Earnings and Quality Differentials in For-Profit

0 downloads 0 Views 215KB Size Report
with respect to age as a proxy of market experience and the qualifications as proxies of ... to choose working in nonprofit homes because they are more likely to be valued ..... employee with the highest earnings who is able to provide the best quality of .... for-profit facilities tend to hire a relatively large number of employees ...
ESRI Discussion Paper Series No.17

Earnings and Quality Differentials in For-Profit versus Nonprofit Long-Term Care: Evidence from Japan’s Long-Term Care Market

By Haruko Noguchi and Satoshi Shimizutani

December 2002

Economic and Social Research Institute Cabinet Office Tokyo, Japan

Earnings and Quality Differentials in For-Profit versus Nonprofit Long-Term Care: 1

Evidence from Japan’s Long-Term Care Market

By

Haruko Noguchi and Satoshi Shimizutani

December 2002

Part I: Are Nonprofit Earnings Differentials Observed in Japan? Evidence from Micro-level Data in Japanese Nursing Homes Part II: Quality Differentials in For -Profit versus Nonprofit Long-Term Care: Evidence from Japanese Micro-level Data

Please address correspondence to: Haruko Noguchi, Toyo-Eiwa University, Yokohama, Japan Email: [email protected] Satoshi Shimizutani, Cabinet Office, Tokyo, Japan Email: [email protected]

1

This research originated in a study on Japan’s long-term care conducted by the Price Policy Division of the Cabinet Office. We thank to Koichi Hamada, Reiko Kanda, Koichi Kawabuchi, Shuzo Nishimura, Takashi Oshio, and Wataru Suzuki for their comments. We also thank to Kaigo Roudou Antei Center for providing us valuable data. The views expressed in this paper do not necessarily represent those of the Economic and Social Research Institute or of the Japanese government.

PART I

ARE NON-PROFIT EARNINGS D IFFERENTIALS OBSERVED IN JAPAN? EVIDENCE FROM MICRO -LEVEL DATA IN JAPA NESE NURSING HOMES

Abstract This research is the first empirical assessment on the wage differentials between the nonprofit sector and profit sector in Japanese nursing homes, and makes use of the Jigyosho Ni Okeru Kaigo Rodo Jittai Chosa (Statistical Survey on Nursing Home Employees) conducted by the Care Worker Support Center Foundation (Kaigo Roudou Antei Center) in September 2000. Theoretically, those who support nonprofit wage premium emphasize the

asymmetry

of

information

frequently

observed

in

services

sectors

and

“non-distributional constraint” in the nonprofit enterprises while those who argue that nonprofit wage is lower than for -profit counterparts believe in the voluntary and altruistic characteristics of employees in the nonprofit sectors.

Previous studies on the nonprofit

premium in the U.S. are inconclusive. We present three major empirical findings. First, our empirical results support to confirm that there is nonprofit wage premium in the long-term care labor market in Japan before and after controlling for nonrandom unobserved self-selection bias.

Second, nonprofit firms are more likely to value workers

with respect to age as a proxy of market experience and the qualifications as proxies of education than for -profit firms.

Third, the long-term care market is segregated between

for-profit and not-for -profit facilities in Japan as a consequence of the presence of comparative advantage on workers’ self-allocation procedure. More experienced workers with higher level of education, therefore who may provide the higher quality of care, tend to choose working in nonprofit homes because they are more likely to be valued in the not-for-profit sector than the proprietary sector.

1

1. Introduction

The growing activities and enhanced presence of the nonprofit sector are remarkable in contemporary Japan. Those employed in the nonprofit sector account for up to 3.8 %2 of the total employment in 2000. The increasingly important role of nonprofit activities was accelerated by the introduction of public insurance in the long-term care industry in the spring of 2000, the most dominant nonprofit sector, together with the Law to Promote Specified Nonprofit Activities , effective from 1999. The implication of what the growing employment in the nonprofit sector will bring about for the labor market, however, is neither deeply understood and nor investigated. For those who emphasize the asymmetry of information frequently observed in the various types of services sectors, like Hansmann type argument (the “Contract Failure”) which argues for higher service equality and wage premiums in nonprofit sectors than those found in the profit sectors—the increased nonprofit activities might spur upward pressure on wages and raise the cost of long-term care in the future. On the other hand, this movement is considered to be a factor to motivate lower wages for those who believe in the voluntary and altruistic characteristics of employees in the nonprofit sectors. Thus, the expanding significance of the nonprofit sector poses a new challenge on wage -setting in Japan.

Empirical investigation of the earnings differentials between the profit and

nonprofit sectors is thus indispensable. Moreover, this is particularly meaningful for the current Japanese downward wage pressure under the historically high unemployment rate, resulting from the “Lost Decade” of the 1990s. This paper aims to investigate the wage differentials between the profit and nonprofit sectors in Japan, focusing on wage rates in Japanese nursing homes, one of the most dominant nonprofit sectors. The main object of the introduction of long-term care insurance is to provide long-term care for the compulsory insured. It also allows for -profit enterprises to enter the home-help market, where incumbent nonprofit enterprises (mainly, 2

Matsunaga, Y. (2002). Chapter 1: Macro estimation. In Yamauchi, N. (ed.), The Japanese nonprofit almanac 2002:The insights into the Japanese nonprofits from the latest data set (pp.7-20). Osaka, Japan: NPO Research Project, Osaka School of International Public Policy.

