Do Minimum Quality Standards Improve Quality? A

Case Study of the Nursing Home Industry�

Haizhen Liny

Abstract

Minimum quality standards are used extensively to solve quality problems in mar-

kets su¤ering from asymmetric information. Using a unique national panel over the

1996 - 2005 period, this paper estimates the impact of minimum sta¢ ng requirements

on the United States nursing home market. To consistently identify the impact of

regulatory policies, various speci�cations are employed and compared to provide com-

prehensive controls for unobserved heterogeneity across states and time. We �nd sig-

ni�cant preference for the dynamic speci�cation as compared to the �xed e¤ect and

the random trend speci�cation. Our result shows that given an increase of 0.5 hours of

minimum sta¢ ng of licensed nurses, the quality of patient care is increased by 15 per-

cent. Minimum sta¢ ng requirements for direct care nurses do not have any signi�cant

e¤ect on quality. Detailed explanations for this lack of impact are also discussed.

�I am deeply indebted to Marc Rysman and Randall Ellis for their guidance in this paper. I also thankCharlene Harrington, Chun-Yu Ho, Kevin Lang, Ching-To Albert Ma, Robert Margo, Claudia Olivetti,Pravin Trivedi and seminar participants at BU microeconomics dissertation workshop, BU empirical microworkshop, 5th International Industrial Organization Conference, and 34th Conference European Associationfor Research in Industrial Economics for helpful suggestions and comments.

yBusiness Economics and Public Policy, Kelley School of Business, Indiana University. Email:hzlin@indiana.edu.

1 Motivation

In recognition of the asymmetric information problem, government regulators, both at the

federal and state level, have imposed minimum quality standards (MQS) on nursing homes in

the U.S. The objective of this regulation is to reduce ine¢ ciencies caused by informational

asymmetries, and to increase quality of patient care. Whether or not this goal has been

achieved, and to what degree, is an important research question given the aging population1

in the U.S. and the widespread concern about low quality of care in nursing home facilities.

This paper is among the �rst to use a national panel to empirically examine the impact

of minimum sta¢ ng requirements on nursing home behavior and performance. What we

have found has important policy implications regarding the rising costs of long-term health

care and ine¤ective regulations of the nursing home industry. Especially at a time where

the e¢ cacy and e¢ ciency of a national health care system is a focal points of public policy

debate, more studies of this kind �studies that provide empirical evidence to maximize the

e¤ect of each health care dollar �are needed.

Due to its popularity in di¤erent economic settings,2 MQS regulation has attracted a

large amount of literature. Theoretical works tend to agree on the entry deterrent in�uence

MQS has on the regulated market. They disagree, however, about the impact of MQS on

the distribution of quality, which highlights the value of empirical work.3 Empirical studies

generally focus on the number of regulated products or services4 and previous investigations

1The older population - persons ages 65 or older - numbered 37.9 million in 2007, which represented 12.6%of the U.S. population, over one in every eight Americans. By 2030, there will be about 72.1 million olderpersons, almost twice their number in 2007. In 2007, 4.4 percent (1.57 million) of the 65+ population lived ininstitutional settings such as nursing homes. The percentage increases dramatically with age, ranging from1.3% for persons 65-74 years to 4.1% for persons 75-84 years and 15.1% for persons 85+ (Administration onAging, 2008).

2For example, drugs must satisfy federal safety standards and many professions have to pass the stateexaminations and ful�ll a series of requirements to be certi�ed or licensed.

3Leland (1979), Shapiro (1986), Ronnen (1991), Crampes and Hollander (1995), Valletti (2000), Jinji andToshimitsu (2004).

4Wiggins(1981) shows that drug regulation reduces the rate of introduction and R&D. Carroll and Gaston(1981) �nd that licensing restriction reduces the provision of seven professional services, including dentists.Gormley (1991) shows that quality regulations lower the number of child care centers. Lowenberg and Tinnin(1992) �nd similar results in the child care market.

1

on quality have been limited.5

One di¢ culty in empirically investigating MQS is the data constraint. Data on qual-

ity information is hard to observe, which is particularly true in markets with asymmetric

information. Previous work typically relies on inputs or outputs alone as indicators of qual-

ity. This type of quality measure is problematic without risk adjustment for heterogeneity

across examined observations.6 However, data required for risk adjustment is usually hard

to obtain. In addition to the issue of quality measure, the lack of variation in regulatory

policies causes identi�cation problems. To address this, some previous work has analyzed

policy variations across states all over the country. One disadvantage of those works is that

they tend to ignore the potential endogeneity problem caused by unobserved heterogeneity

across states. This unobserved heterogeneity has raised an additional di¢ culty for empirical

work. Quite a few works7 have employed panel data to correct for unobserved heterogeneity

using �xed e¤ects. However, their estimation assumes that policy changes are exogenous

once those time-invariant individual �xed e¤ects have been taken into account. Further

investigation is needed either to defend or to relax this assumption.

Given the issues of data constraint and its resulting methodology constraint, the nursing

home industry seems to provide an ideal setting to examine the causal e¤ects of minimum

quality standards on market outcomes. One unique feature of our study is that we have a

panel of observations covering almost all the nursing homes in the U.S. and over a long time

period (1996 to 2005). The quality measure used in this study is based on professional survey

teams�assessment of both the process and outcome of nursing home care, which provides

reliable and valuable quality information.8 Moreover, regulatory variation observed across

5Papers citing quality improvement include Holen (1978), Chipty and Witte (1995), Hotz and Xiao (2005)and Chen (2008). Papers showing deteriorating quality are Carroll and Gaston (1981), Chipty and Witte(1999), and Kleiner and Kudrle (2000).

6Using the number of physician visits is an example of output related quality measure. It may beinappropriate without controlling for the sickness of patients.

7Currie and Hotz (2004), Hotz and Xiao (2005), Siebert and Graevenitz (2005) and Chen(2008).8To be more speci�c, a survey team follows the federal standards to evaluate each surveyed nursing home.

If the nursing home fails to meet any certain standard, one corresponding de�ciency citation will be issued.Our quality measure is based on the number of de�ciency citations and the severity level of each violation.A higher number of de�ciency citations and a higher level of violations indicate lower quality of care.

2

states and time in the nursing home industry helps to precisely identify policy impacts.

Any analysis of policy impacts raises the question of endogeneity. Taking advantage of

our unique dataset, this paper has employed various speci�cations to address the potential

endogeneity problem caused by unobserved heterogeneity. It �rst adopts a model with �xed

e¤ects speci�cation (speci�cation (1)), which takes into account the permanent di¤erences

across states that are unobserved but are likely to be correlated with policy variables. One

disadvantage of this speci�cation is that it assumes away any time-varying individual at-

tributes or unobservables, the resulting being that the endogeneity problem may persist. As

a remedy, the basic speci�cation is extended in two ways. First, individual heterogeneity that

in�uences policy variables may follow individual speci�c trends. For example, a state�s in-

creasing sensitivity to quality issues may have caused more stringent inspections during each

survey (hence systematically lower measures of quality) and more strict minimum sta¢ ng

requirements. Ignoring this heterogeneity would confound the estimates for policy impacts.

The random trends speci�cation (speci�cation (2)) attempts to mitigate this potential bias

by adding market speci�c trends. Second, unobserved heterogeneity may exhibit more com-

plex dynamic behavior. To address this issue, the dynamic speci�cation (speci�cation (3))

introduces the lagged dependent variable and allows for the possibility that policy changes

may be related to those lagged variables. As a �nal extension, speci�cation (2) and (3) are

estimated adding a policy lead dummy indicating whether there will be any policy changes

in the subsequent year. This is to check the possibility of any reverse causality from the

left-hand side variables to policy changes. In the end, speci�cation (3), which includes the

lagged dependent variable, turns out to be our preferred model. Speci�cation (2) has de-

livered quite similar results, except for the case where quality is measured as the weighted

value of de�ciency citations.

MQS regulation imposed in the nursing home industry is characterized by minimum nurs-

ing hours per patient day for licensed nurses and direct care nurses. As di¤erent categories of

nurses a¤ect the care of residents in di¤erent ways (Grabowski (2001)), I examine the impact

3

of MQS separately for licensed nurses and direct care nurses. Estimations from the dynamic

speci�cation have found minimum sta¢ ng to have no signi�cant e¤ect on the number of

nursing homes, contrary to, for example, the child care industry. This is probable, given

that entry into the industry has already been heavily regulated by the government. With

regard to policy impacts on quality of care, minimum sta¢ ng requirement for licensed nurses

is shown to improve quality. To be more speci�c, an half hour increase of minimum sta¢ ng

requirements for licensed nurses increases quality by 15 percent if quality is measured by the

count of citations, and by 20 percent if quality is measured by the value of citations which

takes into account di¤erential severity in violation. Minimum sta¢ ng requirements for direct

care nurses are seen here to have no signi�cant impact on quality.

