Organization of Dental Care Industry
BY
THANH-AN NGUYEN-LE
M.D., Pham Ngoc Thach University of Medicine, 2007
M.P.H., University of Illinois at Chicago, 2010
THESIS
Submitted as partial fulfillment of the requirements
for the degree of Doctor of Philosophy in Public Health Sciences
in the Graduate College of the
University of Illinois at Chicago, 2015
Chicago, Illinois
Defense Committee:
Anthony T. Lo Sasso, Chair and Advisor, Division of Health Policy and Administration
Robert Kaestner, Department of Economics
Lindsey Leininger, Mathematica Policy Research
Marko Vujicic, American Dental Association, Health Policy Institute
Coady Wing, Indiana University, School of Public and Environmental Affairs
This dissertation is dedicated to my parents Nguyen Van Hung and Le Thi Thanh My, to my husband
Doan Trung Kien and my daughter Doan Thao Minh.
ii
ACKNOWLEDGMENTS
I would like to express my deepest appreciation to my advisor, Professor Anthony T. LoSasso, for
all of his support and guidance over the past years. I always feel impressed and grateful for the time
Professor LoSasso devotes for his students including me, given his busy schedule.
His immense
knowledge, insightful guidance and encouragement have always inspired me after each of our regular
meetings. This dissertation would not be possible without him. I am also grateful to other members of
my dissertation committee, Professor Robert Kaestner, Professor Lindsey Leininger, Dr. Marko Vujicic,
and Professor Coady Wing, for their valuable feedback that improved my research in many ways.
My dearest gratitude is for my family. To my mom, for nurturing me and my learning dream,
being the mother, the teacher, and the best friend, all at the same time. To my husband, for sacrificing his
own opportunities to always be on my side, support me in my life, share with me the challenges, and not
let me give up.
Last but not least, I would like to thank the Vietnam Education Foundation and the Division of
Health Policy and Administration at UIC School of Public Health for giving me the opportunity to fulfil
my dream of studying.
TANL
iii
TABLE OF CONTENTS
CHAPTER
PAGE
1.
COMPETITION AND MARKET STRUCTURE IN THE DENTAL INDUSTRY .. 1
1.1
Introduction ............................................................................................................. 1
1.2
The theory ............................................................................................................... 3
1.3
The model specification .......................................................................................... 5
1.3.1 The cost function and its derivatives. ..................................................................... 6
1.3.2 The demand function. ........................................................................................... 10
1.4
Data description and variable construction ........................................................... 10
1.5
Missing Data ......................................................................................................... 26
1.6
Results ................................................................................................................... 46
1.7
Estimation stability ............................................................................................... 67
1.8
Conclusion ............................................................................................................ 85
2.
THE GENDER GAP IN DENTIST EARNINGS ........................................................ 86
2.1
Introduction ........................................................................................................... 86
2.2
Methods................................................................................................................. 88
2.2.1 The Oaxaca-Blinder decomposition ..................................................................... 88
2.2.2 The DiNardo-Fortin-Lemieux decomposition ...................................................... 88
2.3
Data sources .......................................................................................................... 91
2.4
Results ................................................................................................................... 92
2.4.1 An Overview of Gender Differences in Dentists’ Earnings and Related
Characteristics ....................................................................................................... 92
2.4.2 The Oaxaca-Blinder decomposition ................................................................... 103
2.4.3 The DiNardo-Fortin-Lemieux decomposition .................................................... 119
2.5
Conclusion .......................................................................................................... 124
CITED LITERATURE ............................................................................................................ 126
VITA ......................................................................................................................................... 132
iv
LIST OF TABLES
TABLE
PAGE
I.
NET INCOME BY EMPLOYMENT STATUS......................................................... 12
II.
DESCRIPTIVE STATISTICS OF ENTIRE PRACTICE VARIABLES ................... 13
III.
