THE ROLE OF PARENTING STYLE IN CHILD - ResearchGate
Transcript of THE ROLE OF PARENTING STYLE IN CHILD - ResearchGate
THE ROLE OF PARENTING STYLE IN CHILD SUBSTANCE USE
DISSERTATION
Presented in Partial Fulfillment of the Requirements
for the Degree Doctor of Philosophy
in the Graduate School of The Ohio State University
By
Garima Malik, M.A.
* * * * *
The Ohio State University 2005
Dissertation Committee: Approved by Professor Hajime Miyazaki, Adviser Professor Lucia Dunn Professor Raymond Montemayor Professor Patricia Reagan ________________________ Adviser Economics Graduate Program
Copyright by
Garima Malik
2005
ii
ABSTRACT
Strategic interactions between a parent and a child within a family have been deemed
important in predicting the behaviour of the child. The dissertation adopts an inter-
disciplinary approach that uses the methodology of development psychology and the
economics of incentives in order to develop an estimable model of parenting styles on
substance use by children ages 10-14. The dissertation relies on the Baumrind
classification of �authoritative�, �authoritarian�, �permissive� and �disengaged�
parenting types, and constructs parenting styles according to the dimensions of
demandingness and responsiveness. The economics of this dissertation relies on an
underlying economics of intrahousehold bargaining reasoning that interactions between
the parent and the child influence the child�s decision on substance use. The model is
solved based on exogenous parenting style but parenting style could be taken as
�endogenous� as the data rejects the hypothesis of no switching convincingly. The
NLSY-79 Mother-Child dataset is used and in the empirical specification a probit model
is used for the different forms of substance use by the child to estimate the probabilities
of taking substances. Disengaged parents are most likely to have children smoking and
consuming alcohol and authoritative parents are least likely to have children smoking and
consuming alcohol. The dissertation establishes the importance of family background
factors in determining substance use, including parental substance use. Thus the
iii
economic models of household bargaining can be supplemented with variables from the
development psychology literature.
iv
VITA
September 21, 1975����..���� Born � India
August, 1996����������� B.A. Economics, Delhi University
August, 1998�����������M.A. Economics, Delhi School of Economics, Delhi University August, 1999�����������M.A. Economics, The Ohio State University
1998-2004������������Graduate Research and Teaching Associate, The Ohio State University
FIELDS OF STUDY
Major Fields: Economics
v
TABLE OF CONTENTS
Page
Abstract����������������������������. ii
Vita .������������������������.�����.. iv
List of Tables��������������������������. vii
List of Figures�������������������������� ix
Chapters:
1. Introduction
1.1 Introduction��������������������������.1
2. Literature Review
2.1 Parenting Style/Parent-Adolescent Relations .�������������2 2.2 Family Economics����������������������..�.7 2.3 Economics Literature on Incentives and Bargaining...���������..11
3. Towards an Economic Model of Incentives and Bargaining in the Parent-Child Relationship
3.1 Introduction�����.��������������������....15 3.2 Pecuniary or Non-pecuniary aspects����������������...16 3.3 Altruistic Parents�����������������������.�19
3.4 Constrained Nash Bargaining�����������������.��22 3.5 Child�s Incentive Compatibility Constraint�������������....23 3.6 Empirical Comparative Statics������������������...25
vi
4. Classification Scheme
4.1 Classification Schemes���������������������27 4.2 Parenting Styles�����������������������...28 4.3 Psychological Models���������������������..32
5. Empirical Model
5.1 Introduction�������������������������. 36 5.2 Factor Analytic Structural Model�����������������37 5.3 Estimated Probit Model��������������������.. 41 6. Data Description
6.1 Introduction�������������������������..44 6.2 NLSY-79 Dataset description and properties������������..49 6.3 Description of variables and coding scheme������������...54 6.4 Scaling characteristics and cronbach alpha results����������..57 6.5 Reliability and internal consistency����������������.58
7. Results Section 7.1 Results���������������������������. 61 7.2 Correlations������������������������.�.68 7.3 Discussion and Interpretation������������������..69 8. Conclusion 8.1 Conclusion�������������������������.. 74 8.2 Policy Recommendations��������������������75 Bibliography����������������������������.� 81
vii
LIST OF TABLES
Tables Page
6.1 Sample Deletion for different groups��������������. 46
6.2 Variable Definitions���������������������. 47
6.3 Descriptive Statistics��������������������� 48
6.4 T-test for differences in means across samples����������. . 49
6.5 Switching parenting styles across different years for the same child��. 54
6.6 Switching parenting styles across different children for the same year�� 54
6.7 Responsiveness subscale for the parenting style inventory������.. 57 6.8 Demandingness subscale for the parenting style inventory������.. 58 7.1 Random Effects OLS model for BPI��������������� 62
7.2 Random Effects Probit Analysis of Smoking������������ 63