In the case of direct care nurses, this lack of impact on quality of patient care is striking.

One possible explanation is the lack of strict training and certi�cation in the profession. The

certi�cation for licensed nurses requires 2-3 years�education, whereas the certi�cation for

direct care nurses is minimal and informal to a degree that quality is not guaranteed. This

substandard training and certi�cation makes it possible for nursing homes to maintain their

operating costs by substituting less-skilled and cheap labor for direct care nurses after the

imposition of the minimum sta¢ ng requirements. As a result, although the quantity of direct

care nursing input is increased, its quality deteriorates and low-quality sta¢ ng undermines

quality of patient care. Mandating the quantity of nursing input does not necessarily improve

quality of care.

Another possible explanation is based on how nursing homes strategically adjust their

nursing inputs as response to minimum sta¢ ng requirements in a market with asymmetric

information. I�ve found more compressed quality distribution after the imposition of mini-

mum requirements for direct care nurses. As di¤erentiating becomes more costly after policy

regulation increases the lower bound of nursing inputs, nursing homes choose only to meet

the minimum sta¢ ng requirements imposed on direct care nurses. In the end, the average

quality of care does not increase.

4

The remaining paper is organized as follows. The next section reviews previous litera-

ture on minimum quality standards. Then, the nursing home industry is brie�y described.

Section 4 explains the data and provides summary statistics. The model and econometric

speci�cation are presented in Section 5 and the empirical results are discussed in Section

6. Section 7 discusses quality of care in the U.S. nursing homes and possible extensions or

avenue for future work. The last section sets out the paper�s conclusions.

2 Previous Literature

2.1 Theoretical Work

Minimum quality standards are considered a possible solution to the quality deterioration

problem in markets with asymmetric information; there is a large theoretical literature ex-

amining their impacts. Arrow�s work on minimum quality constraints in 1971 (focused on

occupational licensing as applied to professions) highlights MQS�s role in minimizing con-

sumer uncertainty. This increased consumer con�dence in the quality of the licensed service

in fact increases the overall demand for those services. Leland (1979) identi�es types of mar-

kets that are likely to bene�t from minimum quality standards. These markets usually share

the following properties: great sensitivity to quality variations; low elasticity of demand;

low marginal cost of providing quality and low value placed on low-quality service (Leland

(1979)).

Shapiro (1983), by contrast, demonstrates that some types of customers are worse o¤,

either because their preferred quality services are no longer supplied or because prices rise

after the imposition of minimum quality standards. Shapiro (1986) extends his previous

work, noting that minimum quality regulation raises the average quality of service in the

regulated market. Shapiro also concludes that the cost of raising service quality may be so

great as to decrease aggregate consumer surplus, even though licensing bene�ts the segment

of consumers that highly values quality. Di¤ering from the previous work which assumes a

5

competitive environment, Ronnen (1991) models �rms in an oligopolistic market structure:

�rms face quality-dependent �xed costs and compete in quality and price. Imposing a

minimum quality standard leads both low quality and high quality �rms to raise quality.

The intuition is as follows. The disparity between qualities shrinks since low quality �rms

raise quality to meet minimum quality standards. As a result, high quality �rms further raise

quality to di¤erentiate themselves from low quality �rms and to alleviate price competition.

Ronnen shows that all the consumers are better o¤with minimum quality standards enforced,

because of better qualities and lower hedonic prices, which is di¤erent from previous results

presented by Leland (1979) and Shapiro (1983 and 1986).

Crampes and Hollander (1995) consider a similar setting but with quality-dependent vari-

able costs. They di¤erentiate between mildly restrictive and excessively stringent minimum

quality standards.9 Their �ndings with regards to mildly restrictive quality standards are

quite close to Leland (1979) in that social welfare increases when quality standards reduce

the quality gap between �rms. However, the imposition of more stringent minimum quality

standards can eliminate high quality �rms, with the result being a decrease in average market

quality. Valletti (2000) shows that the e¤ect of mild minimum quality standards delicately

depends on the form of competition between �rms. Di¤ering from previous work that as-

sumes �rms compete in a Bertrand Game, his paper considers �rms as Cournot competitors.

He concludes that both low and high quality producers are worse o¤ under minimum quality

standards and social welfare decreases. Jinji and Toshimitsu (2004) revisit minimum quality

standards under a vertically di¤erentiated duopoly. The authors generalize models studied

in Ronnen (1991) and Valletti (2000) and �nd that the results presented in those two works

are quite robust.

9Mildly restrictive standards are slightly above the quality that a low quality �rm would have chosen inthe absence of regulation.

6

2.2 Empirical Work

Previous empirical works agree that the imposition of minimum quality standards has a

negative impact on the number of regulated products or the number of suppliers in the

regulated market. Wiggins (1981) �rst estimates the e¤ects of regulation on new drug

introduction in the 1970s. Drug regulations have a major direct impact on introduction as

well as a signi�cant indirect e¤ect through a reduction in research spending. More precisely,

regulations have reduced drug introduction by roughly 60 percent. Carroll and Gaston

(1981) study seven professional occupations, including electricians, dentists and plumbers.

They claim that for all occupations, government restrictions reduce the number of suppliers

per capita. Gormley (1991) uses state level aggregate data from the child care market

and shows that minimum quality standards, such as higher sta¤-child ratios and high square

footage requirements, reduce the number of child care centers. Lowenberg and Tinnin (1992)

also investigate the U.S. child care market; they conclude that stricter licensing rules are

associated with lower levels of consumption of child care services. In their paper, service

licensing raises entry cost into the industry and raises the supply price more than it increases

consumer utility. In this sense, quality regulations bene�t producers more than consumers.

Previous empirical work on quality is limited and mixed. Holen (1978) studies restrictive

dentist licensing regulations and shows that licensing increases quality of care by reducing

the likelihood of adverse outcomes. However, Carroll and Gaston (1981) show that restrictive

licensing may lower quality. In their study, excess demand (as a result of decreased supply)

increases the market price for regulated service providers, forcing some customers to turn

to unlicensed service providers. For example, they �nd that accident rates, measured by

the number of unintended electrocutions, are higher in states with more stringent licensing

requirements on electricians. Chipty and Witte (1995) use household level survey data to

examine the e¤ect of minimum quality standards in the child care market. They �nd that

regulations are binding and that they have economically large and statistically signi�cant

e¤ects on the equilibrium price of child care service and quality (as measured as sta¤-child

7

ratios). Regulations of di¤erent dimensions have various impacts on quality. For example,

training requirements and group size regulations increase equilibrium quality but minimum

sta¤-child ratio requirements signi�cantly reduce quality. Chipty and Witte (1999) study

child care providers�response to minimum quality standards. They �nd support that min-

imum quality standards improve the average quality of child care in certain markets. But

when regulatory intervention increases the probability of child care center closures, both the

average and the maximum quality observed in the child care market decline.

Kleiner and Kudrle (2000) use unique data on the dental health of incoming Air Force

personnel to empirically analyze the e¤ects of varied licensing stringency across states. Unlike

most of the previous work which uses input as the measure of quality, their paper de�nes

quality in terms of output, namely, the frequency of visits. Their rationale is that an inferior

dentist may require multiple attempts to �ll a tooth while a good dentist requires only one.

They �nd that tougher licensing improves quality and raises prices. Hotz and Xiao (2005)

distinguish their study in two ways. They are the �rst to use a unique panel dataset of the

child care market so that they can control for state and time �xed e¤ect. Second, they use

child care accreditation data as the measure of quality. They �nd that higher sta¤-child ratio

requirements act as a barrier to entry and reduce the number of operating child care centers.

Moreover, they show that the regulation of a higher sta¤-child ratio improves the average

quality of the market due to the exit of low quality providers. The surviving child care centers

also bene�t from this regulation by earning higher revenue and pro�t per employee. On the

other hand, higher sta¤-education requirements have quite the opposite e¤ects: they do not

deter entry and they lead to lower quality and lower pro�t. Based on these �ndings, the

authors conclude that minimum quality standards governing di¤erent dimensions of quality

may have con�icting e¤ects.10

There are quite a few papers examining nurse sta¢ ng levels and quality of care in the

10Previous work in child care markets such as the work done by Chipty and Witte (1995) draw quite similarconclusions in that di¤erent standards may have the opposite e¤ects on quality provided in the regulatedmarket.

8

nursing home industry.11 Among them, Chen (2008) is interested in how nursing homes

strategically choose nursing inputs in response to minimum sta¢ ng requirements. Chen

mainly uses data from two states, and she is focused on the policy regulations of total nursing

hours (the summation of licensed and direct care nurses�s hours). She �nds that minimum

sta¢ ng standards increase total nursing hours per patient day. She also �nds evidence that

nursing homes have higher incentive to di¤erentiate in markets where minimum sta¢ ng

requirements have a bigger impact, which is consistent with the results presented by Chipty

and Witte (1999).