CHANGE IN ENTIRE PRACTICE CHARACTERISTICS OVER TIME ............... 14
IV.
OBSERVED AND PREDICTED INPUT FACTOR PRICES................................... 21
V.
CONSTRUCTION OF DENTIST AND HYGIENIST VISIT PRICE ...................... 22
VI.
DESCRIPTIVE STATISTICS OF DENTIST AND HYGIENIST VISIT PRICES .. 23
VII.
DIFFERENCES IN CHARACTERISTICS BETWEEN MISSING AND NONMISSING CASES (C, YD, YH) ................................................................................. 32
VIII.
DIFFERENCES IN CHARACTERISTICS BETWEEN MISSING AND NONMISSING CASES (WD, WH, WO) ........................................................................... 34
IX.
DIFFERENCES IN CHARACTERISTICS BETWEEN MISSING AND NONMISSING CASES (XD, XH) ..................................................................................... 36
X.
DESCRIPTIVE STATISTICS OF VARIABLES INCLUDED IN THE
ESTIMATION ............................................................................................................ 38
XI.
TREND OF VARIABLES IN THE ESTIMATION FOR INCLUDED
OBSERVATIONS (C, YD, YH) ................................................................................ 39
XII.
TREND OF VARIABLES IN THE ESTIMATION FOR INCLUDED
OBSERVATIONS (WDP, WHP, WOP) .................................................................... 42
XIII.
DIFFERENCES BETWEEN INCLUDED AND EXCLUDED OBSERVATIONS
IN THE ESTIMATION .............................................................................................. 45
XIV.
COST FUNCTION RESULTS ................................................................................... 47
XV.
DEMAND FUNCTION RESULTS ........................................................................... 51
XVI.
ESTIMATION RESULTS WITH BOOTSTRAPPED STANDARD ERRORS ....... 58
XVII.
CHARACTERISTICS OF EXTREME VALUES OF RETURNS TO SCALE
BY TRANSLOG ESTIMATION (RT) ...................................................................... 65
v
LIST OF TABLES (continued)
XVIII.
COMPARISON OF COEFFICIENTS IN THE DEMAND FUNCTIONS FOR
DENTIST VISIT AND HYGIENIST VISIT ............................................................. 70
XIX.
THE TRANSLOG COST FUNCTION ESTIMATIONS FROM DIFFERENT
DATA SETS ............................................................................................................... 73
XX.
THE DIEWERT COST FUNCTION ESTIMATIONS FROM DIFFERENT
DATA SETS .............................................................................................................. 78
XXI.
ESTIMATIONS FROM DIFFERENT DATA SETS................................................. 84
XXII.
CHARACTERISTICS OF FEMALE AND MALE DENTISTS OVER TIME
IN SDP DATA ............................................................................................................ 94
XXIII.
UNCONDITIONAL DIFFERENCE IN MEANS BETWEEN FEMALE AND
MALE DENTISTS IN SDP DATA............................................................................ 97
XXIV.
CHARACTERISTICS OF DENTISTS IN THE CENSUS AND ACS DATA ....... 100
XXV.
THE OB DECOMPOSITION OF EARNINGS DIFFERENTIALS FROM SDP
DATA 1982 TO 2011 ............................................................................................... 107
XXVI.
THE OB DECOMPOSITION OF THE CENSUS AND ACS DATA ..................... 115
vi
LIST OF FIGURES
FIGURE
PAGE
1.
Kernel density estimate for the main variables ................................................................. 18
2.
Kernel density estimate for observed and predicted input prices ..................................... 19
3.
Changes in percentage of time spent for procedures in dental practice ............................ 24
4.
Price of dentist visit and hygienist visit over time ............................................................ 25
5.
Percentages of missing values in the main variables (not imputed). ................................ 27
6.
Percentages of missing values of raw and imputed variables. .......................................... 30
7.
Demand curves for dentist visit and hygienist visit .......................................................... 55
8.
