumers to OBs with a deadweight loss.

Both the demand-side (quality enhancement) and the supply-side hypotheses unambiguously predict higher observed price increases, but the two diverge when predicting the quantity effects of more stringent occupational regulations. We therefore focus on quantity changes and regard our study of state midwifery regulations as one test of whether the dominant effect of regulation is to, on net, increase quantity through quality enhancement or to reduce the quantity consumed through a reduction in the quantity of services. (5) In calculating the effects of both demand and supply shifts in the CNM market, we compare the net effect of average versus minimum state regulations, where minimum regulations would represent parity with OBs.

The paper opens with a discussion of the institution of midwifery in the United States and a brief accounting of the types of regulations on this "credence good" in the 50 states. Next, a theory and empirical model are established to test for the effects of regulation. Finally, we analyze our results and offer some conclusions concerning the outcome of regulations in the market for a service characterized as a "credence good."

2. The Regulation of Midwifery in the United States

Midwifery regulation in the United States takes place under a plethora of methods and means. Table 1 summarizes eight of these methods and identifies the states that use them. According to data obtained from the ACNM in Washington, DC, as of 1995 there are a variety of methods for establishing a regulatory board's authority over nurse-midwifery practice. We have constructed our variable, MEDICAL BOARD AUTHORITY OVER CNM'S, by combining the two states that regulate CNM practice using a board of medicine with the five states that use a department of public health/board of health. Further, there are presently 27 states plus the District of Columbia that require CNMs to meet continuing education and recertification requirements as a condition for license renewal, with seven of those states requiring continuing education as a requirement for prescriptive authority only.

Prescriptive authority, the ability of a CNM to have discretion in the prescribing and dispensing of drugs that are within the scope of practice, is essential for a CNM to function independently of a physician. Twenty-four states plus the District of Columbia grant CNMs full authority to prescribe drugs and medication within their scope of practice as defined by the appropriate regulatory authority. Sixteen states either grant CNMs limited prescriptive authority or require physician control of that authority, while 10 states grant no prescriptive authority to CNMs. (6)

Both state and federal laws discriminate against and limit the ability of CNMs to practice by failing to mandate that third parties (private insurers) reimburse CNMs for services that are within their scope of practice and for services that are identical to physician provided (and third-party reimbursable) services. (7) In some states, Medicaid reimbursement status for CNMs is at a rate that is substantially less than that for physicians for the same service provided. (8)

Guaranteed clinical privileges are also potentially important CNM restrictions since states enact laws that regulated whether hospitals may permit or prohibit hospital facilities use. (9) In addition, we provide a variable that measures CNM control, CNM'S SUPERVISED BY MD'S, that would substitute for the part about CNMs being named in the authorizing statutes. (10)

Table 2 defines and provides sample means for all of the variables used in our tests. All eight regulations described with the state restrictions in Table 1 are included in the test. These variables are largely self-explanatory. We have included the percent of the Hispanic population as an independent variable in order to tract the effects of a social tradition of using midwives in Hispanic cultures.

3. Model Specification

A brief recitation of existing state rules and regulations reveals a wide diversity in midwifery regulation. And such diversity is suggestive of the varying intensities of political and other pressures that provide form to particular regulations affecting that occupation. Interest group strength is a clear determinant explaining forms of regulation in states or regions. Midwives, both lay and CNMs, are certain competitors with OBs. Hospitals, moreover, are competitors with less structured birthing centers and used by both OBs and midwives in some locales. Physician-sponsored state regulation of entry and other market aspects of medicine have been in place for well over a century in most states. Thus, some of the different incarnations of state regulation of midwifery may be explained, in part, by a "tar-baby" effect whereby a strong interest group (physicians) bring a substitute under the umbrella of monopoly. Birthing services are made, at least in some states, more complementary and less substitutable in th e interests of integrated monopoly--a tactic long recognized in economic literature (McKie 1970). (11) There is also the strong possibility that physicians, once in charge of the certification board of CNMs, add credence as to the quality of certified midwifery along with lowered credence in lay midwives. (Our test includes the latter possibility.)

But the wide disparity in the strength of regulations--such as the significant difference in insurance reimbursement rates and the stringency of regulation generally across states--also reflect particular and, it would seem, effective consumer interest groups. Some of this effectiveness, again in particular locales, may be based on customs and practices of ethnic populations. (12) The theoretical model we select to analyze midwifery, is a simple adaptation of supply and demand. As noted in the introduction, the credence characteristics of midwifery, whereby severe information problems mean that quality is unknown before and (sometimes) after the purchase (Darby and Karni 1973), can lead to "underconsumption" of the good. Price competition exacerbates that condition, and, in the "lemons" world of asymmetric information, higher-quality services may be driven from the market (Akerlof 1970). Occupational regulations, in this view, would have the effect of quality assurance, Increasing the demand for midwife servi ces and permitting quality enhancement.

The credence characteristic is of particular importance in the medical fields. An element of "belief" that a correct quality and/or quantity of the good or service will be or has been obtained is demanded of the consumer. Moreover, for midwife services, as with many medical credence goods, such as brain surgery or psychiatry, the full cost of ultimately discovering a "mistake" is apt to be far higher than nominal costs to consumers. The level of quality assurance demanded may well be significantly higher for consumers of these goods than for goods of other types. (13) If licensure and other forms of regulation are successful in improving quality, the demand for these services would be expected to increase. (14) Nelson (1974) provides an important counterpoint to this view, arguing that, under certain conditions, regulation or "certification" of a good or service provides a false sense of security in the purchase leading to a high number of type II errors by consumers.

The well-known alternative view of regulation is that mandatory licensing through a political process restricts entry, competition, and consumer choice. Deleterious supply-side effects reduce consumer welfare and redistribute wealth to members of the occupation--in our case to OBs. Reduced supply would be engendered in this familiar scenario of the effects of more stringent occupational regulations on the scope of midwifery practice.

Predicted effects of the two models have both a common and a divergent characteristic. More stringent occupational regulations will lead to higher observed prices under both the supply-side and the demand-side hypotheses regardless of the level of credence characteristics of the occupation. But, as noted in the introduction, the two hypotheses diverge when predicting the quantity effects of more stringent occupational regulations, and it is at this point that the level of credence characteristics exhibited by the occupation come into play. The supply-side hypothesis suggests that more stringent occupational regulations reduce the quantity consumed of a particular service through a shift in supply, while the demand-side hypothesis suggests that the regulations increase the quantity consumed of the service by eliminating or reducing the low-quality/low-price sector of the market, thereby increasing the demand for the service. Our theoretical model is a test of the dominant, net effect of the alternative regulat ions on the licensing of nurse-midwives. The details of this simple test follow.