3 The Nursing Home Industry

A nursing home is a place of residence for people with signi�cant di¢ culty in daily living so

that constant nursing care is required. Residents include the elderly and young adults with

physical disabilities.12 In 2007, more than 1.4 million people, mostly seniors, live in nearly

16,500 nursing homes nationwide (American Health Care Association, 2007). The United

Sates spent $131.3 billion in 2007 as opposed to $90 billion in 1999 on nursing home care.

Nursing home care is primarily paid for by three sources: Medicare, Medicaid and private-

pay. Medicare is the government health insurance plan for all eligible individuals age 65

and older.13 Averaging over 2001 to 2007, Medicare pays for 12 percent of all nursing

home patients.14 Medicaid is a welfare program jointly funded by the federal and state

governments but is largely administered by the state.15 Medicaid paid for 65 percent of

11Such as Harrington and et al. (2000), Schnelle and et al. (2004), Mueller and et al. (2006), Zhang andet al. (2006).12The majority of nursing home residents were over age 65 and about 10 percent were under age 65

(Decker, 2005). To be more speci�c, nursing home residents include the elderly with chronic disabilities;infants with multiple impairments; young adults with traumatic brain injury, or other physical disabilities;and individuals with short-term rehabilitation or sub-acute treatment needs.13To qualify for Medicare nursing home coverage, an individual must spend at least 3 full days in a

hospital before entering a nursing home. Medicare only covers nursing care up to 100 days. The �rst 20days of nursing care will be fully covered by Medicare and a co-payment will be charged for the remaining80 days. The average paid Medicare nursing home stay was 23 days in 1997, only 1/5 of the allowable time.14Those �gures and �gures below are based on various reports from the American Health Care Association.15To qualify for Medicaid, the potential recipients must pass a means test - their income and assets must

9

residents. Private and other sources paid for the remaining 23 percent of nursing home

residents. As government sources pay for the majority of nursing home residents, it is plain

to see how intimately involved the government is in the industry.

There has been widespread concern about nursing home residents receiving poor quality

care. As a response, the Institute of Medicine published its landmark report in 1986 that

called for major revisions in the way nursing home quality was monitored. Following their

recommendation, Congress passed the Nursing Home Reform Amendment to the Omnibus

Budget Reconciliation Act (OBRA) in 1987. This amendment mandated new standards

of care, including increased minimum sta¢ ng regulations and quality of care monitoring

(Harrington & Carrillo (1999)).

Besides the federal regulations of minimum nurse sta¢ ng, most states have imposed

additional requirements. The highest overall sta¢ ng requirement was adopted in Califor-

nia, which requires 3.2 hours per resident day, excluding administrative nurses (Harrington

(2001)). Despite the regulatory e¤orts stemming from the implementing of OBRA 1987 and

state-imposed sta¢ ng requirements,16 quality problems seem to persist in the industry. A

preliminary analysis of the data shows that the average number of de�ciency citations per

nursing facility decreased from 7.2 in 1994 to 4.9 in 1997, followed by a gradual increase to

7.5 in 2007. The percentage of facilities with de�ciencies that caused harm or immediate

jeopardy to residents rose from 25.7 percent in 1996 to 30.6 percent in 1999, before declining

dramatically to 15.5 percent in 2004; the percentage of facilities with such de�ciencies rose

slightly to 17.6 percent in 2007(Harrington and et al., (2008)).

To what extent has quality of care been a¤ected by regulatory policies? While the above

statistics have provided valuable information, they must be interpreted cautiously. The

confounding components inherent in the data need to be identi�ed and isolated if we are to

accurately evaluate the impact of minimum sta¢ ng requirements on the quality of patient

be lower than a certain level as determined by the individual state.16Due to extended negotiations with the nursing home industry, OBRA 1987 did not take e¤ect till 1995,

8 years after the passage of the law (Wiener, 2007).

10

care.

4 Data and Descriptive Statistics

The data used in this study comes from three sources: (1) state regulatory policies, (2) the

1996 through 2005 Online Survey, Certi�cation, and Reporting (OSCAR) �les, (3) the 2004

Area Resource File (ARF) and the most recent U.S. Population Census. Consistent with

previous work, the county is de�ned as a proxy for the nursing home market.17 The county

may be a reasonable approximation of the market for nursing home care given patterns of

funding and resident origin (Gertler (1989)).18 This section explains each component of our

data and provides descriptive statistics.

4.1 Nursing Home Regulations on Minimum Nurse Sta¢ ng

Data on statutes and regulations is mainly collected from previous literature, which provides

historic regulations back to 1997. More recent regulations are obtained via the internet.

A Medicare and/or Medicaid certi�ed nursing home has to meet the minimum sta¢ ng

levels set by the federal and state government. The federal Nursing Home Reform Act

(NHRA), as part of the Omnibus Budget Reconciliation Act (OBRA) of 1987, sets minimum

sta¢ ng levels for registered nurses (RNs) and licensed practical nurses (LPNs), and minimum

educational training for nursing assistants (NAs). The NHRA requires Medicare and/or

Medicaid certi�ed nursing homes to have: "a RN director of nursing; a RN on duty at least

8 hours a day, 7 days a week; a licensed nurse (RN or LPN) on duty the rest of the time; and

a minimum of 75 hours of training for nurse�s aides." The law also requires nursing homes

17Most studies have used the county as a proxy for the nursing home market (e.g., Cohen and Spector,1996; Nyman, 1985; Zinn, 1993).18Gertler (1989) shows that 75 percent of nursing home residents in New York State had previously lived

in the county where the home was located. Nyman (1989) �nds 80 percent of residents in Wisconsin facilitieschose a nursing home located in the same county of residence. A most recent study by Mehta (2006) �ndsa strong inclination for residents to stay in a nursing home closer to their home. Simulation results suggestthat the county is a good proxy for the market and that all �rms within that area can be assumed to competeequally (Mehta, 2006).

11

"to provide su¢ cient sta¤ and services to attain or maintain the highest possible level of

physical, mental, and psychological well being of each resident" (OBRA 1987). The total

licensed nursing requirements converted to hours per resident day (HPRD) in a facility with

100 residents are around 0.30 HPRD (Harrington, 2001), or 30 hours per day.

Most states have imposed additional requirements for minimum nursing standards. These

standards are quite complex and vary considerably across states. In order to compare these

standards, several steps must be taken. First, standards may apply to only one class of

nursing personnel or to groups of personnel. Given that di¤erent categories of nurses may

a¤ect quality of care di¤erently, I divide those standards into two categories: licensed nurses

(LNs) and direct care nurses (DNs). LN includes registered nurses (RNs), licensed practical

nurses (LPNs) and licensed vocational nurses (LVNs) while DN includes certi�ed nursing

assistants (CNAs), or nursing assistants (NAs) who provide direct nursing care. Second,

standards are set in di¤erent forms.19 For simplicity those standards are converted to the

hours per resident day for a 100 bed nursing facility.20

The federal government has set a minimum sta¢ ng requirement of 0.3 HPRD, regarded as

the lower bound of the regulation for licensed nurses. There is no speci�c federal requirement

with respect to direct care nurses. Up to 2005, 24 states, including the District of Columbia,

had established a minimum sta¢ ng ratio for licensed nurses that was higher than the federal

ratio. The remaining 27 states followed the federal licensed nurse sta¢ ng requirements. As

for minimum sta¢ ng requirements for direct care nurses, 34 states have established their own

standards to date. Regulations varied during our study period. Most of the changes were due

to the adoption of minimum sta¢ ng ratio for either licensed nurses or direct care nurses. Ten

of the states which did not have requirements for licensed nurses in 1996, when the dataset

begins, adopted standards by 2005, when the dataset ends. Similarly, nine states established

19Minimum nursing standards are expressed as either hours per resident day (HPRD), as a ratio of sta¤to residents, or as a ratio of sta¤ to beds. In some cases, two formulations are used. For example, Californiarequires 3.2 hours of direct care per resident day while Maine maintains a direct care sta¤-to-resident ratioof 1 to 5 during the day, 1 to 10 in the evening, and 1 to 15 at night.20More detailed discussion of the conversion can be found in Harrington (2001).

12

requirements for direct care nurses during the time period studied here. During the course

of this study, other states, including Arizona and Missouri, dropped their requirements on

direct care nurse sta¢ ng ratios. Summary statistics on regulatory policies can be found in

Table 1.

Table 1: Descriptive minimum sta¢ ng requirements

Variable Mean Std. Dev. Min Max

Licensed Nurses (Policy Dummy) 0.40 0.49 0 1.00

Direct Care Nurses (Policy Dummy) 0.62 0.49 0 1.00

Licensed Nurses (HPRD) 0.40 0.15 0.30 1.20

Direct Care Nurses (HPRD) 1.15 1.00 0 2.96

Each variable is measured at the state-year level. Observation=400 for 50 states from 1997 to 2004.