7.3 Tobit Model for Smoking Intensity���������������� 64
7.4 Random Effects Probit Model of Smoking with race interactions���� 65
7.5 Random Effects Probit Model for Alcohol������������.. 66
7.6 Tobit Model for Alcohol Intensity���������������� 66
7.7 Random Effects Probit Model for Alcohol with race interactions���� 67
7.8 Correlations between BPI and Smoking�������������.� 69
7.9 Correlations between BPI-Externalizing and Smoking�������.� 69
8.1 Means and Standard Deviations under parenting styles across gender��� 77
viii
8.2 Cronbach alpha measures of coefficient�������������� 78
8.3 Means and variances of factor analysis items������������ 78
ix
LIST OF FIGURES
Figures Page
4.1 Parenting Styles������������������������. 29
4.2 Ecological Model������������������������ 32
4.3 Theoretical Model of Risk-Taking Propensity�..����������� 34
4.4 Schematic Parent-Child Model�����������������.� 35
5.1 Schematic Representation of common factor theory.......................................... 40
8.1 Eigenvalues plot���..��������������������.� 79
8.2 Trends in consumption of smoking and alcohol consumption������� 80
1
CHAPTER 1: INTRODUCTION
This dissertation contributes to the economics of family where household decision
making is made as household members bargain among themselves. My research focuses
on the parent-child interactions in the context of household bargaining. In Chapter 2 I
review the literature on parenting style and child substance use critically both from the
economics of incentives and the psychology of parenting styles on child outcomes. In
Chapter 3 I present a basic economic model of interactions between parents and children
from a game-theoretic viewpoint, and draw several predictions about the relationship
between parenting style and child behaviour. In Chapter 4, I discuss and examine the
classification scheme of parenting styles that is used in development psychology. In
Chapter 5, I present the empirical model which I transform into two different econometric
models for testing purposes. One econometric model uses a single equation probit
specification and the other model. The results of the probit model indicate that
disengaged parenting styles are most likely to lead to increases in the probability of
children taking substances. Authoritative parenting styles leads to the maximum
reductions in the probability of children taking substances. In Chapter 7 I present the
results of the probit models. Therefore both the theoretical and empirical results of this
dissertation demonstrate the potential of the ways in which parenting style can provide a
reasonable and cogent explanation for substance use by young children.
2
CHAPTER 2: LITERATURE REVIEW
There are three basic strands of literature which motivated this research: (1) parenting
styles/parent-adolescent relations; (2) family economics including health
economics/health capital; and (3) economics of incentives, bargaining and behavioral
economics. The discussion below will identify important contributions already available
in each of these areas and will indicate where the present research expands and
contributes to the existing knowledge.
2.1 Parenting Style/Parent-Adolescent Relations
The first of the three major strands comes from development psychology. In this
section, I briefly discuss the methodology and existing contributions form that literature.
More details of classifications, concepts, and methodology from development psychology
are presented in later chapters.
Baumrind classification
This strain of research developed from the seminal paper by Baumrind (1966). In this
work, Baumrind predicts that authoritative parents are more likely to be able to protect
their children from substance use. A general observation from many researchers is that
3
children thrive developmentally when the family environment is characterized by warmth
and psychological autonomy.
Baumrind�s early research created the parenting typologies of �authoritative�,
�authoritarian�, �permissive� and �disengaged� parents along the two dimensions of
demandingness and responsiveness. Demandingness denotes the expectation of parents
for mature behaviour from their adolescent children by setting and consistently enforcing
reasonable rules and standards for their behaviour. Responsiveness refers to parental
warmth and demonstration of physical affection towards the child. Authoritarian parents
attempt to shape, control and evaluate the behaviour and attitudes of children based on
absolute sets of standards, respect for authority and obedience. Authoritarian parents are
more likely to use harsher forms of punishment and are less responsive to the children.
Authoritative parents encourage verbal give and take, explain the reasons behind
demands and discipline, and expect the child to be independent and self-directing. Thus
authoritative parents are both demanding and responsive. Permissive parents are more
likely to give way to the child�s impulses, desires and actions. Permissive parents are less
demanding and more responsive and could be indulgent while disengaged parents are
neither demanding nor responsive and could be termed as neglectful parents.