Structural Equations

To test the theoretical model empirically, a demand-and-supply model of CNM services is specified as follows:

[Q.sub.d] = f(CNMPRICE,[R.sub.i], URBAN, REAL STATE PER CAPITA INCOME,% HISPANIC POPULATION) (15)

so that

[Q.sub.d] = [[alpha].sub.1] + [[alpha].sub.2] CNMPRICE + [summation over (10/j=3)] [[alpha].sub.j][R.sub.j-2] + [[alpha].sub.11] URBAN

+ [[alpha].sub.12] REAL STATE PER CAPITA INCOME + [[alpha].sub.13]% HISPANIC POPULATION + [[epsilon].sub.d]

[[alpha].sub.1] > 0, [[alpha].sub.2] < 0, [[alpha].sub.j] > 0 (j = 3,...,10), [[alpha].sub.11] < 0, [[alpha].sub.12] > 0, [[alpha].sub.13] > 0 (1)

and

[Q.sub.s] = f(CNMPRICE, POBPRICE/HOSPCOSTS, [R.sub.i])

so that

[Q.sub.s] = [[gamma].sub.1] + [[gamma].sub.2] CNMPRICE + [[gamma].sub.3] OBPRICE/HOSPCOSTS

+ [summation over (11/j=4)] [[gamma].sub.j][R.sub.j-3] + [[epsilon].sub.s]

[[gamma].sub.1] < [[alpha].sub.1], [[gamma].sub.2] > 0, [[gamma].sub.3] < 0, [[gamma].sub.j] < 0 (j = 4,...,11) (2)

The variables used in the model are defined in Table 2.

Demand Function ([Q.sub.d])

Following the law of demand, the quantity demanded of CNM services is assumed to be inversely related to the price of CNM services, CNMPRICE. The expected sign of the parameter [[alpha].sub.2] is therefore negative. The term [R.sub.i] is included in the demand function based on the quality certification demand-side hypothesis. This hypothesis suggests that more stringent regulations (discussed later) will increase the quantity demanded of CNM services at all price levels by eliminating the low-quality/low-price sector of the market. (16) Therefore, the expected sign of the parameter [[alpha].sub.j], (j = 3, ..., 10) is positive.

URBAN, the percentage of a state's population that lives in urban areas, is included in the demand function based on the assumption that higher population densities can support a wider variety of services, such as those provided by CNMs. Nurse-midwives have for decades provided care for underserved women in rural and inner city areas (American College of Nurse-Midwives 1994). Yet another study (Scupholme et al. 1992) concluded that twice as many CNMs (attending at least 22% of rural women) are practicing in rural areas than was reported in a limited Health and Human Services sample (Department of Health and Human Services 1992). Therefore, the expected sign of the parameter [[alpha].sub.11] is negative. REAL STATE PER CAPITA INCOME is included in the model on the assumption that CNM services are a normal good; the higher the income level, the greater the demand for these services, ceteris paribus, at any price level. Therefore, the expected sign of the parameter [[alpha].sub.12] is positive. % HISPANIC POPULA TION is included in the model based on the assumption that the greater the number of Hispanics in a particular state, the greater the demand for CNM services, ceteris paribus, at all price levels. (Hispanics have a tradition of utilizing the services of midwives.) Therefore, the expected sign of the parameter is positive.

Supply Function ([Q.sub.s])

Following the law of supply, the quantity supplied of CNM services is assumed to be directly related to own price, CNMPRICE, and hence the expected sign of the parameter [[gamma].sub.2] is positive. OBPRICE/HOSPCOSTS, the average OB price in a state as a percentage of hospital costs in that state, is included in the supply side of the model as a proxy for the cost of production. The hospital costs in each state includes room and board and all ancillary services for an uncomplicated vaginal delivery. The expected sign of the parameter [[gamma].sub.3] is therefore negative. The term [R.sub.i] is included in the supply function based on the interest group supply-side hypothesis. This hypothesis suggests that regulations will decrease the quantity supplied of CNM services at all price levels by increasing the cost of entry to prospective CNMs. Therefore, the expected sign on each of the parameters [[gamma].sub.j] (j = 4, ..., 11) is negative.

4. Empirical Estimates

Appealing to simple supply-and-demand analysis, the quality-enhancing effect of regulation would shift the demand curve rightward, increasing equilibrium price and quantity. If supply restriction occurs, the supply curve shifts leftward, increasing equilibrium price and reducing quantity. Clearly reduced-form equations for price will not allow us to distinguish between the two hypotheses since restrictions increase price in both cases. However, in reduced-form quantity equations, a dominance of the supply effect will reduce quantity, while quality enhancement will positively affect quantity. We therefore concentrate on this fundamental equation.

From an econometric perspective, it should be clear that we wish to estimate a reduced-form quantity equation for CNM services. The parameters for the reduced-form quantity equation are purged of statistical biases resulting from the joint determination of prices and quantities and can therefore be estimated using ordinary least squares (OLS) (Gujarati 1988):

[CNMBIRTHS.sub.i] = [[pi].sub.1] + [[pi].sub.2] MEDICAL BOARD AUTHORITY OVER [CNM'S.sub.i]

+ [[pi].sub.3] CONTINUING [EDUCATION.sub.i]

+ [[pi].sub.4] NO MANDATED INSUR. [REIMBURSEMENT.sub.i]

+ [[pi].sub.5] CLINICAL PRIVILIiEGES NOT [GUARANTEED.sub.i]

+ [[pi].sub.6] NO PRESCRIPTIVE [AUTHORITY.sub.i] + [[pi].sub.7] CNM'S SUPERVISED BY [MD'S.sub.i]

+ [[pi].sub.8] LAY MIDWIVES NOT [PERMITTED.sub.i]

+ [[pi].sub.9] LOW CNM MEDICAID [REIMBURSEMENT.sub.i]

+ [[pi].sub.10] [OBPRICE/HOSPCOSTS.sub.i] + [[pi].sub.11] [URBAN.sub.i]

+ [[pi].sub.12] REAL STATE PER CAPITA [INCOME.sub.i]

+ [[pi].sub.13]% HISPANIC [POPULATION.sub.i] + [[epsilon].sub.i], (3)

where the variables are as defined in Table 2. Simple algebra and the hypothesized signs from the structural equations indicate that reduced form coefficients [[pi].sub.10] through [[pi].sub.13] are unambiguously positive or unambiguously negative. Reduced form coefficients on the regulatory variables, [[pi].sub.2] through [[pi].sub.9], will be signed in accordance with which view dominates: positive if the demand side view dominates and negative if the supply side view dominates. The data for the quantity of CNM services, CNMBIRTHS, Consist of a single observation for each of the 50 states in the survey. (17)

Estimation with Regulatoy Sector Exogenous

Table 3 presents maximum likelihood estimates of the reduced-form quantity equation under two conditions: (i) when the regulatory sector is exogenous and (ii) when the regulatory sector is endogenous.