Marginal cost of dentist (A) and hygienist (B) visit by Diewert and translog estimation 57
9.
Price elasticity of demand for dentist visit (A) and hygienist visit (B) ............................ 60
10.
Market power index for dentists (A) and hygienists (B) by Diewert estimation .............. 62
11.
Market power index for dentists (A) and hygienists (B) by translog estimation .............. 63
12.
Returns to scale by Diewert (A) and translog (B) estimation ........................................... 64
13.
Earnings densities and the unexplained difference, SDP data, 1982-1996..................... 120
14.
Earnings densities and the unexplained difference, SDP data, 1997-2011..................... 121
15.
Earnings densities and the unexplained difference, IPUMS data, 1990, 2000, and 2007-2011.... 122
vii
LIST OF ABBREVIATIONS
ACS
American Community Survey
ADA
American of Dental Association
DFL
DiNardo, Fortin, and Lemieux
GP
General Practitioner
HHI
Herfindahl-Hirschman Index
HIE
Health Insurance Experiment
IPUMS
Integrated Public Use Microdata Series
OB
Oaxaca Blinder
SDP
Survey of Dental Practice
viii
SUMMARY
This research examines the organization of the dental care industry in two different aspects: the
competition among dental care providers at practice level and the earnings of dentists at individual level.
The dental care market in the United States is mainly made up of numerous private solo practice
providers.
However, it could not be characterized as the perfectly competitive market due to the
complexity in the relationship among consumers, sellers, and payers, as well as the professional
regulations. Chapter one uses repeat cross-section data from 1981 to 2011 of the American Dental
Association Survey of Dental Practice (SDP) to estimate the degree of market power in the private dental
practice, how it has changed over time, and the extent of any economies of scale in the provision of dental
services. I identify market power by estimating an index measuring the ability of a firm to markup the
price above the marginal cost. This parameter is estimated from the price index, the demand function and
the system of equations including the cost function and labor input factor demand functions. Both
specifications of flexible cost functions - the generalized Leontief and the translog cost functions - show
that the dentist services market is monopolistic competitive while the hygienist services market is close to
perfectly competitive. I also find that private dental practice shows significant economies of scale and
demand for dental services are inelastic. The results provide important information about practice
organization and pricing behaviors for public welfare and reimbursement policy.
Chapter two explores the gender difference in dentist earnings. Previous research found an
unexplained gap when accounted for age, experience, working hours, parental status, region and race.
Using national data of the Survey of Dental Practice from 1982 to 2012, I add specialty, entrepreneurship,
productivity, and practice size to the explanatory factors for more complete specification. I also use the
census data and the American Community Survey data from the Integrated Public Use Microdata Series
to examine the role of race, marital status, and parental status, which are not available in the SDP. The
Oaxaca-Blinder decomposition finds the earnings gap in the SDP data reducing from 74% in 1982-1986
to 31% in 2007-2011, with the decrease in the explaining role of observable characteristics from 40% to
ix
SUMMARY (continued)
4% and the relatively consistent unexplained part contributing to 23%-29% of the differential. The
census data shows a larger earnings gap, which could be explained by the lower percentage of female
dentists as practice owners as compared to the SDP data. Race, marital status, and parental status are
unobservable in the SDP analysis, but the census analysis shows that these factors only explain for 4-6%
of the earnings difference. Thus, if all factors are combined in an analysis, I still expect an unexplained
earnings difference of about 20% which remains relatively unchanged over time. The semiparametric
DiNardo-Fortin-Lemieux approach finds that the observable characteristics have stronger effect on the
gap for dentists with lower earnings; while for those at higher earnings distribution, the unexplained gap
is large and consistent.
x
1
1.