This dissertation draws on the four-fold parenting style classification given in
Baumrind (1966) and subsequent studies by Baumrind (1991). It explores the two factors
of demandingness and responsiveness. The parenting styles are on the right- hand side as
4
explanatory variables in the empirical framework as well as in the theoretical model
where the parents� utility functions are classified according to the different parenting
styles.
Ecological Factors
There have been studies in development psychology where parenting styles are
classified along the dimensions of support, attachment and learning. These
classifications were not as appropriate for this research as those dimensions would
require �parenting inventories� which are not available here.
This dissertation adapts Bronfrenbrenner�s (1979) model of the ecology of human
development to the parent-child association where the child is at the centre of the system
and is surrounded by the microsystem, which includes parenting factors as well as
neighbourhood factors, which constitute the mesosystem. The ecological paradigm began
with the Lewin model (i.e., Lewin�s Behaviour = f (Person, Environment)) where humans
are active and shape the environments in which they live. In the context of adolescent
psychological development, there are individual factors such as the child�s own
propensity to consume substances, as well as the microsystem and the mesosystem,
which influence the child�s behaviour. Thus the adolescent growing up in the household
has behaviour which consists of the core components of personality and peripheral
components that are constantly adapting to the environment.
5
Maccoby (1990) distinguished parents along the dimensions of permissiveness and
warmth. Maccoby (1980) explored the aspects of parental control stressing the
classifications such as consistent enforcement of demands and rules, high expectations
and training, restrictive parenting, arbitrary power assertion or authoritarian parenting,
open communication patterns and parental warmth and affection.
Steinberg (1982) is the next study in this line of literature which focuses on the
dimensions of acceptance/involvement, strictness/supervision, psychology autonomy
granting, parental involvement in schooling, parental encouragement to succeed, school
performance and school engagement. There are studies by Baumrind (1977) and
Maccoby and Martin (1983) which analyzed parental behaviour through dimensions such
as parental warmth, acceptance, involvement, parental control or strictness. Lamberg,
Mounts, Steinberg and Dornbusch (1991) discus similar dimensions of parenting
practices. The importance of the family in connection to the child�s social and cognitive
development has been highlighted in child development and family studies literature. All
such familial variables that can affect child outcomes � for example parental dispositions,
marital and sibling influences, and the sociocultural context in which the family operates
� are all considered within the interactions between the parent and the child. The parent-
child interaction is characterized by two major parenting dimensions: nurturance (warmth
and support) and control (supervision and discipline). Inadequate parenting characterized
by lack of affection and high levels of criticism and hostility, inconsistent discipline and
supervision, general lack of involvement, provides the foundation for the development of
an aggressive, antisocial behaviour pattern. In addition to parental drinking, there is a
broad range of family influences associated with alcohol problems and externalizing
6
behaviours such as antisocial behaviour and aggression. The family background of
alcohol and other drug users is mostly characterized by marital instability, lack of
support, poor discipline and family conflict.
Techniques and Methodology of Scaling
Scaling techniques were used to generate the different Cronbach alpha values to
check for internal consistency. There are an equal number of items under each scale.
Scaling involves the assignment of objects to numbers according to a rule. Scaling is
different from a response scale, where scales assign numbers according to a common
rule. Scaling is used in this context to see how well these questions �hang together� and
in some instances to score all the responses to generate a single number that represents
the overall construct. Thus a scale refers to a set of items and each item on a scale has a
scale value.
There are three major types of uni-dimensional scaling methods. Scale construction
involves the creation of empirical measures for theoretical constructs and these measures
mostly consist of several items. The process of measurement involves the assignment of
numbers to empirical realizations of the variables of interest. In Thurstone and Guttman
scales, each item represents degrees of the variables of interest, such as the difficulty of
an item. In Likert and Semantic Differential Scales, each item represents degrees of the
variables of interest in a somewhat different way. The differences between the scales
affect the computation of reliability. In the construction of the different types of scales,
many items have been used. The concept of directiveness has been an important element
7
used in the construction of scales (see studies by Courts (1966), Bracken, Brunch, Keith
& Keith (2000), Murphy & Davidshofer (2001) and Schneeweiss & Mathes (1995)).
Scaling has been used as a method to recognize the patterns of the inter-item
correlations which exist among different items within the potential variables of interest.
There are theoretical justifications which underly scale construction. Exploratory factor
analysis was used to isolate three factors for use in this dissertation: (1) acceptance/
involvement (2) strictness/supervision; and (3) psychological autonomy.