Cross-sectional studies often encounter problems with heteroscedasticity, and our results in Table 3 are no exception. Preliminary OLS estimates of the regulatory sector exogenous model indicated a Breusch--Pagan statistic of [chi square] = 16.96, and preliminary instrumental variables (IV) estimates of the regulatory sector endogenous model revealed a Breusch--Pagan statistic of 8.03. Clearly, heteroscedasticity is a problem that we must address.

Traditionally, a generalized least squares (GLS) procedure in which the nonconstant variance is assumed to be proportional to, say, the square of some given explanatory variable is employed to attack this problem. Under this assumption, the GLS transformation amounts to simply weighting all variables by the reciprocal of the given variable. Recently, however, analysts have become more sophisticated in their assumptions concerning the form of the variance function. One popular assumption is one of "multiplicative heteroscedasticity," in which the logarithm of the nonconstant disturbance variance [[sigma].sup.2.sub.i] is assumed to be a linear function of some key variables. Preliminary analysis of the relationship between the squared OLS residuals obtained from estimating Equation 3 and some potential explanatory variables suggested that, for our problem, a variance function of the form

In [[sigma].sup.2.sub.i] is [[phi].sub.0] + [[phi.sub.1] STATE COST OF LIVING INDEX

+ [[phi].sub.2] STATE PER CAPITA INCOME IN 1995 + [zeta] (4)

might be appropriate. (18) It is worth noting that estimating this variance function itself provides a direct test of heteroscedasticity: Statistically insignificant estimates of [[phi].sub.1] and [[phi].sub.2] imply a constant variance (estimated by the antilog of [[phi].sub.0]). and statistically significant estimates of 'Pi and (P2 clearly indicate a nonconstant variance.

Greene (2000) shows that, since the Hessian of the likelihood function is block diagonal, maximum likelihood estimates of the it's in Equation 3 and the [pi]'s in Equation 4 can be found through a simple iterative process. We begin by estimating Equation 3 by OLS. The logs of the squared residuals from Equation 3 are then used to proxy In [[sigma.sup.2.sub.i] in Equation 4 so that the [phi]'s in that equation can then be consistently estimated by OLS. (19) The antilog of the estimated variance function provides estimates of [[sigma.sup.2.sub.i] that can be used to obtain GLS estimates of Equation 3. The log of the squared OLS residuals can then be used to new estimates of Equation 4, which can then be used to obtain new GLS estimates of Equation 3 and so on. The iterations continue until the estimates of both parameter vectors, [pi] and [phi], stabilize. This is the procedure that we used to obtain the parameter estimates presented in Table 3.

The signs on the coefficient estimates in Table 3 conform to our a priori expectations. When the regulatory sector is assumed exogenous, only two of the eight regulatory variables are statistically insignificant, CONTINUING EDUCATION and LOW CNM MEDICAID REIMBURSEMENT, while all four of the nonregulatory variables are statistically significant at traditional levels. These results are not totally satisfactory, however. Sass and Saurman (1995) make a convincing argument that in models such as the one we posit here, the licensing variables are likely to be jointly determined with price and quantity. If this is the case, our reduced-form coefficient estimates in Table 3 (regulatory sector exogenous) are biased and inconsistent. It is therefore essential that we test for the presence of an endogenous political sector. The test introduced by Hausman (1978) has become the standard for evaluating such questions. But Hausman's test requires instruments for the political variables. While there are numerous approaches t o obtaining "acceptable" instruments, they are available on a systematic basis only from estimated political models. Thus, we adopt the following procedure to create our instruments.

We begin by supposing that the parameters of the structural equations explaining MEDICAL BOARD AUTHORITY OVER CNM'S, CONTINUING EDUCATION, NO MANDATED INSUR. REIMBURSEMENT, CLINICAL PRIVILEGES NOT GUARANTEED, NO PRESCRIPTIVE AUTHORITY, CNM'S SUPERVISED BY MD'S, LAY MIDWIVES NOT PERMITTED, and LOW CNM MEDICAID REIMBURSEMENT are jointly determined in an eight-equation system. (20) In principle, these eight equations are part of a larger (10-equation) system that also determines the price and quantity of CNM services. But since we are interested only in whether potential endogeneity of the regulatory variables with equilibrium quantity of CNM service biases the reduced-form coefficient estimates of Table 3, we need to construct instruments only for the eight regulatory variables. Thus, we confine our attention to the smaller system composed of the eight structural equations explaining these regulatory variables.

In any event, we make no attempt to precisely specify any of these structural relationships; there is no need. Recalling that the criteria for an "appropriate" instrument are that it be highly correlated with the variable it purports to measure and uncorrelated with the corresponding disturbance, the reduced-form equations of the system are sufficient to generate satisfactory instruments for the regulatory variables, as is the case in typical two-stage least squares procedures. Consequently, we estimate probit regressions explaining each of the eight regulatory variables with (the same) nine independent variables using data for the 50 states included in our sample (i.e., N = 50). Specifically, the nine explanatory variables include the percentage of the state's senate and of the state's house held by the Democratic Party, the ratio of the state's house to the state's senate, the political party of the governor, the average hospital charges for an uncomplicated vaginal delivery in each state, the percentage of the state's population that lives in urban areas, the number of CNMs per capita, physician deliveries as a percentage of total deliveries in each state, the state's population in 1995, and a constant term. These variables can be taken as all the exogenous variables in the regulatory equation system; all that is required is that each one enters at least one of the eight structural equations. As such, the eight estimated equations comprise the reduced-form equations of the structural system. The predicted values of the dependent variable in each probit regression become the instruments for the corresponding regulatory variables to be used in the reduced form for CNMBIRTHS to perform the Hausman test for endogeneity.