2

social welfare cooperation) occupy a large portion in total number. Put differently, both for-profit and non-profit enterprises began to compete in the same market, though for -profit companies are not allowed to opera te in the institutional care market. The mixture of different types of enterprises in the same market is expected to prevail in Japan, especially in the services sector. The empirical assessment of what the competition among these enterprises brings about thus has important policy implications. Nonetheless, to our best knowledge , there has been little research on the nonprofit earnings differentials in Japan, mainly due to the lack of availability of micro-level data on wages and the characteristics of employers and employees. We take advantage of the “Survey on Employees in Nursing Homes” conducted by the Care Worker Support Center Foundation (Kaigo Roudou Antei Center), which provides monthly, weekly, and hourly wages as well as comprehensive characteristics of individual employees. This valuable data has made it possible to address wage differentials for the first time. Our conclusion demonstrates that there is nonprofit wage premium in the long-term care labor market in Japan before and after controlling for nonrandom unobserved self-selection bias. Moreover, nonprofit firms are more likely to value workers with respect to age as a proxy of market experience and the qualifications as proxies of education than for-profit firms. In addition, the long-term care market is segregated between for-profit and not-for-profit facilities in Japan as a consequence of the presence of comparative advantage on workers’ self-allocation procedure. More experienced workers with higher level of education, therefore who may provide the higher quality of care, tend to choose working in nonprofit homes because they are more likely to be valued in the not-for-profit sector than the proprietary sector. This paper is organized as follows. The next section brief ly reviews some theoretical background and previous research on the nonprofit earnings differentials. Section 3 describes the empirical specifications to be used in this paper. Section 4 describes the data for our empirical work, showing a simple comparison of for-profit and not-for-profit sectors using the summary statistics. Sections 5 and 6 present and discuss the results. The last section concludes.

3

2. Previous research There has been a large body of literature on non-profit wage differentials in the United States. Most of these studies test the competing hypotheses that support either the wage premium in the nonprofit sector or that in the for-profit sector, derived from different behavioral characteristics of enterprises in these sectors. The representative argument on the nonprofit wage premium is Hansmann (1980). Hansmann provides the general theory of nonprofit enterprises. He argues that due to the asymmetry of information between consumers and providers, consumers are unable to judge the quality of services provided and the “non-distributional constraint” under which the nonprofit sectors are prohibited from distributing net earnings makes the nonprofit providers raise the quality of services to avoid opportunistic behaviors observe d in the profit sectors. More concretely, the non-distributional constraint provides two kinds of hypotheses on non-profit management to explain the wage premium. The first explanation, termed “philanthropic wage -setting” (Feldstein (1971)) or “attenuated property rights ” (Borjas et. al (1983)), argue that nonprofit managers have less incentive to lower wages since they do not have to make profits. The other explanation is that managers have less incentive to lower the quality provided since an increase in quality raises their utility in the not-for-profit sector (Newhouse (1970)). Moreover, other research attributes the nonprofit wage premium to the organizational characteristics of the nonprofit sectors such as exemption of corporate taxes or preferential treatment by the government (Frank and Salever (1994)), or soft budget constraints that allow inefficient management of nonprofit enterprises. On the contrary, there are some arguments to support the argument that nonprofit wages are less than the ir counterparts in the profit sectors. The representative of this view is the “labor donation” story, which states that employees in nonprofit sectors place lower value on money and higher value on non-monetary benefits such as working conditions or social responsibility (Rose-Ackerman (1996), Preston(1989)). Another explanation is that

4

nonprofit enterprises are concentrated in less-profitable sectors so that wages are lower in the nonprofit compared to the for-profit sector (Lakdawall et. al. (1998)). Many researchers have assessed quantitatively the wage differentials between the nonprofit sector and the for-profit sector, controlling for various characteristics of both facilities and employees. However, the results are inconclusive3 . One stream of empirical research is to investigate overall and within-industry differentials. Leete (2001) finds that zero or slightly positive economy-wide wage differentials between nonprofit and for-profit employees. The other stream is to investigate the wage differentials in specific sectors. 4 As for nursing homes and day care industries, Borjas et al. (1983) report that, based on the OLS estimates, the non-profit premium in nursing homes is consistent with the property right explanation. Preston (1988) finds it in federally regulated nonprofit day care centers. However, Holtman and Idson (1993) offer a counterargument on the previous findings. They explore why registered nurses in nonprofit facilities earn higher wages than their for -profit counterparts after controlling for sectoral selectivity bias. They follow the line of Hansman to argue that nonprofits compensate market failure caused by asymmetry of information and serve to provide a higher quality of care. Their finding on the selectivity pattern implies that the nonprofit wage premium can be interpreted as compensation for unobserved quality differences, rather than as the comparative advantage theory, which states that employees follow the criterion of choosing the sector that offers the highes t wage. They found that a random allocation procedure would redistribute some high quality workers to the for -profit sector and showed the differential quality hypothesis such that for-profit and nonprofit facilities are producing different products, which are the low and high quality of care. Despite its academic and practical importance, there is little research on the nonprofit wage differentials in Japan. This is partly explained by the unavailability of micro-level data on wages and employees’ characteristics. As previous research pointed out, it is 3

Ruhm and Borkoski (2000) provide a compact survey on empirical studies. Weisbrod (1983) and Goddeeris (1988) investigate wage differentials between private-profit and public -interest lawyers. 4

5

crucial for evaluating wage differentials after controlling for the characteristics of employees and facilities, because such heterogeneity plausibly affects observed wage differentials. To our knowledge, this study is the first attempt in Japan to assess the non-profit wage differentials based on micro-level data.

3. Empirical Specification

Assessing wage differentials between the for -profit and not-for-profit sectors based on simple linear regression, including a sector allocation dummy variable yields unreliable biased estimates, owning to the endogeneity problem such that the sector affiliation of workers is influenced by unobservable various characteristics that also affect wage rates 5. Namely, the nature of characteristics of workers who decide to work in either for -profit homes or not-for-profit homes is alleged to be systematically very different. Moreover, owners in these two sectors tend to hire workers with specific types of characteristics.

For

example, female home helpers with greater experience, having young children, working part-time or not regularly, are more likely to choose working in nonprofit rather than for-profit settings probably because not-for-profit organizations are possibly more concerned with the personal demands of workers, while the owners of for-profit homes might prefer less-experienced employees whom they can hire at relatively low wage rates. Suppose that the i th employee chooses to work in either the for-profit sector or the nonprofit sector to maximize utility defined as:

(1.1)

Vi1 = Wi1ω + Yi γ 1 + υ 1i

(1.2)

Vi2 = Wi 2ω + Yi γ 2 + υi2

The superscript 1 stands for for-profit and 2 stands for nonprofit. Vi1 and Vi2 represents the level of utility of the i th worker in a for-profit and nonprofit firm. 5

Wi1 and Wi 2

Another possible choice, that for workers not in the long-term care industry, should also be considered, but the data on workers who exit from the industry were not available.