States generally rely on the licensing process to monitor and enforce sta¢ ng ratios; meet-

ing minimum ratios is part of state nursing home licensure and regulatory requirements. The

Center of Medicare & Medicaid Services also contracts with each state to conduct annual

onsite inspections that determine whether its nursing homes meet the minimum Medicare

and Medicaid quality and performance standards.21 More details about the inspection and

survey can be found in the following section.

4.2 Nursing Home OSCAR Files

This study uses the On-Line Survey Certi�cation and Reporting System (OSCAR) data from

the �scal year 1996 to 2005. The OSCAR data is based on an annual survey conducted by

state licensure and certi�cation agencies as part of the Medicare and/or Medicaid certi�-

cation process.22 State inspectors collect the OSCAR data every 9 to 15 months to verify

nursing homes�s compliance with all federal and state regulatory requirements. During each

inspection visit, the survey team observes nursing care and sta¤/resident interaction. The

21The state regulations cover many aspects of resident life, from specifying standards for the safe storageand preparation of food to protecting residents from physical or mental abuse or inadequate care practices.There are over 150 regulatory standards that nursing homes must meet at all times. Many are related.22Among the surveyors, there are trained health care professionals in nursing, nutrition, social work,

pharmacy and sanitation.

13

surveyors then �ll in a standard form to determine whether various regulatory standards are

being met for the visited nursing home.

The OSCAR data includes approximately 96 percent of all nursing facilities in the United

States. The dataset is considered the greatest source of reliable information about the U.S.

nursing homes. However, there are limitations to the OSCAR data.23 One concern is that

OSCAR uses a snapshot method of surveyor observation, which may lead to inter-surveyor

variations and inconsistencies. The problem is mitigated by the fact that all the surveyors

have to strictly follow the federal standards for survey visits and �lling in survey forms.

Moreover, we rely on our model speci�cations to address the remaining survey variation

issues, if there are any.

The number of nursing home providers is identi�ed through a nursing home�s presence

and absence from the OSCAR data. In case of a mismatch of the identi�cation number of a

nursing home, I use detailed location information to match observations across years. Since

each survey is done at an irregular interval of 9 to 15 months, our data identi�es the above

variable for the time period of 1997 to 2004.24

The quality measure is based on the annual de�ciency citations at facility levels and is

calculated as the market average over each nursing homes within each market. De�ciency

citations are issued to facilities by state surveyors as a part of the federal survey process.25

There are 185 tags in total to cite, and each tag corresponds to one criterion related to the

quality of nursing home care. If the surveyed nursing home fails to meet or violates one

certain criterion, one corresponding de�ciency citation will be issued. More violations incur

more citations and therefore indicate lower quality of care.

23One concern is that each nursing home provides information on resident characteristics, and only someof the residents are selected to be veri�ed by the surveyors. This may cause the problem of measurementerrors. Fortunately, these parts of the data are not used in this analysis.24For example, if a nursing home is not observed in the survey of 2005, I cannot identify whether it has

exited the market in 2005 or it has not exited but its survey was going to be conducted sometime in 2006(but I cannot observe the survey as our data ends at 2005).25The process and the outcomes of nursing home care in 15 major areas are assessed by state surveyors.

Each of these areas has speci�c regulations which state surveyors review to determine whether or not facilitieshave met the standard. In July 1995, the Health Care Financing Administration consolidated the total of325 tags (individual criteria) to a total of 185.

14

Some violations do not relate to nurse sta¢ ng levels, so do the corresponding citations

issued as the result of those violations. 26 Since our main interest is to examine policy

impacts of minimum sta¢ ng requirements, it�s optimal to isolate those non-sta¢ ng related

citations for the calculation of our quality measure. To do that, I �rst di¤erentiate between

sta¢ ng and non-sta¢ ng related citations based on the detailed tag information of each

citation.27 I then focus on those sta¢ ng related citations in this study.

Besides the use of de�ciency citations, other quality measures in the literature include

resource use and patient outcome. Both measures need to be adjusted using detailed infor-

mation on the severity of patient illnesses at each nursing home. However, this information

is hard and expensive to obtain and any unobserved information regarding severity of illness

will lead to biased quality measures. Due to these reasons, de�ciency citations have become

the most common quality measures (Mukamel and Spector (2003)).

To provide a quantitative measure of quality, I �rst simply use the count of the total

citations issued to each nursing home. Given that such measure ignores di¤erences in the

severity of each violation, I therefore also calculate a value measure of de�ciency citations,

which takes into account the scope and severity level of each citation.28 This value measure

follows a weighting method used by Gannett News Service where a score is assigned to each

de�ciency based upon the citations�scope and severity.29

In the end, sta¢ ng related de�ciency citations provide two quality measures (Q_c and

Q_v) to compare and evaluate quality of patient care in nursing homes nationwide. The

variable Q_c is the count measure of sta¢ ng related de�ciency citations, and the variable

Q_v is the value measure of those citations using a weighting method that takes into ac-

26For example, an environment/cleaning violation will incur a citation, but it does not necessarily relateto nurse sta¢ ng levels.27Among all the tags, the following are considered as non-sta¢ ng related: F151-177 (resident rights), F201-

208 (admission, transfer and discharge rights), F360-372 (dietary services), F385-390 (physician services),and F454-469 (physical environment).28E¤ective in July 1995, each de�ciency is also rated on its scope and severity. An alphabetic score (from

A to L) is given to each de�ciency based on the combination of the de�ciency�s scope and severity indicator.29For example, a de�ciency with a scope and severity of D is scored as a 5, whereas a de�ciency with a

scope and severity of K receives a score of 45. More detailed information can be found at Matthews-Martin(2003).

15

count the di¤erential severity in violation. Similarly, the total de�ciency citations (including

sta¢ ng and non-sta¢ ng related citations) provide two measures for overall quality (T_c

and T_v). These two variables are used as instruments for the estimation of the dynamic

speci�cation.

The unit of observation of this study is the county so that county-wide quality is measured

as the average across nursing homes within a county. Not shown in this paper, another

measure of the county-wide quality is weighted by the number of beds per facility and this

alternative quality measure provides quite similar results.

Summary statistics are presented in Table 2. Counties with missing values for demo-

graphics are deleted from the sample so that the data covers a total of 3,073 counties in the

U.S. during the time period of 1997 to 2004. When it comes to the study of quality, the data

covers 2,507 counties as observations which lack information on de�ciency citations or their

severity levels are also dropped.30 Note that I add negative sign to those log transformed

quality measures so that higher value means higher quality of care. Also note that the value

of zero for quality measures means no de�ciency citation, indicating the highest level of

quality.

Table 2: Descriptive establishment and quality measures from 1997 to 2004

Variable Mean Std. Dev. Min Max

Ln(N): natural logarithm of the number of nursing homes 1.55 0.79 0 6.14

Ln(Q_C): natural logarithm of the related count citations -1.57 0.89 -4.08 0

Ln(T_C): natural logarithm of the total count of citations -1.78 0.91 -4.66 0

Ln(Q_V): natural logarithm of the related value citations -3.05 1.21 -7.09 0

Ln(T_V): natural logarithm of the total value of citations -3.30 1.21 -7.48 0

Each variable is measured at the county level. Observation=24,584 for ln(N), =20,056 for quality measures.

4.3 ARF (2004) and Other Data

The 2004 Area Resource File (ARF) is used as a data source for market factors. ARF

collects several thousand variables on population characteristics, socioeconomic features, and30Among those 3,073 counties, roughly 6 percent do not have data of quality because they do not have

any nursing homes.

16

health care resources from more than 50 di¤erent sources such as the National Center for

Health Statistics, the American Medical Association, and the American Hospital Association.

Variables used in this analysis include average income, the size of the senior population,

Medicare reimbursement rate and Medicaid nursing home reimbursement rate. State level

Certi�cate of Need policies are also used in the paper.31 Table 3 summarizes the variables

discussed above.

Table 3: Descriptive other control variables, market level

Variable Mean Std. Dev. Min Max

Ln(Income): natural logarithm of income 10.02 0.23 8.50 11.40

Ln(Elder): natural logarithm of elder population 8.27 1.33 2.48 13.80

Medicare Rate 274.74 44.38 108.30 634.93

Medicaid Rate 98.11 24.23 57.08 253.48

CON: certificate of need program 0.71 0.45 0 1

Each variable is measured at the county level. Observation=24,584.

5 Empirical Speci�cation

The main goal of this study is to identify the e¤ects of minimum quality standards on

outcomes of the nursing home market. Minimum sta¢ ng requirements serve as proxies for

regulatory policies on minimum quality standards. Nurses are divided into two categories: li-

censed nurses and direct care nurses. State regulators may set minimum sta¢ ng requirements

for either licensed nurses, or direct care nurses, or both. Minimum sta¢ ng requirements are

measured both as binary policy dummies and as continuous measures of minimum nursing

hours per patient day. The following work focuses on continuous measures; however, policy

dummies provide very similar results.