Before turning to the conduct, outcome, and implications of this test, we note that all the explanatory variables in the reduced forms are well grounded in a public choice approach to modeling the supply and demand for CNM regulations. (21) Each variable is a measure of the extent to which some factor affects the incentives of legislators to bargain among themselves, the accountability of legislators to the public, or the size of some interest group that might wish to influence regulation-related legislation. Previous studies have found these types of variables significant in explaining the existence of various regulations. (22)

Our point is that it is quite possible to specify a set of reduced-form equations, well grounded in theory and precedence, without specifically positing the underlying structural system. Since our sole object in developing a political model is to obtain legitimate instruments for the regulatory variables in our CNM market model, we choose to follow this course of action.

Estimation with Regulatory Sector Endogenous

Table 3 (regulatory sector endogenous) presents IV estimates of Equation 5 using the instruments for the political variables developed in the previous section. Based on the OV (omitted variables) version of the Hausman test (Kennedy 1992), the test statistic was a chi-square (8) of 52.4828. This exceeds the critical value of a chi-square (8) at the .05 level of 15.5073. Therefore, the null hypothesis of consistent estimation of the parameters of the reduced-form quantity equation is rejected at any reasonable level. This result suggests that our initial estimates of the quantity equation must be corrected for simultaneity bias. Therefore, we now shift our focus to the IV estimates.

Our results for the quantity equation bear directly on the competing demand- and supply-side hypotheses concerning the effects of CNM regulations. The result for the nonregulatory variable OBPRICE/HCOSTS suggests that the higher the ratio of OB prices to total hospital costs, the higher the quantities consumed of CNM services, although the parameter estimate is not statistically significant. Higher income levels and the greater the percentage of a state's population that is Hispanic have a positive effect on the number of CNM deliveries. The parameter estimates for both of these variables, REAL STATE PER CAPITA INCOME and % HISPANIC POPULATION, are positive and significant at the .01 level. For each thousand-dollar increase in real per capita income in a state, CNM deliveries increase by about 1 percentage point, or about 18%. (23) In addition, for each percentage-point increase in the Hispanic population in a state, CNM deliveries increase by approximately .22 percentage points, or about 4%.

The parameter estimates for two of the eight regulatory variables, NO PRESCRIPTIVE AUTHORITY and LOW CNM MEDICAID REIMBURSEMENT, are not statistically significant. It appears that allowing CNMs either full or limited prescriptive authority in a particular state has no bearing on the number of CNM deliveries in each state. A low level of Medicaid reimbursement for CNMs, as compared to physicians, also appears to have no effect on the number of CNM deliveries in each state.

The parameter estimates of the regulatory variables MEDICAL BOARD AUTHORITY OVER CNM'S and CONTINUING EDUCATION support the demand-side hypothesis. Both parameter estimates are positive and are statistically significant at the .01 and the .05 level, respectively. If CNMs are supervised by a regulatory board other than a board of nursing, midwifery, or certified nurse midwifery or a board that includes nurses or has nurse input, then the number of CNM deliveries roughly doubles in that particular state. As suggested earlier, this regulation (as measured by our variable) provides "credence" to the services of CNMs while simultaneously reducing perceived quality of lay nurse-midwives. Requiring CNMs to enhance their practice skills through continuing education requirements for license renewal increases CNM deliveries by approximately 1.4 percentage points, or 29%, compared to those states that do not have such requirements.

The parameter estimates of the four remaining regulatory variables, NO MANDATED INSUR. REIMBURSEMENT, CLININCAL PRIVILEGES NOT GUARANTEED, CNM'S SUPERVISED BY MD'S and LAY MIDWIVES NOT PERMITTED are all negative in sign and statistically significant at either the .05 or the .01 level. The signs and significance of these estimates lend support to the supply-side hypothesis. Private insurance reimbursement mandates or AWP laws increase CNM deliveries by about 1.8 percentage points, or 40%, compared to those states that have no such mandates. Both the guarantee of hospital admitting privileges to CNMs and their ability to practice independently of physicians have a dramatic impact on the number of CNM deliveries in a particular state, resulting in an increase in CNM deliveries of approximately 73% and 109%, respectively. (24) The ban on the practice of lay midwifery results in a decrease in CNM deliveries of about 3 percentage points, or about 46%, compared to those states that do not ban this practice. While th is seems contrary to a priori expectations, as lay midwives can be viewed as competitors to CNMs, it appears that this variable is a proxy for the tendency to oppose midwife practice (both lay and CNM) in general in a particular state.

5. Summary and Conclusion

The theory and empirical model developed in this paper analyzes the theoretical effects of regulation through supply and demand on prices and quantities and develops an empirical model to analyze the quantity of CNM services. Regulation of CNMs is a specific case of regulation that must be analyzed and interpreted relative to the regulation of OBs. Since the use of either supply-side (Stigler-Peltzman) or demand-side (quality assurance) hypotheses predicts higher prices from increased regulation of CNMs, we focus on the quantity effects from increased regulation.

The two hypotheses diverge in their predictions concerning the effects of increased regulation of CNMs when it comes to the quantities consumed of CNM services. Our results suggest that the supply-side (quantity-reducing) effects dominate the demand-side (quality assurance and quantity enhancement) effects. When evaluated at their respective means and at their sample minimums, the resulting effect of minimum regulations versus mean regulations on CNMs is to increase the percentage of CNM births from approximately 5.76% to 11.12% of all births in the 50 states. The results support the hypothesis that the more restrictive a state's statutes concerning CNM regulations, that is, those that reduce parity with OBs, the less will be the quantities consumed of those services in that state. Although CNM services can clearly be regarded as having some fairly significant credence characteristics--and these effects are important to exchange in the CNM market--it appears that regulation of this type of service has detrime ntal consumer welfare effects. (25) In a time when many medical service delivery systems are in chaos, the advantages to deregulation of such fundamental activities should not be minimized.

Appendix

Data Sources

Council of State Governments. The Book of the States (1992/1993).

Statistical Abstract of the United States. 1996.

U.S. Bureau of the Census. 1990.

U.S. Bureau of the Census. 1992. Current Population Reports.