6

indicate the natural logarithmic value of potential wage rate if the i th worker chooses to be self-selected into the for-profit sector or the nonprofit sector. Yi represents measurable factors and υ 1i and υi2 include unobserved factors, determining the i th worker’s utility. And ω , γ 1 , and γ

2

represent parameters. The i th worker is

expected to be self-allocated in the sector where he or she achieves greater utility, along with the selection rule as follows (Goddeeris (1988)):

(2)

Z *i = Vi1 − Vi2 = ( Wi1 − Wi 2 )ω + Yi (γ 1 − γ 2 ) + υi1 − υ i2

Z i = 1 if Z *i ≥ 0 or Vi1 ≥ Vi 2 Z i = 0 if Z *i < 0 or Vi1 < Vi 2

Z *i represents for the i th worker’s unobserved propensity to be allocated to the for-profit

sector, such that the observable dichotomous counterpart variable Z i = 1 , meaning the i th worker actually chooses a for-profit home, if Z *i ≥ 0 , and Z i = 0 , implying that the i th worker chooses a not-for-profit home, otherwise. In equation (2), the expected natural logarithmic wage of the i th worker can be shown as the following function:

(3.1)

Wi1 = X iß + ϕ 1 + τ 1i

(3.2)

Wi2 = Xiß + ϕ 2 + τ i2

X i includes various factors determining the i th worker’s wage rate, such as human capital characteristics. ß , ϕ 1 , and ϕ 2 are parameters to be estimated, where ϕ s are selection biases showing the potential wage differential between for-profit and nonprofit firms to be focused on in this study. τ 1i and τ i2 show unobservables that would influence his or her wage rates. Substituting (3.1) and (3.2) into (2), we obtain:

7

Z *i = Vi1 − Vi2 = (X i ß + ϕ 1 + τ 1i − X i ß − ϕ 2 − τ i2 )ω + Yi (γ 1 − γ 2 ) + υ 1i − υ i2

(4)

= (ϕ 1 − ϕ 2 )ω + Yi (γ 1 − γ 2 ) + (τ i1 − τ i2 )ω + (υ i1 − υi2 )

= ϕω + Yi γ + τ i ω + υi ϕω can be considered as a constant term and τ iω + υ i as u i , various unmeasured

factors that might affect workers’ choice on the sectoral allocation. Then, simplifying equation (4), in order to correct the influence of a nonrandom allocation of workers to either the for -profit sector or the not-for-profit sector, the model has to include a reduced equation with a self selection rule, as follows:

Zi* = Yiγ + u i

(5)

Zi = 1 if Z*i ≥ 0 or − u i ≤ Yi γ Zi = 0 if Z*i < 0 or − u i > Yi γ where

u i ~ N ( 0, σ u2 )

Again, we specify a separate log-linear wage equation for for-profit and not-for-profit sectors, as follows:

is observable if Z i = 1

(6.1)

Wi1 = X 1i π 1 + ε1i

(6.2)

Wi2 = X i2 π 2 + ε 2i

where

ε1i ~ N( 0, σ12 ) and ε 2i ~ N( 0, σ 22 )

is observable if Z i = 0

In these equations, all the factors that determine workers’ allocation do not necessarily influence wage rates. In other words, Yi and X i are not essentially the same.

For example, varia bles describing the family backgrounds of workers, such as

marital status and number of children, would have significant effects on the sector selection, yet they would not be pertinent to wage rates. In contrast, variables that represent the

8

workers’ status and occupation in homes after the choice was made would influence wage rates, but they might not be relevant to the selection. Since these variables relevant to one equation and irrelevant to another are not available on the data used for the empirical analysis in this study. We will, however, control for the selection and wage equations using the same explanatory variables. So as to obtain unbiased estimates of parameters, we will apply Heckman’s two-stage approach (Lee (1978), Heckman (1979), Willis and Rosen (1979), Greene (1981)).

First,

maximum likelihood method is used for the probit regression equation to estimate the parameter vector γ of the probability that Z *i ≥ 0 , implying the likelihood for the i th worker of being allocated to a for-profit home in Equation (5). Then, for each worker, using probit coefficient vector γ , we will be able to obtain Z *i and hence the inverse of Mill’s ratio, λ i or the truncated mean of the normal density with truncation point Yi γ / σ u , due to selection. Including λ i in the regression along with potential determinants of wage rates will correct truncation and selectivity bias. Therefore, second, rewriting equations (3.1) and (3.2), a linear regression will be performed for Wi on X i and λ i in the for-profit and nonprofit sector, separately, as follows:

(7.1)

Wi1 = X 1i δ1 + λ1i θ 1 + ω1

(7.2)

Wi2 = X i2 δ 2 + λ2i θ 2 + ω 2

where

E (ω i1 ) = E (ω i2 ) = 0

6

λ1i = − φ(Yi γ / σ u ) / Φ( Yi γ / σ u ) and λ2i = φ( Yi γ / σ u ) /[1 − Φ (Yi γ / σ u )] θ1 = σ 1u / σ u = cov( ε1i , u i ) / σ u and θ 2 = σ 2 u / σ u = cov( ε 2i , u i ) / σ u

Note that regression estimates of the variances of ω and ω based on the least square residuals 1 2 are downward-biased estimates compared to the true variance of εi and ε i , because the error 1

6

2

structure is heteroskedastic as a result of the nature of truncated data and λ i s are unknown (Heckman (1979) and Lee (1978)). Consistent estimates of Equations (7.1) and (7.2) can, however, be obtained by replacing predicted λ i based on the reduced-form probability equation using a probit technique.