1.1
COMPETITION AND MARKET STRUCTURE IN THE DENTAL INDUSTRY
Introduction
In the United States dental care is mainly provided by dentists in private solo practice. This should
create a competitive market with plethora of suppliers. However, given the complexity in the relationship
among consumers, sellers, and payers, as well as the professional regulations, the extent to which dental
health care markets could be described as perfectly competitive is unclear. It has long been known that
dentistry in the U.S. is an internally well-organized industry in which dentists are operating in a strongly
regulated environment but professional regulations are mostly made by market participants (Lipscomb and
Douglass 1982). Local and state dental societies are mainly made up of dentists. These entities decide on
local and state regulations, and ultimately impact national policy under which dentists practice their
profession. Lipscomb and Douglass (1982) concluded that “the professional structure has reinforced
dentistry's leverage over governmental legislation and regulations affecting the profession thereby
influence the market of dental care”. This is a reason for the common perception that dental practice can
exploit some market power to control price. However, recent judgement against the North Carolina State
Board of Dental Examiners (the Board) in the case against the Federal Trade Commission is a sign that
such dynamics in dental care market could be changing. The Board which is made up of six elected
dentists, one hygienist and one consumer member was accused of violating antitrust laws for excluding
non-dentists from the teeth-whitening services market. The main question of the case is whether a state
agency involving market participants is exempt from antitrust laws. The Supreme Court decides that a
state regulatory board operated by market participants is subject to federal antitrust laws unless having
active state supervision (Supreme Court of the United States 2015). The Court decision will affect state
regulatory boards and could particularly change the legal environment of dental practice and dental care
market. Thus, studying the competitive landscape and market structure of the dental care market is policy
relevant. Furthermore, understanding the degree of competition, which in turn affects pricing behaviors in
dental care market, is important for public welfare and reimbursement policy.
1
2
In medicine, there is a rich literature with the objective of characterizing the market structure of
hospitals and physicians. In some studies, researchers directly measure the degree of competition using
the Hirschman-Herfindahl Index (HHI) (Farley, 1985; Zwanziger & Melnick, 1988; Gruber, 1994; Kessler
& McClellan, 2000) or the function of some structural variables such as provider density, population
density, market size, or entry barriers (Joskow, 1980; Robinson & Luft, 1985; Wilson & Jadlow, 1982;
Gruber, 1994). In other studies, researchers infer the market structure by giving evidence of competitive or
anticompetitive conduct such as barriers to entry (Kessel 1970; Rayack 1967), advertising bans (Feldman
and Begun 1978; Hass-Wilson 1986), or pricing behavior (Kessel 1958; McCarthy 1985; Gunning and
Sickles 2013) . Measurement of competition is useful to study its effect on prices, quality of care, and
health outcomes, as well as to examining the determinants of market characteristics such as antitrust work
(L. C. Baker 2001).
Empirical studies dealing with the structure of the dental care market are surprisingly sparse and
dated. Most economic papers on the dental care market focus on estimating price elasticity of demand or
supply, rather than identifying market structure or measuring the degree of competition. Since the 1970s,
several estimates of the demand for dental care have been published by Feldstein (1973), Holtmann and
Olsen (1976), Phelps and Newhouse (1973), and Maurizi (1975). However, they were criticized for
methodology issues including lack of controlling for price variation, environmental and personal
characteristics, or under-identifying of demand curves (Manning and Phelps 1979). Later estimates of
price and co-insurance elasticity of demand by Hu (1981) and Mueller and Monheit (1988) are also
questioned due to small sample size and specification errors (Grembowski et al. 1988). The number of
studies on the supply side is even smaller and they provided mixed results: Kushman and his team made
several efforts to identify whether the dental care market was monopolistic or competitive. In a study
using the supply and demand equation approach they concluded a monopolistic model of dentists’
behavior (Kushman and Scheffler 1978), while in another paper using the marginal product and wage
approach they found that dentists’ behavior was consistent with competitive profit-maximizing (Kushman
3
et al. 1978). As a result, Kushman (1981) suggested that there should be a model including some
characteristics of both market structure extremes. The answer to the question of how competitive the
dental care market is remains unclear based on the prior evidence. Additionally, most studies used crosssectional data from several decades ago. There is a need for a study specifying the structure of dental care
market using recent time-series data. As data on market shares of dental practices are unavailable, I will
not directly measure competition using HHI or practice density. Instead, I will characterize the market
structure by estimating the market power of dental practices in pricing in relative to their marginal costs.