Considering that observations within each state are likely to be dependent, all of the re-

gressions are adjusted for clustering at state-year level. Failure to account for clustering may

cause the researcher to greatly understate the standard errors on the estimated coe¢ cients

31Certi�cate of Need and Moratorium policy allows the government to be involved in the process ofestablishing a new nursing home and change of bed capacity of an existing home. The policy claims toration resources so that there will not be an uncontrolled growth of facilities.

17

for the state-level variables (Moulton (1990)).

Several speci�cations are studied in this paper and they are detailed in the following

discussion. The dynamic speci�cation, which comes last, is our main and preferred model.

Fixed E¤ect Speci�cation

The basic speci�cation uses a di¤erence-in-di¤erence methodology to estimate policy

impacts. To be more speci�c, the outcome equation is written as:

Yist = �0 + �i + �t +MQSst � �1 +Xist � �2 + "ist (1)

where Yist represents various dependent variables at state s market i in time t, such as the

number of nursing homes and quality measures. The variables MQSst control for minimum

sta¢ ng policies, which vary at state-year level. The coe¢ cients �1 are our primary research

interests. The variable Xist is the vector of variables representing market characteristics such

as average income, size of the elder population, and Medicaid and Medicare reimbursement

rates. Variables �i and �t are market and time �xed e¤ects respectively.32 Note that time-

invariant state �xed e¤ects are automatically taken care of with the inclusion of market �xed

e¤ects.

The inclusion of year dummies provides controls for unobserved national attributes that

may a¤ect the dependent and the policy variables. The inclusion of market �xed e¤ects has

two advantages. It provides controls for market (state) heterogeneity that may a¤ect the

dependent variable, such as the quality of nursing home care. More importantly, it provides

controls for unobserved time-invariant factors that may also relate to the policy changes

across states.

Random Trends Speci�cation

The �xed e¤ect speci�cation takes control of unobserved heterogeneity but assumes they

32States vary substantially in the stringency of nursing home regulations. Furthermore, some states havechanged their regulations frequently enough that it is possible to use variation over time within states tocontrol for state �xed e¤ects and to use variation across states within time to control for time-�xed e¤ects.I exploit this across-state and over-time variation in state regulations to examine the impact of minimumquality standards on behaviors of the nursing home market.

18

are unchanged over time. However, unobservables that are correlated to policy variables can

vary within a state over time. For example, there may exist state speci�c trends that have

caused more stringent inspections of de�ciency citations (hence lower the measure of quality)

and more stringent minimum sta¢ ng requirements. Another simultaneity example is how

increasing concern for quality of care has driven both policy changes and better quality of

care. Or it might be the case where not more stringent policies, but increasing concern for

quality of care has caused better quality of nursing care. Ignoring those unobservables would

confound the estimates for policy impacts. To mitigate the bias, the following speci�cation

adds market speci�c trends into the �xed e¤ect speci�cation where

Yist = �0 + �i + �t + �it+MQSst � �1 +Xist � �2 + "ist (2)

Speci�cation (2) is also referred to as a random trend model.33 This speci�cation captures

the impact of policy changes on deviations of the left-hand side variables from their market

growth paths. To estimate equation (2), the �rst-di¤erence is taken to get rid of �i so that

the equation is transformed to

�Yist = �t + �i +�MQSst � �1 +�Xist � �2 +�"ist (3)

and then equation (3) is estimated using the �xed e¤ects method to get rid of �i.

Dynamic Speci�cation

The random trend speci�cation provides a more �exible way to control for heterogeneity

in unobservables that may bias the estimation of �1, but it restricts the market speci�c trends

to follow a linear pattern for the purpose of identi�cation. To account for the possibility

that some unobserved factors may exhibit more complex dynamic behavior, speci�cation

(3) includes the lagged value of the dependent variable. Taking the analysis of quality as

33Some works, such as Friedberg (1998), have shown the importance of including those individual speci�ctrends.

19

an example, the lagged quality of care captures state dependence in quality of care and

provides a good proxy for factors determining policy changes. Given a statewide problem

of deteriorating quality, the state government may be more likely to impose more stringent

minimum sta¢ ng requirements as a remedy. Speci�cation (3) is given as follows:

Yist = �0 + �i + �t + �Yist�1 +MQSst � �1 +Xist � �2 + "ist (4)

Again the above equation is transformed to the equation below by taking the �rst-order

di¤erence

�Yist = �t + ��Yist�1 +�MQSst � �1 +�Xist � �2 +�"ist (5)

and then the above equation is estimated using instruments for �Yist�1.

Endogeneity Issues

There may exist other types of endogeneity that have not been addressed under speci�-

cation (2) and (3). For example, there might be an artifact of a spurious correlation between

the quality of nursing care and the propensity for a state to adopt or change its regulatory

policies regarding minimum sta¢ ng requirements.

To further check for the existence of endogeneity problems in MQS policies, I include

in speci�cation (2) and (3) an additional dummy variable for whether there will be any

policy changes in the subsequent year.34 Since two policy variables are examined in this

study, I allow the dummy variable to be one whenever one policy variable has changed in

the subsequent year. The estimated coe¢ cient on the lead dummy should be insigni�cant.

Otherwise, there should be concerns for reverse causality from the left-hand side variable to

policy changes. The similar strategy has been employed by Gruber and Hanratty (1995).

34An earlier version of this paper has adopted two variables re�ecting trends in hospital nurse sta¢ nglegislation as instruments for �MQSst: intro (whether sta¢ ng legislation for hospitals has been introducedin a state at a particular year) and enact (whether sta¢ ng legislation for hospitals has been enacted ina state at a particular year). However, the inclusion of these two instruments has caused quite impreciseestimate for the parameters of interests under speci�cation (2) and (3). This may be due to the lack ofvariation across state and time in those two instrumental variables.

20

6 Results

This section presents and discusses the estimation results. All of the regressions are clustered

at state-year level. In the interest of length, coe¢ cients for time dummies are not presented.

Table 4 examines the impacts of minimum sta¢ ng requirements on the number of nursing

homes at the county level. Their impacts on quality of care is presented in Tables 6 and

7, where Table 6 uses the number of de�ciency citations as the measure of quality and

Table 7 uses the value measure of quality. In addition, Table 5 provides results for the

extended models that include the policy lead dummy to test for the reverse causality of

policy regulations.

6.1 E¤ects on the Number of Nursing Homes

Column 1 of Table 4 presents the ordinary least squares (OLS) estimation without controls

for the market �xed e¤ects. The coe¢ cient for licensed nurses measures how a change

of minimum nursing hours of licensed nurses a¤ects the number of nursing homes at the

market level. The results from the OLS indicate that a half-hour increase of licensed nursing

requirements increases the number of homes by more than 17 percent. Direct care nursing

requirements, on the other hand, have a negative but largely insigni�cant impact. Medicare

and Medicaid reimbursement rates are found to decrease the number of nursing homes.

Column 2 shows the estimation results for speci�cation (1) using �xed e¤ects estimation.

By contrast to the OLS results, there are no longer any signi�cant negative e¤ects of Medicare

and Medicaid reimbursement rates. Instead, we see a signi�cant negative e¤ect of minimum

nursing hours of direct care nurses. Moreover, the positive e¤ect of licensed nurses is largely

reduced in magnitude. This positive e¤ect could have di¤erent interpretations. It might

re�ect the demand-expanding e¤ect that imposing minimum sta¢ ng requirements reduces

uncertainty over quality of care and increases overall demand for nursing home care (Arrow,

1971). It might as well just indicate an artifact of a spurious correlation between the number

21

of nursing homes and regulatory policies regarding minimum nurse sta¢ ng.

Estimation of speci�cation (2) is presented in Column 3. Under this speci�cation, there

are no longer any signi�cant impacts regarding minimum sta¢ ng requirements for licensed

nurses and direct care nurses. Note that under this speci�cation, the identi�cation of policy

impact relies on the deviation of the number of nursing homes from its market growth trend

rather on the deviation from its market average level across time. Also note that this paper

uses data covering 3,073 counties from 1997 to 2004, which gives us a total of 24,584 obser-

vations. Taking the �rst-order di¤erence leaves 21,511 observations under speci�cation (2).

The remaining columns are estimation results for speci�cation (3). Column 4 uses dl2:Yist�2,

the di¤erence of the lagged two-period Yist�2 and the lagged three-period Yist�3 number

of nursing homes, as an instrument for �Yist�1. Column 5 uses Yist�2 and Yist�3 as instru-

ments. The dynamic speci�cation has provided quite similar results compared to the random

trend speci�cation. One striking di¤erence as compared to the FE speci�cation is that we

see no evidence of any positive impacts from licensed nurses, indicating the importance of

controlling for other sources of heterogeneity.

Table 5 reexamines speci�cation (2) and (3) with the inclusion of the policy lead dummy.