9

For the above equations for truncated mean λ i , φ and Φ indicate the standard normal probability density and the standard normal cumulative density functions, respectively. Due to the nature of the model setting and the truncated observations in our data, it is unlikely to estimate the residual in the selectivity equation, σ u . If σ u is standardized and assumed to be 1, then the coefficient of truncated mean adjustment factors for selection bias ( θ1 and θ 2 ) can be directly interpreted as the covariance between the residuals in the selectivity equation and the residuals in the wage equations, measuring the relation between wage rates and the unobservable characteristics that influence the sector allocation. Consequently, θ1 and θ 2 are key variables to examine our hypothesis based on asymmetric information caused by the coexistence of the for -profit and nonprofit sectors in the nursing facility industry. If the selection behavior among workers in the long-term care labor market depends merely on the well-known comparative-advantage hypothesis, then they will choose to work in the sector that rewards their characteristics the most, such as ability, talent, education, and experience. The comparative-advantage argument suggests the condition such that θ1 > 0 if unobserved factors have impacts on both the likelihood of being in the for-profit sector and wage rates in the same direction, while it is θ 2 < 0 if unobservables have a positive influence on the probability of being in the for-profit sector and if they have a negative effect on wage rates, or vice versa. If θ1 > 0 and θ 2 < 0 are met, then the employee with the highest earnings who is able to provide the best quality of services in the for-profit sector would also do the best performance with the greatest wages in the not-for-profit sector, along with a strictly hierarchical self-allocation pattern based on ability. In other words, θ1 > 0 implies a positive selection bias among workers in the for-profit sector, such that E[W i1 | Z *i ≥ 0] > E(Wi1 ) , or those who earn greater than predicted wages derived from observed factors of workers in for-profit homes are more likely to be working in the for-profit sector than the predicted probability. In contrast,

θ 2 < 0 indicates a negative selection bias among nonprofit workers, such that

E[Wi2 | Z*i < 0] > E(Wi2 ) , or those who earn greater than observed predicted wages in the

10

nonprofit sector, are less likely to be allocated to the for -profit sector than the expected prospect (Willis and Rosen (1979), Holtmann and Idson (1993)). Instead of the particular directions for θ , Trost (1981) claims that θ1 > θ 2 , or the relation between the unobservable characteristics that influence the sector allocation and wage rates is greater in the for -profit than nonprofit sector, as the necessary condition for the comparative-advantage hypothesis. Explicitly, E[Wi1 | Z*i ≥ 0] > E[Wi2 | Z*i ≥ 0] , for-profit workers with greater than conditional mean wages of nonprofit employees if they decide to work in the for-profit sector are more likely to be allocated to for-profit homes than the expected probability; and E[Wi2 | Z*i < 0] > E[Wi1 | Z*i < 0] , not-for-profit workers with greater than conditional mean wages of for-profit workers if they choose to work in the nonprofit sector are less likely to be allocated to the for-profit sector than the expected prospect. The statistical significance of θ , such that the hypothesis θ = 0 can be rejected by t-test, shows whether or not workers’ observed characteristics in the model captures unobservable effects on sector allocation and wage rates better, compared to no sector allocation adjustment in the nursing care industry.

4. Data

The empirical analysis of this study is based on the Jigyosho Ni Okeru Kaigo Rodo Jittai Chosa (Statistical Survey on Nursing Home Employees, administered in September 2000), conducted by the Care Worker Support Center Foundation (Kaigo Roudou Antei Center). This survey is not based on a random sample of the population, including data on various characteristics of approximately 39,261 employees in 1,347 nursing facilities all over Japan, which are self-enumerated by the nursing home’s administrator. In this study, we choose a sample of 24,967 employees, due to the lack of information on ownership and various indicators used in our empirical analysis. There are seven major types of ownership in the industry: for-profit, and six not-for-profit facilities—social welfare cooperation, medical cooperation, authorized Non-Profit Organization (NPO), Co-op, agricultural cooperative, and other charitable

11

corporation. In this survey, 2,823 (11.3 percent) and 22,144 (88.7 percent) employees are working at proprietary and nonprofit nursing homes, respectively. Table 1 describes some basic statistics of major characteristics of both employees and facilities that could affect wages in for-profit and nonprofit nursing homes. This indicates significant differences between these sectors in the nursing industry in Japan. Mean daily wages are 4.7 percent higher in nonprofit facilities (the mean wage is 9,634 yen in for-profit homes and 10,217 yen in not-for-profit homes). Not-for-profit facilities are more likely to hire male employees in relatively young age groups from 20 to 49 years old. Also, nonprofit facilities are seen to prefer either full-time or part-time but regular workers. Regarding employees’ job qualifications 7 , a higher percentage of workers who are qualified to perform advanced skills, e.g., physicians, staff nurses, social workers, social care workers, and physical therapists are employed in nonprofit homes. On the contrary, a higher percentage of home helpers with any degree of skill, care managers, and service managers are hired by for -profit homes. Unfortunately, we do not have such information as experience, education, marital and family status, which are controlled for in previous works.

However, employee age can be considered as a proxy of years of experience and

the qualification may reflect education, which might indicate the selectivity of both employers and employees influencing the mean wage between two sectors in the nursing industry. Instead, using marital and family status as proxy indicators for employees’ stability in previous studies (Holtmann and Idson (1993), and Mocan and Tekin (2000)), we may use employees’ full-time status as a proxy of the stability as well as a facility-level aggregated indicator in the survey directly measuring the stability of full-time and part-time employees at facilities. For-profit homes currently hiring workers claim greater instability of employees for either full-time or part-time status, as compared to nonprofit homes; – 13.2 percent versus 7.5 percent for full-time and 21.1 percent versus 14.6 percent for part-time employees, respectively. Also, a higher percentage of for-profit homes are seen to face difficulty in hiring both full-time and part-time workers than nonprofit homes. 7

In Table 1, a worker with two and more qualifications is double- (or multiple-) counted so that the total percentage points for qualification dummies are not one hundred.

12

We include various facility characteristics 8 that might affec t wage rates. First, for-profit facilities tend to hire a relatively large number of employees compared to not-for-profit, though this is not the case for for-profit homes with fewer than 29 employees. Hence, higher wage rates in the nonprofit sector tha n in the for-profit sector cannot be explained simply by the positive relation between employer size and wage rates that a large number of studies have shown (Brown and Medoff (1989), Idson and Feaster (1990), Troske (1999)).