The method is from Bresnahan (1989) and will be described in the next section. The research objective is
to estimate the amount of market power in the private dental practice market, as well as the extent of any
economies of scale in the provision of dental services. The use of the repeat cross-section data allows for
describing the trend of these estimations over time.
1.2
The theory
Economic theory predicts that profit-maximization in a competitive market with costless entry and
exit implies cost-minimization. The equilibrium output level y is chosen such that the marginal revenue
equals the marginal cost. In the perfectly competitive market, firms are price-takers which produce at the
output level y such that the marginal cost equals the equilibrium price MR MC P . The monopolist,
on the other hand, has to reduce price in order to sell an additional output (dP/dy <0); thus, the marginal
revenue is always less than price, as given by:
MR
d[ P * y]
dP
y dP
P y
P 1
dy
dy
P dy
As the price elasticity of demand is given by eD
P dy
, I have
y dP
[1]
4
1
1
MR P 1 P 1
eD
eD
[2]
Combining the formula with the condition for profit maximization, MR MC , Lerner (1934)
proposed a measurement of monopoly power:
Lerner index
P MC
1
P
eD
[3]
The Lerner index implies that the extent to which a monopolist sets the price above the marginal
cost depends on the inverse of the price elasticity of demand. Firms in markets between the two extremes,
perfect competition and monopoly, also have some market power to set the price above the marginal cost.
If the market is highly inelastic, a small increase in price would not significantly change the quantity
demanded, which allows firms to markup more. The extent of the markup implies the degree of market
competition and the elasticity of demand. Based on prior evidence, it is reasonable to presume that the
private dental practice market structure is neither perfect competitive nor monopolistic, and use the
theoretical framework as described in Bresnahan (1989) to measure market power, the ability of providers
to markup price above marginal cost, given the elasticity of demand for dental care services.
Considering the general cost function of a private dental practice i with quantity of output yi, the
marginal cost (MC) function is given by:
Ci C ( yi , wi , Zi , , ci )
MCi
Ci
C1 ( yi , wi , Zi , , i )
yi
[4]
where wi is a vector of factor prices, Zi are exogenous variables that shift cost, are parameters to
be estimated, and i is the error term. The demand function is given by:
5
yi D( Pi , X i , , di )
yi
D1
Pi
[5]
where Xi are exogenous variables that shift demand, are parameters to be estimated, and di is the error
term. The marginal revenue is:
MRi Pi yi
Pi
y
Pi i
yi
D1
[6]
Combining [4] and [6] I have the condition for profit-maximizing or cost-minimizing equilibrium:
C1 Pi
yi
D1
(C P )
1 i D1
yi
[7]
Equation [7] illustrates the relationship between the marginal cost incurred by a firm and the price
it sets. In perfect competition, price equals marginal cost and thus 0 . The marginal revenue for a
monopolist is MR P y / D1 which means 1 . The parameter is therefore an index of the market
competition, measuring the ability of a firm to markup the price above the marginal cost, occurring in
market structures departing from perfect competition. Credibly estimating the parameter is the main
purpose of this research.
1.3
The model specification
The market power index will be estimated from the demand function and the system of three
equations including the cost function and two labor input factor demand functions. The method is adapted
from Escarce and Pauly (1998) and Gunning and Sickles (2013), who estimate physician practice cost
functions as well as the competition and market power of physicians in private practice. Due to the large
number of parameters to be estimated in the cost function, Gunning and Sickles (2013) add the factor
6
demand functions into the system to increase efficiency (as the factor demand functions are the derivative
of the cost function with respect to the factor prices, the parameters of the factor demand functions are
restricted to be identical to the corresponding parameters in the cost function). The specifications of the
equations are detailed below.