U.S. Bureau of the Census. 1994. City and County Data Book.

U.S. Department of Commerce. 1992. Census of Service Industries.

U.S. Department of Commerce. 1992. Bureau of Economic Analysis.

U.S. Department of Labor. Dictionary of Occupational Titles.

Table 1

State-Mandated Regulatory Restrictions over Certified Nurse Midwives

CNM Restriction States with Restriction

Medical board authority over CNMs CT, DE, HI, NJ, NM, PA, RI
Continuing education requirement AL, AK, AZ, AR, GA, ID, IN,
 IA, KS, ME, MD, MI, MS, MT,
 NV, NM, ND, OR, RI, SC, TX,
 UT, VT, WA, WAV, WI, WY

Insurance reimbursement mandated or AK, CA, CO, CT, DE, FL, GA,
 any willing provider laws ID, IL, IN, KY, LA, MD, MA,
 MI, MN, NV, NH, NJ, NM, NY,
 OH, OK, OR, PA, SD, UT, WA,
 WV, WY

Clinical practice privileges FL, GA, OH, OR, VA
 guaranteed
Prescriptive authority for CNMs AK, AZ, AR, CA, CO, CT, FL,
 ID, IN, IA, KS, ME, MD, MA,
 MI, MN, MS, MO, MT, NE, NV,
 NH, NJ, NM, NY, NC, ND, OR,
 RI, SC, SD, TN, TX, UT, VT,
 VA, WA, WV, WI, WY

Supervised by MDs AL, AR, CA, CO CT, FL, HI, ID,
 KS, LA, ME, MD, MA, MS, MO,
 NE, NV, NJ, NM, NY, NC, OH,
 PA, SC, SD, VA, WI

Lay midwives permitted in state AL, AK, AZ, AR, CA, CO, FL,
 GA, KY, LA, ME, MA, MI, MN,
 MS, MO, MT, NE, NH, NJ, NM,
 NY, OK, OR, PA, RI, SC, TN,
 TX, UT, VT, VA, WA, WV, WI,
 WY

Medical reimbursement 80% or lower AL, AZ, AR, FL, HI, IL, IN,
 than MD rate IA, KS, KY, MD, MT, NV, NJ,
 ND, RI, SC

Table 2

Variable Names, Sample Means, and Descriptions

Variable Name Sample Mean Description

CNMBIRTHS 5.76% CNM attended births as a percentage
 of total births in each of the 50
 states for 1995.

MEDICAL BOARD 0.14 Indicates the committee, board, or
 AUTHORITY agency that regulates
 OVER CNM'S nurse-midwifery practice in a
 particular state. A dummy variable
 is used with a 1 indicating that
 CNMs are regulated by a board of
 medicine or a department of public
 health/board of health in a
 particular state. A value of 0
 indicates that CNMs are regulated
 in a particular state by any of
 the following: board of nursing,
 board of nursing with board of
 medicine input, certified
 nurse-midwifery board, board of
 midwifery, or jointly by a board
 of nursing and a board of
 medicine.

CONTINUING 0.54 Indicates whether a state requires
 EDUCATION continuing education units for
 CNMs to renew their license to
 practice in that state. A dummy
 variable is used with a 1
 indicating that the state requires
 this or a 0 indicating if it does
 not.

NO MANDATED INSUR. 0.40 Indicates whether a state mandates
 REIMBURSEMENT private insurance reimbursement
 for CNM services or if the state
 has enacted an "any willing
 provider" (AWP) law. A dummy
 variable is used with a 1
 indicating that the state does not
 have this mandate or AWP law or a
 0 indicating that it does have
 this mandate or AWP law.

CLINICAL PRIVILEGES 0.90 Indicates whether a state has
 NOT GUARANTEED enacted statutes that either
 permit hospitals to grant CNMs
 clinical practice privileges or
 prohibits hospitals from
 discriminating against CNMs in the
 granting of these privileges. A
 dummy variable is used with a 1
 indicating that the state does not
 have either statute or a 0
 indicating that it has one or the
 other statue.

NO PRESCRIPTIVE 0.20 Indicates whether a state grants
 AUTHORITY prescriptive authority to CNMs. A
 dummy variable is used with a 1
 indicating that a state does not
 grant either full or limited
 prescriptive authority to CNMs or
 a 0 indicating that it does not
 grant CNMs full or limited
 prescriptive authority.

CNM'S SUPERVISED 0.54 Indicates reduced support CNM
 BY MD'S independence in a particular
 state. A dummy variable is used
 with a 1 indicating that a state's
 nurse-midwifery practice act
 includes, uses, or refers to (i)
 protocols rather than practice
 guidelines, (ii) terms such as
 "medical functions" or "delegated
 medical acts," or (iii) terms such
 as "supervision" or "direction" to
 describe the CNM's relationship
 with physicians. A 0 is used to
 indicate that CNMs have greater
 independence from physicians in a
 particular state.

LAY MIDWIVES 0.28 Indicates whether lay midwives are
 NOT PERMITTED allowed to practice in the state.
 A dummy variable is used with a 1
 indicating that the state outlaws
 lay midwives or a 0 indicating if
 it does not.

LOW CNM MEDICAID 0.34 Indicates the extent to which
 REIMBURSEMENT Medicaid reimburses CNMs for
 delivery services compared to
 physicians. A dummy variable is
 used with a 1 indicating that the
 Medicaid reimbursement rate for
 CNMs is 80% or lower than the
 physician reimbursement rate in a
 particular state. A 0 indicates
 that CNMs are compensated for
 delivery services by Medicaid at
 a rate higher than 80% of the
 physician reimbursement rate.

RATIO OF .6867 The ratio of average obstetrician
 OBPRICE/HOSPCOSTS prices to average total hospital
 charges for un uncomplicated
 vaginal delivery in each of the 50
 states for 1993, inflated to 1996
 price levels by the medical cost of
 living index.

URBAN 68.18% Percentage of the population that
 is urban in each of the 50 states.

REAL STATE PER 22,384 State per capital income adjusted
 CAPITA INCOME by the cost of living index for
 each state.

% HISPANIC POPULATION 5.2802% Percentage of the population that
 is Hispanic in each of the 50
 states.