The second notable attribute of facility characteristic is that non-wage

benefits–employment insurance, work-related accident compensation insurance, health insurance, and public pensions –are more often offered by for -profit homes than by nonprofit facilities. Probably the cost-sensitive managers in the for-profit sector are motivated to lower wage rates. On the other hand, nonprofit homes are seen to support employees differently from for -profit homes; additional pay to basic salary, professional training inside and outside the facility, health examinations, subsidies for uniforms and equipment, and preventive care for work-related back pain and accidents. Only a few papers pay attention to these factors for non-wage benefits and work-related supports to employees, but they would be key indicators to control for measurement error in wage rates of workers, in particular, on a part-time basis (Montgomery and Cosgrove (1995), and Mocan and Tekin (2000)). Our data shows that, on average, about 70 percent and 50 percent of full-time employees are offered any type of insurance and public pension by for-profit and nonprofit employers, respectively, while the part-time employees of either sector are much less likely to own these non-wage benefits. Although we do not focus on 8

Nursing homes are often producing various community-based long-term care in addition to institutional care, e.g. at-home help, bathing, nursing care, outpatient treatments, guidance to care management at home, short-term stay, and rental care equipment. For-profit nursing homes are more likely to supply daily needs to the elderly, rather than intensive long-term care such as rehabilitation, nursing care, and medical treatments often performed at special nursing homes operated merely by nonprofit organizations due to legal restrictions. Thus, employers in two sectors are demanding workers with different types of job qualifications or skills possibly so that higher wage rates in the nonprofit sector may possibly be explained by the demand for more advanced skilled labor rather than less-skilled labor. Therefore, these variables are not included into our regression analysis because they may lead to biased results due to multi-collinearity caused by the strong correlation between products, facilities provided, and employees ’ qualifications. In order to avoid multi-collinearity, we applied the same sectoral allocation probit method to the segregated sample by qualifications- home helpers care managers, staff nurses, and service managers. We discuss this result in our conclusion.

13

a wage differential between full-time and part-time employees in this study, non-wage benefits can be considered as labor costs for any firms; they are likely to be significant determinants on wage rates. To examine the effects of these factors on wage determination, we first run a simple semi-logarithmic human capital wage function for a pooled full sample, using a dummy variable that indicates whether or not the ownership is proprietary. Then, to examine our suspicion that the labor market in the nursing home industry may be completely segregated between two sectors on account of not the difference in the quality but contents of services they are producing, we perform the same analysis separately by employees’ qualification9 . The results appear in Table 2. After adjusting for all characteristics in Table 1, we found a negative and statistically significant for-profit wage differential by –1.6, –5.3, –3.7, –2.9, and –9.3 percentage points for full-sample, helpers, care managers, staff nurses, and service managers, respectively, that is consistent with previous findings for the nursing home industry (Borjas, Frech III, and Ginsburg (1983), Holtmann and Idson (1993), Ruhm and Borkoski (2000)). Consequently, a wage differential among employees with the same qualification is seen to be in favor of not-for-profit workers, regardless of whether the qualification requires advanced or less -advanced skills and education. Therefore, we may still pursue an examination of the relation between wage rates and the quality of service in the for-profit and nonprofit sectors. Note that adjusted for -profit wage differentials for helper and staff nurses become even larger than unadjusted ones. This implies that the standardized percentage difference for either home helpers or staff nurses with similar characteristics between two sectors is even larger than the gross for-profit effects. For evaluating wage differentials across ownerships in the nursing home industry, however, simple least square regressions are not appropriate since the allocation of workers with various characteristics in our data to the for-profit and nonprofit sectors is nonrandom. When examining sample selection in our data, a Chow test also proves that we can reject 9

As regards physicians, social workers, social care workers, and physical/occupational therapists, the regressions adjusted for all characteristics are not statistically significant. We could not find valid wage differentials between for-profit and not-for-profit facilities.

14

the hypothesis that equations for for-profit and nonprofit homes are the same at more than 99.9 percent probability level where the F-statistic is 119.5. Therefore, applying for Heckman’s two-stage approach (Lee (1978), Heckman (1979)) is an appropriate procedure to obtain reliable estimates on the effects of a variety of individual and firm factors on the sectoral allocation and wage rates. There are only a few studies that address self-selection of workers into different sectors, in particular, with control for firm effects.

5. Results Table 3 shows the estimated coefficients of the for-profit selection equation and wage equations, including workers with any qualifications. The results of the sector allocation probit regression in the first column basically replicate unadjusted comparisons between the two sectors in Table 1. Workers in the age groups from 40 to 49 years old, who are supposed to have relatively long labor market experience and relatively advanced levels of skill or education, remain less likely to self-select into proprietary homes, which is consistent with the simple comparison between the two sectors. Since full-time workers are more likely to be self-allocated into the nonprofit sector, profit-seeking facilities might face in-short and unstable supply of full-time workers needed so that they are more likely to be hiring full-time workers. The difference in various facility characteristics such as firm sizes, professional training and supports to employees between two sectors largely reflects our simple comparison analysis. However, the probit sectoral allocation setting underlines the for-profit nursing homes’ characteristics of the compensation policy for the employees more clearly than our summary statistics. Basically, employees are more likely to be offered any type of insurance as non-wage compensation in proprietary homes, while this is not the case for public pensions, used mainly for supporting living expenses after retirement. According to the “attenuated property rights” assumption, efficient managers in for -profit facilities who offer solely work-related non-wage insurances to attract stable workers in the liquid labor market would not be tolerated for non-wage benefits that are not directly

15

relevant to current job performance such as public pensions; they might even have incentives to lower wages instead. The remainder of columns in Table 3 show the estimated coefficients of wage equations for proprietary (7.1) and not-for-profit homes (7.2), corrected for self-selection bias.

Although, in general, male and/or full-time employees are seen to be more

appreciated in both sectors, the direction, magnitude, and significance of a number of important indicators ’ effects on wage rates vary between for -profit and not-for-profit facilities.