2.2 Family Economics
Intergenerational Human Capital Models
The most relevant study to date is by Akabayashi (1998). It uses the NLSY-Child
dataset linking the parent and child in an inter-generational human capital framework
which takes parental incentives as endogenous to examine their influence on cognitive
and behavioural indicators. The NLSY-Child dataset data on children collected through
self-administered questionnaires, as well as extensive information on parental substance
and family background variables.
A longitudinal study by Brook (1990) focused on parenting variables as the major
psychosocial influence on a child�s development of alcohol use and abuse patterns.
Brooke found that the level of mutual warmth, support, and control within the parent-
adolescent relationship predicts the significance of the risk of adolescent drug use.
Adolescent personality characteristics such as sensation-seeking, rebelliousness, and
tolerance for deviance were robust predictors for adolescent alcohol use. A positive
relationship between the parent and adolescent served as a protective factor offsetting the
8
risk of alcohol use associated with peer alcohol use. Studies of the individual effects have
included the role of the parent role modeling and alcohol expectancies in determining the
behaviour of children of alcoholics. Dyadic effects come from the interactions of two
family members, focusing on the parents� marital relationship and the child�s relationship
with the siblings.
Household Economics
Hao, Hotz and Jin (2003) present a game-theoretic model involving parents and
daughters. They further test the model on different family formation structures. In
families which typically have multiple siblings, daughter and family-specific fixed effects
are used. The impact of families on juvenile substance use is examined in Mach (2001),
who looks at the impact of families on juvenile substance use using the NLSY97 dataset1.
She finds that family formation can be an important factor in explaining juvenile crime.
This approach uses county crime rates to look at the influence of parents as well as
siblings on the consumption of substances by youth
Among the various dyads, the parent-child relationship has received the most
attention in the study of alcohol-specific family influences. These dyads are divided
along the lines of father-daughter, father-son, mother-son, and mother-daughter
relationships. With regard to parenting effects on alcohol use, strong associations were
found between the child�s conduct disorder/adolescent delinquency/adult antisocial
behaviour and adult alcoholism. Almost 20% of alcoholics studied met the criteria for
1 The National Longitudinal Surveys of Youth 1997 (NLSY97) is one of the biannual rounds of the survey which individuals born between 1957 and 1964. More detail is given in Chapter 6.
9
antisocial personality disorder, which is characterized by a disregard for and violation of
the rights of others. The associations between antisocial personality disorder and
alcoholism indicate that a parent-child interaction that promotes aggressive, antisocial
behaviour plays a role in the alcoholism of both children of alcoholics and of non-
alcoholics.
The family has been recognized as the primary support system and socializing
institution for children; the better the family operates, the more likely it is that a child will
develop well. Fundamental to positive family dynamics are the relationships that parents
develop with their children. Parental support is significantly related to child and
adolescent development and also to other behaviours which are less deviant than the ones
focused on here.
The study of the relationship among adolescents� reckless behaviours, employment,
and opportunities for risk-taking and parenting styles is not new to psychology or
economics. However, a full understanding of adolescence behavour requires
consideration of the rapidly changing individual in a developmental context.
There is an extensive literature in development psychology and behavioural economics
that seeks to explain key background variables impacting children�s cognitive and
behavioural development.2 These variables include children�s and parent�s background
factors, poverty status, parent�s cognitive support and other key parenting measures. It is
essential to understand what support or protective measures can help children overcome
these risk factors. Mothers� cognitive ability, represented by a mother�s low intelligence
2 See references at end.
10
quotient, is shown to have detrimental effects on her children. It has also been shown that
lower academic levels result in adverse outcomes such as poor parenting.
There are links between poverty, economic resources, and child outcomes. For
children who face persistent poverty, there are substantial developmental deficits.
Low-income families may not be able to afford adequate food, shelter and other
material goods or to provide a warm and stimulating home environment. These absent
conditions are the ones which if present would foster healthy cognitive and social
development of children. Thus economic deprivation has been linked to both
externalizing and internalizing behaviour problems among children and adolescents.3
However, as explained in the NLSY79 Handbook, the effects of low socioeconomic
status are more clear for externalizing problems than internalizing problems. Economic
resources account for half of the difference in children�s outcomes for single versus two-
parent families, thus providing a powerful explanation for family structure effects on
academic achievement and cognitive outcomes. However, even when income is
controlled, children from disrupted families demonstrate greater behaviour problems than
children from intact families, indicating that differences in economic resources do not
fully account for the effect of family structure on child outcomes.
Health Economics/Health Capital
In the case of alcohol and smoking, linear regression models can be used to measure
the amount of alcohol and smoking consumption. When we focus on a decision to smoke
3 Internalizing behaviour includes depression and anxiety. Externalizing behaviour includes hyperactivity and antisocial actions.