Table 3

Reduced-Form Quantity Estimates (Assuming Multiplicative
Heteroscedasticity)

 Maximum Likelihood Estimates
 Regulatory Sector Exogenous
Variable Coefficient t-ratio

INTERCEPT 0.00134742 0.022466
MEDICAL BOARD AUTHORITY 0.0448621 3.5665
 OVER CNM'S
CONTINUING EDUCATION 0.00920926 1.32656
NO MANDATED INSUR. -0.02355 -3.06799
 REIMBURSEMENT
CLINICAL PRIVILEGES NOT -0.0363571 -3.79327
 GUARANTEED
NO PRESCRIPTIVE AUTHORITY -0.0149473 -1.70286
CNM'S SUPERVISED BY MD'S -0.0119984 -1.88629
LAY MIDWIVES NOT PERMITTED -0.0235863 -3.10315
LOW CNM MEDICAID 0.00839065 1.12452
 REIMBURSEMENT
RATIO OF OBPRICE/HOSPCOSTS 0.0615535 2.85874
URBAN -0.00123594 -3.00491
REAL STATE PER CAPITA INCOME 0.000570785 1.93224
% HISPANIC POPULATION 0.00172976 2.94605

Variance Function Estimates

Sigma 0.000594417 1.10661
State cost-of-living index 0.157043 5.18109
State per capita income -0.00036956 -3.37809

Summary Statistics (c)

N 50
[R.sup.2] 0.46
[chi square](16) 47.0844

 Maximum Likelihood Estimates
 Regulatory Sector Endogenous (a)
Variable Coefficient t-ratio

INTERCEPT -0.0738628 -1.1594
MEDICAL BOARD AUTHORITY 0.0577459 4.30184
 OVER CNM'S
CONTINUING EDUCATION 0.0142667 1.99793
NO MANDATED INSUR. -0.0182948 -2.16778
 REIMBURSEMENT
CLINICAL PRIVILEGES NOT -0.0391105 -2.85241
 GUARANTEED
NO PRESCRIPTIVE AUTHORITY -0.00267303 -0.25134
CNM'S SUPERVISED BY MD'S -0.0412903 -5.84603
LAY MIDWIVES NOT PERMITTED -0.0299316 -3.68219
LOW CNM MEDICAID 0.0100652 1.1973
 REIMBURSEMENT
RATIO OF OBPRICE/HOSPCOSTS 0.0271361 1.32729
URBAN -0.00110181 -2.87024
REAL STATE PER CAPITA INCOME 0.00102681 3.30707
% HISPANIC POPULATION 0.00221497 4.0252

Variance Function Estimates

Sigma 0.00296663 1.10661
State cost-of-living index 0.099501 3.2827
State per capita income -0.000274214 -2.50655

Summary Statistics (c)

N 50
[R.sup.2] 0.69
[chi square](16) 67.0874

(a)Exogenous variables in the probit models used to determine the
predicted values for the regulatory variables include hospital costs,
percentage urban, state population (1995), political variables (the
ratio of House size to Senate size, whether the state had a Republican
governor, and the percentage of Democrats in the Senate), and variables
indicating the size of competing interest groups (the number of midwives
per capita and the percentage of total births conducted by MDs). The
variable UNKNOWN was also included in the NOCLINPP probit in order to
avoid perfect multicollinearity between its predicted value with the
constant term.

(b)The coefficients arise when we use the predicted values from the
estimated probit equations outlined in note a as instrumental variables
to avoid potential simultaneity problems.

(c)Summary statistics: N is the sample size; [R.sup.2] is the
coefficient of determination (its meaning is unclear in instrumental
variables models); [chi square] (16) is the statistic for testing the
joint significance of the slope coefficients (its critical value for 16
degrees of freedom at the 5% level of significance is 26.2923).

Received January 2001; accepted March 2002.

(1.) Dueker et al. (2000) suggest that this result may obtain because physicians substitute physician assistants for APNs for self-interested reasons.

(2.) Graddy and Nichol (1989) explore the effects of public licensing board members on legislative regulatory reforms using four health-related occupations (chiropractors, licensed practical nurses, physicians, and registered nurses). Their results suggest that the more public members (not members of the occupation being licensed) an occupational licensing board has, the more effective the board is "in reducing the number of nonsense requirements (morality, age, residency/citizenship) that limit entry into the four health occupations studied" (1989, p. 623). Graddy's (1991) study covers dietitians, nurse-midwives, occupational therapists, physician assistants, psychologists, and social workers. See also Gaumer (1984), who reviews the empirical literature in the area.

(3.) The ACNM reports that the states with the most restrictive regulations have the lowest percentage of CNM-attended births, 1.7% (1991 figures), while those states that are moderately supportive and supportive of CNMs have 4.5% and 6.0%, respectively, of all births attended by CNMs.

(4.) Occupational regulations for credence goods, including some aspects of midwifery, have been explored. Sass and Nichols (1996), for example, explain why nonphysician health care professionals might demand less regulation (meaning less physician controls) in spite of income reductions for themselves. Using a "full-value" argument, they argue that, for some professionals, the nonmonetary rewards of independence may be high.

(5.) While we do not formally develop an analysis of price effects in this paper, we estimate, using unique price data, an empirical model that allows us to make preliminary welfare calculations. The calculations are reported later in this paper, and the empirical underpinnings are available from the authors on request.

(6.) As will be seen, we construct our variable so as to lump full prescriptive and limited prescriptive authority together. Decomposing these variables yields less "robust" results.

(7.) Twenty-one states mandate private insurance reimbursement of nurse-midwifery services, while nine states have enacted an "any willing provider" (AWP) law. According to she American College of Nurse-Midwives (1995), AWP laws include "CNMs, either specifically as CNMs or as ANPs (Advanced Nurse Practitioners) or ARNPs (Advanced Registered Nurse Practitioners). AWP laws typically require HMOs or other categories of managed care plans to permit any health care professional to become a participating provider in that plan, so long as s/he is willing to accept the terms and conditions the plan offers to its chosen participating providers. Variations on such laws are 'freedom of choice' statutes, which prohibit class-based discrimination against certain categories of health professionals."

(8.) Reimbursement rates vary as a percentage of the physician fee schedule or on the basis of services provided. For the states covered in this study, the range is between 70% and 100% of the physician fee schedule, with a full 27 states providing reimbursements at the highest level. (Utah reimburses CNMs according to a CNM schedule.) Table 1 includes only those states (17) that reimburse CNMs at lower levels.