Wage rates are seen to increase proportionally along with the age until 59 years

old in either the for-profit or nonprofit sector. Assuming workers’ age as a proxy of years of experience, the effect is larger in the not-for-profit than the for -profit settings, which is consistent with the results of previous studies. It is interesting that, regardless of stability, the hiring status of full-time and part-time workers has different influences on wage rates between the for-profit and nonprofit sectors. As we discussed, due to the short and unstable supply of full-time workers for -profit facilities might be facing, full-time workers are more appreciated than part-time workers within the profit-seeking sector. Compared to the nonprofit sector, however, the proprietary sector rewards full-time employees less and part -time employees more 10. As regards qualification dummies, home helpers with the first or third degrees of skill, physicians, staff nurses, social care workers, physical and occupational therapists who are less likely to self-select into the for -profit sector are valued relatively high in nonprofit

10

Even though both full-time and part-time workers are positively valued in both sectors, the hiring status has a positive impact on wage rates of full-time workers in the for-profit sector, but negative in the nonprofit sectors. Based on a result of the sector allocation probit equation (1), we found that full-time workers are seen to be less likely to be self-selected into for-profit homes and more likely to be unstable in the labor market of the private sector. To attract stable workers in the labor market, therefore, profit -seeking facilities demanding full-time workers tend to value them more than homes where the demand is met. On the other hand, in the labor market of the nonprofit sector, facilities that claim demanding full-time workers would be able to find workers at lower wage rates than facilities that are not hiring people, probably because qualified workers may have already been employed and the rest of workers may have to compete for fewer positions in the nonprofit nursing care industry. In contrast to full-time workers, profit-seeking homes demanding part-time workers would be able to attract them even at lower wage rates and not-for-profit homes would have to offer higher wage rates to meet the demand for part-time workers. This may reflect the difference in part-time workers’ labor supply between the for-profit and not-for-profit sectors.

16

settings. Moving down to the effects of facility characteristics, consistent with previous studies, wage rates become greater along with the firm size, and the statistical significance is slightly higher in for -profit than nonprofit homes. As regards non-wage benefits, for either full-time or part-time workers, proprietary facilities offering public pensions would lower wage rates. The provision of employment insurance and wages are appeared to be trade-offs in nonprofit homes. In general, non-wage compensations other than insurance such as professional trainings and various supports to employees seem to have negative influences on wage rates in both sectors. However, the results are slightly more significant in for-profit than in nonprofit homes. Consequently, it seems that nonprofit facilities would recompense workers’ identifiable experience, some human capital with advanced skills, and their stability more than for-profit homes, while some non-wage compensation benefits appear to lower wage rates in both sectors. Finally, it would be important to discuss that the coefficients of truncated mean adjustment factors for selection bias, θ1 and θ 2 in the wage equations, to test the hypothesis that the labor market in the nursing home industry is segregated between the for-profit and not-for-profit sectors providing different quality of services. The estimates in this study show that both θ1 and θ 2 are significantly positive, which implies that unobserved factors extensively affect employees’ allocation and wage rates. Further, the necessary condition ( θ1 > θ 2 ) for the comparative advantage theory suggested by Trost (1981) is met. θ1 > θ 2 shows that the relations between unobserved factors and both sectoral allocation and wage rates are larger in the proprietary than nonprofit sectors. Therefore, our estimates on the coefficients of truncated mean adjustment factors for selection bias are seen to demonstrate that there is a self-selection pattern between two sectors along with comparative advantage among workers who tend to be choosing the sector offering the highest wages. This implicitly means that the mean wages of employees who are actually self -allocated into profit facilities are greater than wages that would have been observed if they chose to work in not-for-profit facil ities.

On the

contrary, the mean wages of employees who in fact chose to work in the nonprofit sector

17

are greater than wages that would have been observed if employees with observed similar characteristics were self-allocated into the profit sector. The result is entirely inconsistent with the strict hierarchical sorting assumption such that employees who do better performance than average in the proprietary sector would always do better than average in the nonprofit sector. In other words, they are valuating seemingly similar workers with various unobserved characteristics somehow differently in a way that nonprofit homes value workers with longer market experience and higher level of skills or education who suppose to provide better quality of care than for-profit homes. Therefore, we can conclude that the market is segregated between for -profit and not-for-profit facilities that are assume to be providing different quality of services. Since proprietary firms just started entering in the long-term care market, we suspect that the quality of service in the for-profit sector has not caught up yet with the one in the nonprofit sector.

6. Factor decomposition analysis of wage differentials between two sectors

Based on the estimates from the wage equations corrected for self -selection bias in Table 3, we will apply a simple decomposition analysis to the wage differential, in order for distinguishing the contributions of attributes from differential contributions in valuation of key explanatory factors between for -profit and not -for-profit homes to the nonprofit wage premium.

(8)

where

The total effects are decomposed in the following way:

∑ W − ∑ W = ∑ (X δ + λ θ ) − ∑ (X δ + λ θ ) = {∑ δ 2 (X 2i − X1i ) + ∑ θ 2 (λ2i − λ1i )}+ {∑ (δ 2 − δ1 )X1i + ∑ (θ 2 − θ1 )λ1i } 2

1 i

i

{∑ δ (X 2

2 i

)

2 i

2

2 i

2

1 1 i

1 1 i

( )} is the summation of each factor ’s qualitative {∑ (δ − δ )X + ∑(θ − θ )λ } is the sum of differential

− X1i + ∑ θ 2 λ2i − λ1i

differential contributions and

2

1

1 i

2

1

1 i

contributions in valuation of each explanatory variable. The results are shown in Tables 4 and 5. Looking at Table 4, the summation of wage differentials between nonprofit and