The �rst two columns are results for the case where the dependent variable is the number of

nursing home providers. The estimated coe¢ cients on the policy lead dummy are all very

small in magnitude and in statistical signi�cance, suggesting that the causality goes from

policy changes to the dependent variable.

In conclusion, policy variables have no signi�cant e¤ects on the number of nursing home

providers. Although we cannot make any inferences regarding the e¤ects of regulations on

the behavior of either the demand or the supply side based on our reduced form analysis, this

insigni�cant impacts may be due to the fact that entry into the nursing home industry has

been heavily regulated by state government. The imposition of minimum sta¢ ng require-

ments may increase overall demand for nursing home care. However, supply is regulated

so that it fails to meet the increasing demand, or supply is decreased as minimum sta¢ ng

22

requirements increase costs for the supply side. As a result, there is no signi�cant impact on

the number of nursing homes.

Our �ndings also indicate that the size of the elderly population is an important deter-

minant of the number of nursing homes. This result is quite robust across speci�cations.

Note here that the variable Certi�cate of Need program is dropped out of the analysis un-

der speci�cation (3) because of the lack of variation within states during the sample period

(2000-2004) under speci�cation (3). As a result, its impact gets picked up by the state �xed

e¤ects.

6.2 E¤ects on Quality

Quality of care has long been a hot debate in the nursing home industry. The imposition

of minimum sta¢ ng requirements is aimed at improving quality of patient care. Whether

this goal has been reached, and to what extent, will have profound policy implications.

Estimation results from various speci�cations are presented in Table 6 and 7 where quality is

measured as the count of de�ciency citations and the value of citations respectively. Column

1 of both tables correspond to the basic �xed e¤ect speci�cation. Both regulatory policies

are found to have no signi�cant impacts on quality.

The �xed e¤ects speci�cation (speci�cation (1)) takes into account the permanent di¤er-

ences across states that are likely to be correlated with policy variables. One disadvantage

of this speci�cation is that it assumes away time-varying individual attributes or unobserv-

ables. The ignorance of those time-varying attributes may have confounded the estimation

of policy impacts if they are correlated with the regulatory policy changes across states and

time. For example, states that have shown rising concern about quality of care may see both

the adoption of quality regulations and lower quality measures (due to more stringent survey

investigations). If this is the case, estimations of policy impacts will tend to be downward

biased. Another example that will cause downward biased estimations is the selection prob-

lem: states with decreasing quality of care may be more likely to adopt quality regulations to

23

improve quality. There are also cases where estimations would be upward biased. For exam-

ple, a state may experience increasing demand for high quality of care, which could improve

quality as well as increase a state�s minimum sta¢ ng requirements. Without capturing this

unobserved heterogeneity, it is unclear whether it is the minimum sta¢ ng requirement or

the increasing demand for quality that has driven up quality of care.

Speci�cations (2) and (3) add controls for time-varying unobserved heterogeneity that

has been assumed away by speci�cation (1). The results for speci�cation (2) are shown in

Column 2 of Table 6 and 7; and results of speci�cation (3) are listed in the remaining columns

of both tables. Column 3 of Table 6 uses the lagged two-period quality measure Yist�2 as

an instrument for �Yist�1; Column 4 uses Yist�2 and a lagged two-period overall quality

measure as instruments. Note that throughout the paper, quality of care, as the dependent

variable, is measured using sta¢ ng related de�ciency citations. The overall quality measure

(using both sta¢ ng and nonsta¢ ng related de�ciency citations) is correlated to the measure

of quality of care because they can be considered as decisions made within the same nursing

facility, or as the survey outcomes delivered by the same survey team.

Using both the count and the value measures of de�ciency citations (those related to

nurse sta¢ ng), I �nd signi�cant improvement in quality as the result of minimum sta¢ ng

requirements for licensed nurses. To be more speci�c, results from the dynamic speci�cation

show that an extra half hour�s sta¢ ng requirement for licensed nurses increases the quality

level by 15 percent if quality is count-measured and by 20 percent if quality is value-measured.

This quality-increasing e¤ect is consistent with previous research �ndings that more licensed

nurses improve quality. Note here that quality measures are rescaled so that a positive

coe¢ cient indicates a quality increasing e¤ect. As opposed to the impact of licensed nurses,

the estimated parameters for direct care nurses remain insigni�cant under speci�cation (2)

and (3). More detailed discussion about this insigni�cant impact will be provided in the

next section.

As a comparison between speci�cation (2) and (3), the estimated coe¢ cient for licensed

24

nurses is larger in magnitude for the random trend speci�cation. As shown in Table 6, the

size of the coe¢ cient is 0.44 as opposed to 0.31, and the di¤erence is even bigger for the

case (Table 7) where quality is measured taking into account di¤erential severity in levels

of violation. To further examine which speci�cation should be preferred, we add the policy

lead dummy to both speci�cations and the results are presented in Table 5 (Columns 3-6).

As shown in column 3 and 4 where quality is count measured, the coe¢ cients on the lead

dummy are both insigni�cant. However, the coe¢ cient turns to be signi�cantly negative

for the random trend speci�cation where quality is value measured (Column 5). This is

most likely caused by an artifact of the correlation between policy variables and the quality

measure. With the inclusion of the policy lead dummy, the measured coe¢ cient for licensed

nurses shrinks to be closer to the one estimated under the dynamic specialization, validating

the dynamic speci�cation as our preferred model.

Medicare Reimbursement rates are positively correlated to both measures of quality of

care. The Medicaid reimbursement rate is found to have a negative e¤ect on quality. This

�nding is opposite to some recent work such as Grabowski (2001), but consistent with pre-

vious work such as Nyman (1985) and Gertler (1989).

To sum up its impact on quality, the imposition of minimum sta¢ ng requirements im-

proves quality of care. A half-hour increase of minimum sta¢ ng requirement for licensed

nurses improve quality by 15 percent if quality is measured as de�ciency count, and by 20

percent if quality is measured as de�ciency value. The signi�cant quality-improving e¤ect of

licensed nurses could be due to the fact that licensed nurses play a supervisory role, and it

is very likely that increased sta¢ ng at licensed nurse levels is e¤ective in increasing quality

of care. This result is consistent with previous �ndings.35

35Such as Cohen and Spector (1996), Schnelle et al. (2004), and Zhang and Grabowski (2004).

25

7 Discussion and Future work

By imposing a minimum lower bound, minimum quality standards (MQS) are intended to

improve quality of the regulated product. Considering the important role that nurses play in

providing quality health care, quality is expected to improve after the imposition of minimum

sta¢ ng requirements. By examining the MQS separately for licensed nurses and direct care

nurses, we �nd that whereas minimum sta¢ ng requirements for licensed nurses increase the

quality of patient care, similar requirements for direct care nurses have no signi�cant impact.

Imposing minimum direct care nursing requirements does not necessarily improve quality.

One possible explanation is related to nursing homes�incentive to substitute cheaper laborers

to reduce operating costs. Labor expenses constitute the largest component of a nursing

home�s operating expenses. The imposition of minimum sta¢ ng requirements increases

labor costs and one unintended consequence is that nursing homes may compensate quality

for quantity to maintain their labor costs. The di¤erence in policy impacts from licensed

nurses and direct care nurses can be explained by the di¤erence between the two labor

markets: nursing homes are more likely to hire cheap and less skilled substitutes for direct

care nurses as compared to licensed nurses.

To become a certi�ed registered nurse, an individual has to obtain a degree in registered

nursing (which normally takes 2-3 years to complete) and pass a national licensing exam-

ination. In this sense, the quality of licensed nurses can be guaranteed. On the contrary,

for the profession of direct care nurses, the current certi�cation requirement is minimal: a

minimum of 75 hours of entry-level training, 12 hours of supervised clinical training and a

competency exam within 4 months of employment. Many nursing homes provide their own

free training program to their job candidates. This kind of minimal and informal training

makes it possible for nursing homes to hire cheap and low-skill labors as substitutes for direct

care nurses.36 Nursing homes may also have incentives to force overtime work, which will

36One characteristic of the direct care nursing labor market is a very high turnover rate caused by lowsalaries, few bene�ts and a heavy workload. A report by the American Health Care Association shows thatthe turnover rate for certi�cate nurse aids was over 71% in 2002 nationwide. "Average annual CNA turnover

26

deteriorate the quality of care as well. According to a national survey on compliance with

minimum wage, overtime and child labor violations in nursing homes, only 40 percent of

the sample was in compliance with these requirements in 2000, while the compliance rate

was 70 percent in 1997 (GAO, 2001). To summarize, mandating quantity of care does not

necessarily guarantee quality of care. The neglect of quality in direct care nurses undermines

quality of patient care. Future work could investigate how the labor market of direct care

nurses in�uences quality of care in the nursing home industry.