(9.) According to the American College of Nurse-Midwives (1995), 45 states have "no statutory or regulatory provisions (that) either require hospitals to grant admitting or other clinical privileges to CNM's or prohibit discrimination against CNM's" (p. vi).

(10.) Regarding CNM supervision (CNMs supervised by MDs), the American College of Nurse-Midwives (1995) reports that there are certain "signs" that indicate whether the Nurse-Midwifery Practice Act in a state is supportive of ACNM guidelines and standards for CNM practice. The "signs" in the state's practice act that indicate reduced support for CNM independence include (i) whether the practice act refers to protocols rather than practice guidelines, (ii) whether the scope of nurse-midwifery practice uses terms such as "medical functions" or "delegated medical acts," and (iii) whether the practice act uses terms such as "supervision" or "direction" to describe the CNM's relationship with physicians. The Nurse-Midwifery Practice Act in 27 states indicates reduced support for CNM independence by including some or all of the preceding language in the "act." The ACNM says that you have a "good" Nurse-Midwifery Practice Act if "the practice act defines nurse-midwifery practice as independent (either directly or in directly) and does not contain requirements for physician supervision or direction" or "the practice act references or directly quotes the ACNM definitions of consultation, collaboration and referral to describe the CNM relationship with physicians."

(11.) Our tests treat OB prices as independent of midwifery charges, however. A more elaborate test--given data availability, of course--would account for the possibilities of a "tar-baby" effect and their joint determination. Further, it would clearly be in the interest of both OBs and CNMs to pass regulations suppressing lay midwives. Our empirical findings support the fact that CNMs are substitutes for lay midwives.

(12.) An interesting and valid avenue of inquiry--one not addressed in this paper--would be to explain why regulations are as they are in each of the 50 states. The state of Texas, for example, with a large Hispanic population that carry traditions of midwifery, would be expected to experience less stringent regulations on midwife practices. Our more limited concern, however, is with the effects of these regulations on efficiency and economic welfare once they are in place.

(13.) Little empirical evidence has been produced in this area, but see Ekelund, Mixon, and Ressler (1995), where evidence is provided on relative intensities of information for credence and experience goods vis-a-vis search goods in Yellow Pages advertising. For some categories, such as child day care, chiropodists, optometrists, psychologists, and marriage/family counseling, information intensities (measured by licensing, certification, and other quality attributes) were not significantly different from "experience" goods but of (statistically) greater intensity than for search goods. This result was perhaps quite significant given the traditional prohibitions against advertising in "medical" fields.

(14.) Some evidence exits which links quality measures to what may be termed "credence" services. Carroll and Gaston (1981b) found that states with more restrictions in the legal profession had higher quality rankings. Holden (1978) found that higher failure rates on entry exams for dentists was associated with better service quality. However, Carroll and Gaston (1981a) found contrasting results for dentists.

(15.) The term [R.sub.i] is a vector of restrictions, MEDICAL BOARD AUTHORITY OVER CNM'S, CONTINUING EDUCATION, NO MANDATED INSUR. REIMBURSEMENT, CLINICAL PRIVILEGES NOT GUARANTEED, NO PRESCRIPTIVE AUTHORITY, CNM'S SUPERVISED BY MD'S, LAY MIDWIVES NOT PERMITTED, LOW CNM MEDICAID REIMBURSEMENT, which is included in both the demand and the supply functions (see Table 1).

(16.) Leland (1979) uses as an example the market for physicians, arguing that there is informational asymmetry between doctor and patient concerning the quality of medical services rendered. Since "patients ... have difficulty in distinguishing the relative qualities of physicians ... all doctors must therefore command the same fees, which wilt reflect the average quality of medical services. Doctors with above-average opportunities elsewhere may not he willing to remain in (or enter) the market, since the price they receive will reflect the lower average quality of service. Their withdrawal from the market lowers the average quality of medical services, the price falls, and further erosion of high-quality physicians occurs" (p. 1329). Leland suggests that licensing, or other forms of minimum quality standards, may he a relatively inexpensive way of eliminating this informational asymmetry resulting in the elimination of the low-quality/low-price sector of the market.

(17.) Data for this variable have been obtained from the Statistical Resources Branch Division of Vital Statistics of the U.S. Department of health and Human Services for 1995. The data are for total CNM-attended births as a percentage of total births in each of tie 50 states. Sources of other data are the Council of State Governments, the Census Bureau, the Department of Commerce, the Department of Labor, and the Statistical Abstract of the United States listed at the end of the references to this paper.

(18.) State cost of living indices are not as easy to find as one might think. The measure we use comes from a paper by Izraeli and Murphy (1997).

(19.) Technically, consistent estimation of the complete parameter vector [phi] requires adding a constant (1.2704) to the constant term.

(20.) Assuming that the political variables are (contemporaneously) jointly determined may gloss over some important dynamics intrinsic to the implied relationships. Both legislative and constitutional values change over time, the latter far less frequently. Unfortunately, no adequate or well-specified model of regulatory change yet exists with which to explain institutional evolution. While we look forward to such a model, a potential gap in our specification is that we use current rather than original magnitudes to explain our regulatory variables in our subsequent reduced-form regressions. Legislators can modify (or eliminate) regulations if they choose, but cost levels suggest that licensing requirements change infrequently. Our use of current values implicitly suggests that legislative change is costless. In that sense, we assume away potentially important problems.

(21.) A more complete description of the explanatory variables (along with sample means) and the empirical results from estimating the regulatory reduced-form equations (accompanied by a behavioral analysis of the results) is available from the authors on request.

(22.) For example, McCormick and Tollison (1981) found that variables such as the size of the legislature, the relative size of the two houses, and the percentage of the population living in urban areas affect the ease with which special interests can accomplish their lobbying goals. Jackson, Saurman, and Shughart (1994) showed that election term length affects legislative action to institute legal change. Maurizi (1974) and Graddy and Nichol (1989) found that state occupational licensing board members have an influence on the legislative process.

(23.) Recall from Table 2 that CNMBIRTHS are 5.76% of total births so that a 1-percentage-point increase would amount to an 18% increase in CNMBBIRTHS. Subsequent analysis makes use of this type of calculation.

(24.) These increases, percentage-point-wise, are 3.9 and 4.1, respectively.