1.3.1 The cost function and its derivatives.
I use the multi-product flexible cost function approach, which is common in recent empirical work
to estimate the costs of hospital and physician practice production with multiple outputs. The two outputs
in my context are dentist visits and hygienist visits. Prior to the widespread use of the flexible cost
functions, ad hoc models which estimated average cost as a function of various determinants were used but
often criticized for the lack of theoretical foundation in the assumptions of a production function. The
flexible cost function, on the other hand, is derived from a theoretically appropriate production function
without imposing a priori restrictions on factor substitution elasticities. The most appealing feature of the
flexible cost function that serves my objective is its measurement of scale and scope economies. The
original flexible function only includes output quantities and input prices. I employ the “hybrid” flexible
form which adds other explanatory variables that could affect costs and thus combines the distinct features
of the ad hoc function and the original flexible function. I estimate two different widely used cost
functions in order to compare the robustness of my results across the different underlying models.
The translog cost function. The most popular flexible function in empirical work is the translog
cost function developed by Christensen, Jorgenson, and Lau (1973). The advantage of this form is the
interpretability of coefficients and the manageable number of parameters to be estimated, compared to the
quadratic and the Diewert cost functions. When first proposed, it was applied to single product firms and
did not permit zero output level. In 1980, Caves, Christensen, and Tretheway developed the generalized
translog multiproduct cost function, which replaced the natural log for the outputs in the original translog
function by the Box-Cox transformation. This way, the cost function could allow for zero output level and
for imposition of linear homogeneity in input prices restrictions. Linear homogeneity is the required
7
property of the cost function for the existence of a duality relationship between the cost and the production
function. I apply the generalized translog cost function for two outputs: dentist visits (yd) and hygienist
visits (yh). The Box-Cox transformation is given by:
yiB yi 1 /
[8]
where λ is the Box-Cox parameter; yiB represents ydB and yhB, which are the Box-Cox transformation for
dentist visits and hygienist visits. I estimated a regression of total cost on untransformed outputs, input
factor prices, and control variables to obtain the maximum likelihood estimate of the Box-Cox parameter.
Based on this regression I obtained 0.9530704 .
The hybrid generalized translog cost function has the form:
3
1
1
2
2
ln C 0 d ydB h yhB d ydB
h yhB
dh ydB yhB di ydB ln wi
2
2
i 1
3
3
4
1 3
hi yhB ln wi i ln wi ij ln wi ln w j m Z m T 1
2 i , j 1
i 1
i 1
m 1
[9]
where wi are input factor prices including wage of dentists (w1), wage of hygienists (w2), and office price
(w3); T is a vector of year indicators; Zm are the non-input cost shifters including practice location (urban or
rural area), ownership (incorporated or unincorporated), practice size, and an indicator of public program
participation; and is the error term. From the cost function given by equation [9], I can derive the
marginal cost functions:
8
C
MCd
yd
C * ln C
yd
1
2
d ydB 2 d ydB dh ydB yhB 3 di ydB ln wi
C
y
y
y
yd
i 1
d
d
d
2
1
2
d yd 1 /
3
d yd 1 /
y
y
1
/
ln
w
y
1
/
dh
hB
d
di
i
d
2
C
yd
yd
yd
yd
i 1
3
d yd 1 yd 1
1
C d yd
dh yhB yd 1 yd 1 di ln wi
i 1
3
[10]
C d d ydB dh yhB di ln wi y ,
i 1
1
d
for dentist visits.