Another possible explanation for this outcome may be seen in how nursing homes strate-

gically choose nursing inputs after the imposition of minimum sta¢ ng requirements in an

industry su¤ering from asymmetric information problems. Imagine a model where patients

cannot perfectly observe quality information and a nursing home has to set its sta¢ ng ra-

tio a lot higher so as to distinguish itself from its competitors. Let�s assume the di¤erence

in sta¢ ng levels has to exceed one for di¤erentiation. But obtaining a high sta¢ ng ratio

is costly. Consider a simple case with two nursing homes in the market, where sta¢ ng is

zero for nursing home A and one for nursing home B. Now a minimum sta¢ ng regulation

is imposed at the level of 0.5. Nursing home A increases its sta¢ ng just to the minimum

required level at 0.5. Nursing home B can choose to be at level 1.5 so that patients will

acknowledge its high quality or it can choose to lower its sta¢ ng to the minimum required

level at 0.5. High labor costs and shortage of direct care nurses may deter B from hiring

more and as a result, the new market equilibrium sta¢ ng is 0.5, which equals the equilibrium

sta¢ ng before the imposition of the policy. This model can also be extended to the case

where quality of care ends up lower as a result of minimum sta¢ ng requirements.

Had the model explained the di¤erences in policy impacts for the two types of nurses,

we should expect more compressed variance in nursing inputs for direct care nurses after the

imposition of minimum sta¢ ng requirements, but not for licensed nurses. Unfortunately,

rates were below 40% in only 4 percent of states, and 60% or less in only 35 percent of states. CNA turnoverrates exceed 60% in 65 percent of states, exceed 80% in 37 percent of states, and were above 100% in 20percent of states." (Decker and et al., 2003)

27

data on nursing inputs is not available for this study. Instead I rely on quality data based on

de�ciency citations to provide some evidence. If it is the regulatory policies on the direct care

nurse, but not the licensed nurse that reduce the variance in quality measure, the explanation

based on the model seems plausible. To proceed, I focus on the value measure of quality

which takes into account di¤erential severity levels in violation and I undertake empirical

analysis both at the market level and at the state level. For market level analysis, the quality

measure at the nursing facility level is used to calculate the standard deviation of quality at

the county level. The standard deviation turns out to be signi�cantly smaller for markets

with regulatory policies for both licensed nurses and direct care nurses.37 I further run a

regression of the standard deviation of quality on the two policies (dummies), including

market level characteristics used in the main analysis. I�ve found that the imposition of

minimum sta¢ ng requirements on direct care nurses, but not licensed nurses, signi�cantly

reduces the size of the standard deviation of quality measures. This �nding seems to be

consistent with the model discussed above, and it also explains the di¤erence in policy

impacts for these two types of nurses.

In addition to the market level analysis, I also perform similar analysis at the state

level. I use quality measures at the market level to calculate quality standard deviation

at the state level and I �nd states with sta¢ ng requirements for direct care nurses have

seen signi�cantly smaller standard deviation (the mean is 0.81 versus 0.88). Moreover, a

�xed e¤ect regression with the inclusion of time and state dummies has shown that the

imposition of minimum sta¢ ng requirements on direct care nurses is found to signi�cantly

reduce standard deviations by 0.1, while sta¢ ng requirements for licensed nurses have no

signi�cant impact on standard deviations. Based on the above analysis, we can conclude

that policies for these two types of nurses seems to a¤ect the decisions of nursing homes in

di¤erent ways. Given extra data on sta¢ ng input at nursing home level, further investigation

37The standard deviation has the mean of 1.19 and 1.23 for markets with and without sta¢ ng requirementsfor direct care nurses. The counterpart number is 1.19 versus 1.22 for markets with and without sta¢ ngrequirements for licensed nurses.

28

on nursing homes�strategic interaction would provide more insightful policy implications.

Meanwhile, it�s also interesting to explore how the extent of asymmetric information a¤ects

strategic interactions among regulated �rms from a theoretical point of view.

8 Conclusion

This paper empirically examines the impacts of minimum sta¢ ng requirements on the nurs-

ing home market using a unique national panel during the time period of 1996 to 2005. The

paper highlights the importance of controlling for unobserved heterogeneity in examining

policy impacts. It also shows that the extent to which one controls for unobserved hetero-

geneity considerably a¤ects the estimation results. The basic �xed e¤ect speci�cation deals

with time-invariant heterogeneity but fails to provide consistent results due to the ignorance

of heterogeneity from other sources. By contrast, the dynamic speci�cations have success-

fully provided more comprehensive controls for unobserved heterogeneity. The estimation

reveals a quality-improving e¤ect from the minimum sta¢ ng of licensed nurses: a half-hour

increase in the minimum sta¢ ng requirement increases quality by 15 percent. Equivalently,

it means one standard deviation increase of minimum licensed nursing hours will improve

quality by four percent. There is no evidence of any e¤ect from the minimum sta¢ ng of

direct care nurses. This �nding has an important policy implication: mandating the quantity

of direct care nursing does not guarantee quality of care.

29

ReferencesAdministration on Aging. 2008. "A Pro�le of Older Americans: 2008." U.S. Depart-

ment of Health and Human Services.

American Health Care Association. 2007. "Trends in Nursing Facility Characteris-

tics." July 2007.

Arrow, Kenneth J. 1971. Essays in the Theory of Risk Bearing. Amsterdam: North

Holland.

Carroll, S., and Gaston, R. 1981. "Occupational Restrictions and the Quality of

Service Received: Some Evidence." Southern Economic Journal, 47 (4): 959-976.

Chen, M. 2008. "Minimum Quality Standards and Strategic Vertical Di¤erentiation:

An Empirical Study of Nursing Homes." working paper.

Chipty, T., and Witte, A. 1995. "Economic E¤ects of Quality Regulations in the

Day-care Industry." American Economic Review, 85 (3): 419-424.

Chipty, T., and Witte, A. 1999. "An Empirical Investigation of Firms�Responses to

Minimum Standards Regulations." Children and Youth Services Review, 21: 111-146.

Cohen, J., and Spector, W. 1996. "The E¤ect of Medicaid Reimbursement on Quality

of Care in Nursing Homes." Journal of Health Economics, 15(1): 23-48.

Crampes, C., and Hollanderb, A. 1995. "Duopoly and Quality Standards." European

Economic Review, 39 (1): 71-82.

Currie, J., and Hotz V. 2004. "Accidents Will Happen? Unintentional Childhood

Injuries and the E¤ects of Child Care Regulations." Journal of Health Economics, 23(1):

25-59.

Decker, H., 2005. "Nursing Homes, 1977�99: What Has Changed, What Has Not?"

U.S. Department of Health and Human Services.

Decker, H., and et al. 2003. "Results of the 2002 AHCA Survey of Nursing Sta¤

Vacancy and Turnover in Nursing Homes.�American Health Care Association.

Friddberg, L. 1998. "Did Unilateral Divorce Raise Divorce Rates? Evidence from Panel

30

Data.�American Economic Review, 88(3): 608-627.

General Accounting O¢ ce. 2001. "Nursing Workforce: Recruitment and Retention

of Nurses and Nurses Aides Is a Growing Concern.�GAO-01-750T.

Gertler, J. 1989. "Subsidies, Quality, and the Regulation of Nursing Homes.�Journal

of Public Economics, 38(1): 33-52.

Grabowski, D. 2001. "Medicaid Reimbursement and the Quality of Nursing Home

Care.�Journal of Health Economics, 20(4): 549-459.

Gruber, J., and Maria, H. 1995. "The Labor Market E¤ects of Introducing National

Health Insurance: Evidence from Canada.� Journal of Business and Economic Statistics,

13(2), 163-73.

Gormley, W. 1991. "State Regulations and the Availability of Child Care Services.�

Journal of Policy Analysis and Management, 10(1): 78-95.

Harrington, C. 2001. "Nursing Home Sta¢ ng Standards in State Statutes and Regu-

lations.�University of California Working Paper.

Harrington, C., and Carrillo, H. 1999. "The Regulation and Enforcement of Federal

Nursing Home Standards, 1991-1997.�Medical Care Research and Review, 56(4): 471-494.

Harrington, C., and et al. 2008. "Nursing Facilities, Sta¢ ng, Residents and Facility

De�ciencies, 2001 Through 2007.�University of California Working Paper.

Harrington, C., and et al. 2000. "Nursing Home Sta¢ ng and Its Relationship to

De�ciencies.�The Journals of Gerontology, 55B(5): 278-287.

Holen, A. 1978. The Economics of Dental Licensing. Final report submitted to the

U.S. Department of Health and Human Services.

Hotz, J., and Xiao, M. 2005. "The Impact of Minimum Quality Standards on Firm

Entry, Exit and Product Quality: The Case of the Child Care Market.�NBER Working

Paper 11873.

Institute of Medicine. 1986. "Improving the Quality of Care in Nursing Homes.�

National Academy Press.

31

Jinji, N., and Toshimitsu, T. 2004. "Minimum Quality Standards under Asymmetric

Duopoly with Endogenous Quality Ordering: A Note.� Journal of Regulatory Economics,

26(2): 189-199.