(25.) Price equations were estimated, in part by using phone survey data, in preparatory econometric modeling for this study. In a supply-and-demand model, we found that when all regulatory variables (seven in that model) were evaluated at their respective means and at their sample minimums, the resulting effect of mean regulations (average price at about $2041) versus minimum regulations (average price about $1149) on CNMs is to decrease the average price of CNM services for an uncomplicated vaginal delivery by about $892, roughly a 44% decrease. Losses to CNMs and consumers as a result of mean regulations versus minimum regulations are approximately $184 million per year with deadweight losses estimated at $6.5 million per year. While small, such deadweight losses are not unexpected given the lowered price sensitivity engendered by third-party payments. These results are available from the authors on request.

References

Adams, Constance J. 1989. Nurse-midwifery practice in the United States, 1982 and 1987. American Journal of Public Health 77:1038-9.

Akerloff, George A. 1970. The market for lemons: Qualitative uncertainty and the market mechanism. Quarterly Journal of Economics 84:488-500.

American College of Nurse-Midwives. 1994. Nurse-midwives: Quality care for women and newborns. Washington, DC: American College of Nurse-Midwives.

American College of Nurse-Midwives. 1995. Nurse-midwifery today: A handbook of state legislation. Washington, DC: American College of Nurse-Midwives.

Butter, Irene H., and Bonnie J. Kay. 1988. State laws and the practice of lay midwifery. American Journal of Public Health 78:1161-9.

Carroll, Sidney L., and Robert J. Gaston. 1981a. A note on the quality of legal services: Peer review and disciplinary service. In Research in Law and Economics 3, edited by Richard O. Zerbe. Greenwich, CT: JAI Press, pp. 251-60.

Carroll, Sidney L., and Robert J. Gaston. 1981b. Occupational restrictions and the quality of service received: Some evidence. Southern Economic Journal 47:959-76.

Coburn, Karen Ann. 1997. The battle over home delivery. Governing 6:40-1.

Darby, Michael R., and Edi Karni. 1973. Free competition and the optimal amount of fraud. Journal of Law and Economics 16:67-88.

Department of Health and Human Services. 1992. A survey of certified nurse-midwives. Atlanta: Office of the Inspector General.

DeVries, Raymond G. 1985. Midwives, medicine, and the law. Philadelphia: Temple University Press.

Dueker, Michael J., Stephen J. Spurt, Ada K. Jacox, and David E. Kalist. 2000. The practice boundaries of advanced practice nurses: A panel study of U.S. states. Unpublished paper, Wayne State University.

Ekelund, Robert B. Jr., Franklin G. Mixon Jr., and Rand W. Ressler. 1995. Advertising and information: An empirical study of search, experience and credence goods. Journal of Economic Studies 22:33-43.

Gaumer, Gary L. 1984. Regulating health professionals: A review of the empirical literature. Milbank Memorial Fund Quarterly 62:380-416.

Graddy, Elizabeth. 1991. Interest groups or the public interest--Why do we regulate health occupations? Journal of Health Politics, Policy and Law 16:25-49.

Graddy, Elizabeth, and Michael B. Nichol. 1989. Public members on Occupational licensing boards: Effects on legislative regulatory reforms. Southern Economic Journal 55:610-25.

Greene, William H. 2000. Econometric analysis. 4th edition. Upper Saddle River, NJ: Prentice Hall.

Gujarati, Damodar. 1988. Basic econometrics. New York: McGraw-Hill.

Hausman, J. 1978. Specification tests in econometrics. Econometrica 46:1251-71.

Holden, Arlene S. 1978. The economics of dental licensing. Arlington, VA: Public Research Institute of the Center for Naval Analysis.

Izraeli, Oded, and Kevin Murphy. 1997. Convergence in state nominal and real per capita income: Empirical evidence. Public Finance Review 6:555-76.

Jackson, John D., David S. Saurman, and William F. Shughart. 1994. Instant winners: Legal change in transition and the diffusion of state lotteries. Public Choice 80:245-63.

Kennedy, Peter. t992. A guide to econometrics. 3rd edition. Cambridge, MA: MIT Press.

Leland, Hayne E. 1979. Quacks, lemons, and licensing: A theory of minimum quality standards. Journal of Political Economy 87:1328-46.

Maurizi, Alex. 1974. Occupational licensing and the public interest. Journal of Political Economy 82:399-413.

McCormick, Robert E., and Robert D. Tollison. 1981. Politicians, legislation, and the economy. Boston: Kluwer Academic.

McKie, James. 1970. Regulation and the free market: The problem of boundaries. Bell Journal of Economics and Management Science 1:6-26.

Nelson, Phillip. 1974. Advertising as information. Journal of Political Economy 82:729-54.

Peltzman, Sam. 1976. Toward a more general theory of economic regulation. Journal of Law and Economics 19:211-40.

Rottenberg, Simon. 1980. Introduction in occupational licensure and regulation. Washington, DC: American Enterprise Institute.

Safriet, Barbara J. 1992. Health care dollars and regulatory sense: The role of advanced practice nursing. Yale Journal on Regulation 9:417-88.

Sass. Tim R., and Mark W. Nichols. 1996. Scope-of-practice regulation: Physician control and the wages of non-physician health-care professionals. Journal of Regulatory Economics 9:61-81.

Sass, Tim R., and David S. Saurman. 1995. Advertising restrictions and concentration: The case of malt beverages. Review of Economics and Statistics 77:66-81.

Scupholme. Anne, Jeanne DeJoseph, Donna M. Strobino, and Lisa L. Paine. 1992. Nurse-midwifery care to vulnerable populations, phase I: Demographic characteristics of the national CNM sample. Journal of Nurse-Midwifery 37:341-8.

Shapiro, Carl. 1986. Investment, moral hazard, and occupational licensing. Review of Economic Studies 53:843-62.

Stigler, George J. 1971. The theory of economic regulation. Bell Journal of Economics and Management Science 1:3-21.

A. Frank Adams III, *

Robert B. Ekelund Jr., +

John D. Jackson ++

* Department of Economics, Kennesaw State University, Kennesaw, GA 30144, USA.

+ Auburn University and Trinity University (San Antonio), Department of Economics, 215 Lowder Business Building, Auburn University, Auburn, AL 36849, USA; E-mail bobekelund@prodigy.net; corresponding author.

++ Department of Economics. 215 Lowder Business Building, Auburn University, Auburn, AL 36849, USA.

We are grateful so Michael Dueker and his coauthors for sharing their unpublished manuscript on advanced practice nurses with us. We are, of course, liable for any errors in our paper.