Similarly,
MCh
C
yh
C * ln C
yh
3
C h h yhB dh ydB hi ln wi yh 1 ,
i 1
for hygienist visits
Differentiating the cost function with respect to input prices yields the input factor demand
equations. I estimate two labor input factors including dentists and hygienists.
ln C
d 1 ydB h1 yhB 1 11 ln w1 12 ln w2 13 ln w3
ln w1
[11]
ln C
d 2 ydB h 2 yhB 2 22 ln w2 12 ln w1 23 ln w3
ln w2
[12]
ln X d
ln X h
I estimate the translog cost function parameters by the seemingly unrelated regression of three
equations [9], [11] and [12]. I impose the following restrictions for the cost function to be homogeneous
of degree one and symmetric in factor prices:
9
3
i 1
i
1
11 12 13 0
12 22 23 0
13 23 33 0
d1 d 2 d 3 0
h1 h 2 h 3 0
ij ji
[13]
The Diewert cost function. Another form of the cost function that I use, for comparison and
robustness checking with the translog form, is the generalized Leontief cost function developed by Diewert
(1971). An advantage of the Diewert cost function is that it can handle zero values of outputs and inputs,
which appear in the data as a practice can have either dentists or hygienists or both. In addition, as the
Diewert cost function is linear homogeneous in factor prices by construction, I do not need to impose
restrictions on parameters as in the case of the translog form thus allow more flexibility.
The hybrid generalized Leontief cost function has the form:
C 0 yd
y
3
2
d
i 1
di
3
1i j 3
wi y
dij wi1/ 2 w1/j 2 yd di wi yh
i 1
3
2
h
i 1
hi
1i j 3
3
hij wi1/ 2 w1/j 2 yh hi wi
i 1
3
3
4
i 1
i 1
m 1
wi yd yh i wi i wi m Z m T 1
[14]
The marginal cost functions derived from the generalized Leontief cost function are specified as:
MCd
3
3
3
C
dij wi1/ 2 w1/j 2 di wi 2 yd di wi yh i wi
yd 1i j 3
i 1
i 1
i 1
3
3
3
C
MCh
hij wi1/ 2 w1/j 2 hi wi 2 yh hi wi yd i wi
yh 1i j 3
i 1
i 1
i 1
[15]
The two labor input factor demand equations to be estimated together with the cost function in the
seemingly unrelated regression are:
10
Xd
Xh
C
w1
C
w2
1
2
1
2
3
yd d 1 j
j 2
yd d 2 j
j 1,3
w1/j 2
1/ 2
1
w
w1/j 2
w1/2 2
yd d 1
yd d 2
1
2
1
2
3
yh h1 j
j 2
yh h 2 j
j 1,3
w1/j 2
1/ 2
1
w
w1/j 2
w1/2 2
yh h1 yd2 d 1 yh2 h1 yd yh 1 1
[16]
yh h 2 yd2 d 2 yh2 h 2 yd yh 2 2
[17]
1.3.2 The demand function.
I specify the demand function linearly in price, its second-order term, practice characteristics
(vector of variables Z m as in the cost equation) and time variable.
6
4
y 0 1P 2 P 2 i Z m 10T 3
[18]
i 3 m 1
1.4
Data description and variable construction
I use 31 years of data from the American Dental Association (ADA) Survey of Dental Practice
(SDP) from 1982 to 2012. The SDP is an annual survey conducted by the ADA sent to a random sample
of general practitioners and specialists in private practice across the U.S., regardless of ADA membership.
The sample is drawn with simple random probability method, from the ADA Sampling Frame which
includes all active dentists who graduated from an accredited dental school in the U.S. and work in private
practice. The response rates vary from 30 to 50 percent (American Dental Association, 2011; Vujicic,
Lazar, Wall, & Munson, 2012). The questionnaire includes a core set of questions that remain roughly
unchanged over the years and other questions that vary depending on whether a short form or long form
version of the survey was administered in a given year. In 1982, a part of the sample received the short
form questionnaire while the rest received the long form. In all other years, the entire sample received the
same questionnaire. The long form was sent out every third year. The survey asked about the experiences
- Xem thêm -