Kleiner, M., and Kudrle, R. 2000. "Does Regulation A¤ect Economic Outcomes?

The Case of Dentistry.�Journal of Law and Economics, 43(2): 547-582.

Leland, H. 1979. "Quacks, Lemons, and Licensing: A Theory of Minimum Standards.�

Journal of Political Economy, 87(6): 1328-1346.

Lowenberg, A., and Tinnin, T. 1992. "Professional versus Consumer Interests in

Regulation: the Case of the US Child Care Industry.�Applied Economics, 24(6): 571-580.

Matthews-Martin, L. 2003: "Comparing Nursing Home Quality and Performance:

An Evaluation of the Basic Method in Nursing Home Ranking Systems.�American Health

Care Association.

Mehta, A. 2006. "Spatial Competition and Market De�nition in the Nursing Home

Industry.�Boston University Working Paper.

Moulton, B. 1990. "An Illustration of a Pitfall in Estimating the E¤ects of Aggregate

Variables in Micro Units.�Review of Economics and Statistics, 72(2): 334-338.

Mueller, C., and et al. 2006. "Nursing Home Sta¢ ng Standards: Their Relationship

to Nurse Sta¢ ng Levels.�The Gerontologist, 46(1): 74-80.

Mukamel D., and Spector W. 2003. "Quality Report Cards and Nursing Home

Quality.�The Gerontologist, 43(2): 58-66.

Nyman, A. 1985. "Prospective and �Cost-Plus�Medicaid Reimbursement, Excess

Medicaid Demand, and the Quality of Nursing Home Care.�Journal of Health Economics,

43(3): 237-259.

Nyman, A. 1989. "Excess Demand, Consumer Rationality, and the Quality of Care in

Regulated Nursing Homes.�Health Services Research, 24(1): 105�128.

Ronnen, U. 1991. "Minimum Quality Standards, Fixed Costs, and Competition.�Rand

Journal of Economics, 22(4): 490-504.

32

Schnelle, F., and et al. 2004. "Relationship of Nursing Home Sta¢ ng to Quality of

Care.�Health Services Research, 39(2): 225-250.

Shaprio, C. 1983. "Premiums for High Quality Products as Returns to Reputations."

Quarterly Journal of Economics, 98(4): 659-679.

Shapiro, C. 1986. "Investment, Moral Hazard, and Occupational Licensing.�Review

of Economic Studies, 53(5): 843-862.

Siebert R., and Graevenitz G. 2005. "How Licensing Resolves Hold-Up: Evidence

from a Dynamic Panel Data Model with Unobserved Heterogeneity.�CEPR discussion paper

no 5436.

Valletti, T. 2000. "Minimum Quality Standards under Cournot Competition.�Journal

of Regulatory Economics, 18(3): 235-245.

Wiener, M., and et al. 2007. "Nursing Home Care Quality Twenty Years after the

Omnibus Budget Reconciliation Act of 1987.�The Henry J. Kaiser Family Foundation.

Wiggins, S. 1981. "Product Quality Regulation and New Drug Introductions: Some

New Evidence from the 1970s.�Review of Economics and Statistics, 63(4): 615-169.

Zhang, N., and et al. 2006. "Minimum Nurse Sta¢ ng Ratios for Nursing Homes.�

Nursing Economics, 24(2): 78-85.

Zhang, X., and Grabowski, D. 2004. "Nursing Home Sta¢ ng and Quality under

the Nursing Home Reform Act.�The Gerontologist, 44 (1): 13-23.

Zinn, S., 1993. "The In�uence of Nurse Wage Di¤erentials on Nursing Home Sta¢ ng

and Resident Care Decisions.�The Gerontologist, 33(6): 721-729.

33

Table 4: E¤ects of minimum sta¢ ng requirements on the number of nursing homes

The Number of Nursing Homes(1) (2) (3) (4) (5)

OLS FE RT DYI DYII

Licensed Nurses 0.3497*** 0.1151*** 0.0073 0.0019 -0.0016(0.0918) (0.0411) (0.0452) (0.0207) (0.0213)

Direct Care Nurses -0.0081 -0.0175** -0.0162 -0.0012 0.0021(0.0101) (0.0083) (0.0112) (0.0030) (0.0030)

Lag Dependent Var. 0.1007** 0.1293**(0.0497) (0.0520)

Ln(Income) 0.3589*** 0.0113 -0.0041 -0.0226 -0.0205(0.0382) (0.0184) (0.0187) (0.0173) (0.0176)

Ln(Elder) 0.5045*** 0.0996*** 0.0283* 0.0447** 0.0501**(0.0071) (0.0167) (0.0170) (0.0196) (0.0197)

Medicare Rate -0.0007*** -0.0001 0.0001 0.0001 0.0001(0.0002) (0.0001) (0.0001) (0.0001) (0.0001)

Medicaid Rate -0.0035*** -0.0000 -0.0001 0.0002 0.0003(0.0004) (0.0002) (0.0003) (0.0002) (0.0002)

CON -0.0152 0.0037 -0.0107(0.0246) (0.0065) (0.0145)

Constant -5.7735*** 0.6251** 0.0038 -0.0106*** -0.0136***(0.3326) (0.2481) (0.0033) (0.0039) (0.0039)

Observations 24584 24584 21511 15365 15365Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table 5: E¤ects of minimum sta¢ ng requirements on the number of nursing homes and

quality of care, with a policy lead dummy

Number of Homes Count Quality Measure Value Quality Measure(1) (2) (3) (4) (5) (6)RT DYI RT DYI RT DYI

Licensed Nurse 0.0060 0.0061 0.3677** 0.2877** 0.5658** 0.3334**(0.0500) (0.0188) (0.1865) (0.1433) (0.2288) (0.1689)

Direct Care Nurse -0.0165 0.0016 -0.0211 0.0076 -0.1078 0.0268(0.0115) (0.0025) (0.0557) (0.0343) (0.0714) (0.0434)

Lead Policy Dummy -0.0008 0.0063 -0.0447 -0.0308 -0.0986** -0.0845(0.0107) (0.0048) (0.0348) (0.0442) (0.0420) (0.0563)

Lag Dependent Var. 0.1000** 0.2151*** 0.1461***(0.0499) (0.0262) (0.0223)

Observations 21511 15365 17549 15042 17549 15042Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

34

Table 6: E¤ects on quality of care measured by citation counts

The Count Quality Measure(1) (2) (3) (4)FE RT DYI DYII

Licensed Nurse 0.0745 0.4444*** 0.3161** 0.3307**(0.1444) (0.1654) (0.1485) (0.1480)

Direct Care Nurse 0.0548 -0.0046 0.0212 0.0208(0.0353) (0.0580) (0.0340) (0.0340)

Lag Dependent Var. 0.2147*** 0.2131***(0.0263) (0.0263)

Ln(Income) 0.2620 0.0957 0.1774 0.1694(0.2120) (0.2468) (0.2762) (0.2761)

Ln(Elder) -0.3776** 0.3154 0.1449 0.0846(0.1651) (0.2785) (0.2650) (0.2605)

Medicare Rate 0.0141*** 0.0068*** 0.0058** 0.0066***(0.0008) (0.0016) (0.0026) (0.0025)

Medicaid Rate -0.0052** -0.0067** -0.0038 -0.0040(0.0023) (0.0026) (0.0026) (0.0026)

CON -0.0846*** -0.0482(0.0288) (0.0509)

Constant -4.4171* -0.1634*** -0.1272* -0.1426*(2.5129) (0.0555) (0.0751) (0.0741)

Observations 20056 17549 15042 15042Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

35

Table 7: E¤ects on quality of care measured by citation values

The Value Quality Measure(1) (2) (3) (4)FE RT DYI DYII

Licensed Nurse 0.2461 0.7348*** 0.4115** 0.4231**(0.1976) (0.2138) (0.2027) (0.2024)

Direct Care Nurse 0.0702 -0.0715 0.0642 0.0636(0.0592) (0.0749) (0.0443) (0.0443)

Lag Dependent Var. 0.1454*** 0.1437***(0.0223) (0.0222)

Ln(Income) 0.3830 0.1765 0.2840 0.2593(0.2905) (0.3758) (0.4234) (0.4228)

Ln(Elder) -0.3777* 0.7251* 0.4848 0.3875(0.1929) (0.4350) (0.4016) (0.3930)

Medicare Rate 0.0150*** 0.0090*** 0.0072*** 0.0076***(0.0009) (0.0018) (0.0024) (0.0024)

Medicaid Rate -0.0039 -0.0082* -0.0029 -0.0029(0.0030) (0.0042) (0.0035) (0.0035)

CON -0.1304*** -0.0078(0.0501) (0.0631)

Constant -7.4241** -0.1597** -0.0521 -0.0557(3.3233) (0.0740) (0.0767) (0.0766)

Observations 20056 17549 15042 15042Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

36