18

for-profit sectors over all explanatory variables is 17.8 percentage points (433 yen in real value) after controlling for unobserved self-selection bias, which is different from a simple linear estimate without controlling for self-allocation that might be biased in Table 2. The characteristics of employees and facilities (attributes) and the sector valuation account for 3.2 percentage points (–437 yen) and 14.5 percentage points (870 yen) of total, respectively. Table 5 shows the decomposition of sectoral wage differentials in real value by factor, where positive (or negative) values enlarge (or reduce) total nonprofit wage differential. In other words, without the positive (or negative ) contribution of a factor, the total nonprofit wage premium will be reduced (or enlarged). Both qualifications and various non-wage supports to employees other than insurance are seen to contribute to the higher observed wage in not-for-profit facilities. For qualifications, the magnitude of effect is largely due to the attributes, whereas, for non-wage supports, it is mainly due to higher value on the characteristics. Age as a proxy of experience does also increase wages in nonprofits above proprietary homes, which depends on higher valuation on the attributes in the not-for-profit sector. On the other hand, the rest of the factors are affected in the opposite way: the pattern of male, full-time status, and most notably facility characteristics such as size of facility, insurance, hiring status, and professional training. For these factors, for -profit facilities value these attributes much higher than in the nonprofit sector. Therefore, wages in the proprietary sector are seen to grow at a faster rate with size, the presence of insurance, current hiring status, and the provision of professional training, than in the nonprofit sector. These are consistent with the previous empirical results shown by Idson and Feaster (1990) and Holtmann and Idson (1993) 11. Focusing on the selectivity, we found that the total effect of selectivity is negative (-592 yen) and the net of selectivity (reducing the effect of selectivity from the total effect) is 1025 yen. Also, since the truncated mean is positive in the for -profit sector, the positive 11

Our results, however, draw a complicated picture. Nonprofit managers are less likely to be cost-sensitive in total under the non-distributional constraint. They are, therefore, more likely to value workers’ productivity associated with the quality of services related to age, qualifications, and various non-wage supports other than insurances. Alternatively, we cannot fully support the property-rights hypothesis because employees in not-for-profit facilities are less rewarded than for-profit homes, with respect to facility-related characteristics.

19

coefficient on λ1i entails that wages for the average observed self-selected for-profit employees are higher than that would be observed if employees are randomly allocated across two sectors. This implies that a random allocation of employees would lead to even lower wages in the for-profit sector relative to wages in the nonprofit sector and to enlarge the wage differential between two sectors. In contrast to the results of Holtmann and Idson (1993), our result shows that a large number of workers who actually chose to work in nonprofit homes according to their comparative advantage are redistributed to for-profit institutions by random reallocation procedure, decreasing for-profit relative to nonprofit wages. As shown in Table 5, this negative selectivity effect is mostly contributed by differential characteristics rather than sector valuation of the selectivity that performs in the positive direction, which are canceled with each other, eventually helping to expand the wage differential between two sectors as total. We would like to conclude that the presence of a larger number of workers who have unmeasurable comparative advantage in the not-for-profit relative to for-profit sectors would be a more serious cause of nonprofit premium rather than the not-for-profit sector ’s lavish and inefficient rate of return to workers in the long-term care market in Japan.

7. Conclusion Unlike the long-term care market in the United States, the Japanese nursing care market had been legally restricted to be run by the nonprofit sector. The recent introduction of the long-term care insurance lastly allows for -profit facilities to enter only the home-help market and for-profit and not-for-profit firms have just started to compete in the long-term care market in Japan. The current strong political incentive for the Japanese government to restructure administrative and industrial organizations will lead to accelerating the entry of for -profit firms into the conventional nonprofit sector and to stimulating competition between different types of firms in the same market. Under such a drastic change in socio-economic circumstance, employment in the labor market where both for-profit and nonprofit sectors coexist will also expand. Therefore, the empirical

20

assessment of the introduction of for-profit into the nursing industry as a social experiment gives us significant clues regarding policy implications from various aspects. In this paper, we focus on the differences in economic behavior on wage setting between for-profit and nonprofit facilities. We hypothesize that the nonprofit wage premium in the long-term care market may be accounted for by the segregated labor market between two sectors providing observed and unobserved different quality of services owing to “non-distributional constraint.” Under “non-distributional constraint,” nonprofit homes are prohibited from distributing net earnings, which makes nonprofit providers raise the quality of services to avoid the opportunistic behavior often observed in for -profit facilities. Since long-term care is a labor -intensive product, the difference in quality of workers between for-profit and nonprofit facilities is the most likely to clarify the quality of product provided. We have a couple of empirical and theoretical challenges in this context. An empirical challenge is sample selection bias caused by the endogeneity problem such that both sector affiliation of workers and wage rates are correlated with various unobserved individual and facility characteristics. We present three major empirical findings. First, our empirical results support to confirm that there is nonprofit wage premium in the long-term care labor market in Japan before and after controlling for nonrandom unobserved self-selection bias.

Second, nonprofit firms are more likely to value workers

with respect to age as a proxy of market experience and the qualifications as proxies of education than for-profit firms. Third, the long-term care market is segregated between for-profit and not-for-profit facilities in Japan as a consequence of the presence of comparative advantage on workers’ self-allocation procedure. More experienced workers with higher level of education, therefore who may provide the higher quality of care, tend to choose working in nonprofit homes because they are more likely to be valued in the not-for-profit sector compared to the proprietary sector. Our theoretical challenge is related to how to measure and interpret the quality of services provided by both sectors. Most previous studies positing that higher quality of products in the not-for -profit sector produced by high-quality workers might account for

21

nonprofit wage premium control for various individual and facility characteristics like education, market experience, marital and family status, full-time or part-time status, facility affiliation, non-wage compensations, and professional supports, and so forth, as observed quality measures (Preston (1988), Preston (1989), Holtmann and Idson (1993), Mocan and Viola (1997), Mocan and Tekin (2000), Leete (2001)). Although proprietary firms have just started entering into the long-term care market, we observe that several verifications that nonprofit facilities might be producing better quality of care than proprietary institutions. For example, those who are supposed to have relatively long market experience (measured by age as a proxy indicator ) and advanced level of skills and education (assessed by qualification as a proxy factor) are more likely to be self-allocated into the nonprofit sector and they are unsurprisingly given a greater rate of return for their high-quality work in not-for-profit firms. We also suspect that for -profit and nonprofit institutions would be producing completely different contents of products, rather than high and low quality of services, in the sense that for-profit nursing homes are more likely to supply daily needs to the elderly, as opposed to intensive long-term care such as rehabilitation, nursing care, and medical treatments often performed at special nursing homes operated merely by nonprofit organizations due to legal restrictions. To avoid multicollinearity, dummy variables for contents of services are not included into our previous estimation procedure. As regards home he lpers, care managers, staff nurses, and service managers that appear to have nonprofit wage premiums based on simple semi-logarithmic wage functions, we applied the same empirical procedure separately, controlling for service dummies. Tables 6 and 7 show the results. These separated regression analyses clarify a significant aspect more clearly than our previous over-all regression; that is, for-profit firms might unfairly value both observed and unobserved productivity-related workers ’ attributes, leading to a much larger number of workers with comparative advantage in the not-for-profit sector. After correcting for selection bias, the summations of wage differentials between nonprofit and profit sectors over all explanatory variables are 1.8 percent (1,006 yen) for home helpers; -1.7 percent

22

(-226 yen) for care managers whose nonprofit wage differential turns negative; 3.9 percent (884 yen) for staff nurses; and 18.3 percent (2,161 yen) for service managers. In particular for home helpers and staff nurses, the sector valuation on their attributes related to productivity such as age, full-time status, and professional training, which are supposed to guarantee the quality of service provided, appears to contribute to the majority of the nonprofit wage premium. In contrast, the sectoral valuation on service contents is slightly higher for home helpers or even largely lower for staff nurses in nonprofit homes, relative to for-profit facilities, implying that the nonprofit sector is not necessarily behaving inefficiently. Nevertheless, focusing on the selectivity, again we found that the total effect of selectivity is slightly negative, lending to expand the wage differential between two sectors. Examining our suspicion that the presence of unmeasured comparative advantage in the nonprofit to for -profit sector remains a crucial challenge for further theoretical and empirical investigations. Further, we should continue to investigate empirically how the difference in wage setting probably associated with the quality of care in the for-profit sector and the nonprofit sector will be changing after proprietary firms settle down in the long-term care industry in Japan.

23

References

Borjas, George J., Frech III, H.E., and Ginsburg, Paul B.(1983). “Property Rights and Wages: The Case of Nursing Homes.” Journal of Human Resources, vol.18, pp231-246.

Brown, C. and Medoff, James (1989). “The Employer Size Wage Effect.” Journal of Political Economy, vol.101, pp483-496

Frank, Richard G. and Salkever, David S. (1994). “Nonprofit Organizations in the Health Sector.” Journal of Economic Perspective, vol.8, pp129-144.

Feldstein, Martin (1971). The Rising Cost of Hospital Care. Information Services Press. Washington D.C.

Goddeeris, John H. (1988). “Compensating Differentials and Self -Selection: An Application to Lawyers.” Journal of Political Economy , vol.96, pp411-428.

Greene William (1981). “Sample Selection Bias as a Specification Error: Comment” Econometrica 49, pp795-798.

Hansmann, Henry (1980). “The Role of Nonprofit Enterprise.” Yale Law Journal, vol.89, pp835-901.

Heckman, James (1979). “Sample Selection Bias as a Specification Error.” Econometrica, vol.47, pp153-162.

Holtmann, A.G., and Idson, Todd L. (1993). “Wage Determination of Registered Nurses in Proprietary and Nonprofit Nursing Homes.” Journal of Human Resources, vol.28, pp55-79.

24

Idson, Todd, and Daniel J. Feaster (1990) “A Selectivity Model of Employer-Size Wage Differentials.” Journal of Labor Economics, 8(1), pp99-122.

Lakdawall, Darius and Philipson, Tomas (1998). “Nonprofit Competition and Production. ” NBER Working Papers, no.6377.

Lee, Lung-Fei (1978) “Unionism and Wage Rates: A Simultaneous Equations Model with Qualitative and Limited Dependent Variables.” International Economic Review, 19(2), pp415-33.

Leete, Laura (2001). “Whither the Nonprofit Wage Differential? Estimates from the 1990 Census.” Journal of Labor Economics, vol.19, pp136-170.

Mocan, H. Naci and Deborah Viola (1997). “The Determinants of Child Care Worker ’s Wages and Compensation: Sectoral Difference, Human Capital, Race, Insiders and Outsiders,” NBER Working Papers, no.6328.

Mocan, H. Naci and Tekin, Erdal (2000). “Nonprofit Sector and Part-Time Work: An Analysis of Employer-Employee Matched Data of Child Care Workers.” NBER Working Papers, no.7977.

Montgomery, Mark and Cosgrove, James (1995). “Are Part Time Women Paid Less? A Model with Firm Specific Effects.” Economic Inquiry, vol.38, pp119-133.

Preston, Anne E.(1988). “The Effects of Property Rights on Labor Costs of Nonprofit Firms: An Application to the Day Care Industry. ” Journal of Industrial Economics, vol.36, pp337-350.

Preston, Anne E.(1989). “The Nonprofit Worker in a For -Profit World.” Journal of Labor

25

Economics , vol.7, pp438-463.

Ruhm, Christopher J., and Borkoski, Carey (2000). “Compensation in the Nonprofit Sector.” NBER Working Papers, no.7562.

Troske, Kenneth R. (1999). “Evidence on the Employer Size -Wage Premium from Worker-Establishment Matched Data.” Review of Economics and Statistics, vol.81, pp15-26.

Trost, Robert P. (1981). “Interpretation of Error Covariances with Non-Random Data: An Empirical Illustration of Returns to College Education.” Atlantic Economic Journal, 9(3), pp85-90.

Weisbrod, Burton A. (1983). “Non-Profit and Propr ietary Sector Behavior: Wage Differentials Among Lawyers.” Journal of Labor Economics, vol.1, pp246-263.

Willis, J. Robert and Sherwin Rosen (1979). “Education and Self -Selection. ” Journal of Political Economy, 87 (supplement), ppS7-S36.

26

Table 1: Key Variable definitions and summary statistics

For-profit (n=2,823) Variable

Nonprofit (n=22,144)

Definition

*: Reference indicator to be excluded from regression

Mean

Standard deviation

Mean

Standard deviation

I. Employees' characteristics natural log of daily wage logdw =1 if male q5_021 =1 if 20