11
or not or to consume alcohol or not, however, the appropriate model is of a discrete
nature with a probabilistic outcome being regressed on all the explanatory variables.
There are studies in the Health Economics literature by Hill (1987), Seo (1998), Yin
(2000), Lane, Gerstein, Huang & Wright (1997) which deal with alcohol and smoking as
continuous or discrete variables.
The literature also has an extensive section on the ethnic and ecological perspectives
on socialization in Family Theory. Our purpose here is to present empirical
generalizations and theoretical propositions about relationships between characteristics of
children and parent variables of support. Thus we will evaluate the generalizations and
propositions in social psychology.
2.3 Economics Literature on Incentives and Bargaining
Inter-generational human capital models and models of household bargaining have
been used to examine the effects of parental behaviour. The economic view of
psychological interactions between parents and children considers the effects of family
background as basic human capital with inputs chosen by both parents and children
interactively. Human capital theory has contributed to explaining the level, pattern and
interpersonal distribution of life cycle earnings.
The issue of endogeneity in these kinds of transmissions of family endowments is
important to the extent that this enables a deeper understanding of the theory of
intergenerational mobility. Intergenerational human capital formation is distinct from
�self� investment in human capital.
12
The existing intergenerational human capital theories can be applied to the early
formation of human capital models since children in earlier stages of development must
be different from adults and fully controlled investment in children is not possible by
parents. There are only certain kinds of behaviours which are considered appropriate in
these situations in the process called �socialization�. Personality is a set of characteristics
which determine how individuals respond to experiences and how they get along with
others and themselves. A competent child is characterized as possessing independent,
self-reliant, self-controlled, explorative, and self-assertive, high linguistic, analytic and
logical abilities.
Psychology is concerned with the structure and components of family influence on
several dimensions of children�s development- cognitive, emotional and psychological.
They are exploring the relationship among the adolescents� reckless behaviours (i.e.
alcohol use and non-normative behaviour), parenting practices, adolescents� employment,
and adolescents� opportunities for risk-taking.
The propensity event theory examines the opportunity variables which can mediate the
effects of other explanatory variables on the adolescent�s participate in risky behaviours.
This model of risk-taking behaviour is adapted from Irwin & Millstein (1986) and the
definition of risk-taking inherent in psychosocial development is that risk-taking is a
result of an interaction between the biopsychosocial processes of adolescence and the
environment.
The development psychology literature accepts that some risk-taking is necessary in
the natural developmental process, but extreme forms of risk-taking have severe
consequences. The underlying strand of thinking indicates that young children do not
13
have an adequate understanding of the long-term consequences of their actions and
therefore may take actions that are potentially destructive in nature.
An economic study by Weinberg (2000) examines the impact of parental income on
the incentives offered to children where the more punitive measures are being offered by
low-income households such as grounding while measures such as taking away the
child�s allowance are being offered by higher income households. This is investigated
through an incentive model of parental actions. These economic studies use rational
choice models and the utilitarian framework to model these family choices. This
approach focuses on the �Black-box� and derives reasonable equilibrium results and
testable predictions for the model. However one potential gap within these approaches is
it does not adequately capture the role of parenting styles in these choice theories and
thus does not use the psychological theories to emphasize the association between parents
and children.
Economic Psychology and Behavioural Economics
The literature in the area of economic psychology and behavioural economics deals
with self-control and addiction behaviours. This includes the part on rationality of
decision making processes and the cognitive influences.
There is an extensive literature which examines risky behaviours in the context of the
societal influences. The game theoretic models essentially model rational, socially
14
interdependent decision making processes with psychological explanations as described
in details in Brandstatter and Guth (1994).
There is an extensive literature on the psychology of risk-taking find that it is
positively correlated with behavioural aberrations such as creative activity, courage and
resiliency. Lipsitt and Mitnick (1991) examines different self-regulatory behaviours and
the causal factors of such behaviours.
15
CHAPTER 3:
TOWARDS AN ECONOMIC MODEL OF INCENTIVES AND BARGAINING IN
THE PARENT-CHILD RELATIONSHIP
3.1 Introduction
In this chapter, I discuss how the economic model implied in this research differs
from the standard principal-agency models that economists use. To economize on
terminology and usage according to the standard principal-agent model, in this chapter I
shall sometimes use a one parent vs. one child framework. To interpret the parent-child
relationship, it may appear to be straightforward that the parent is the principal and the
child is the agent. This is correct in a sense that a family, and the parent-child
relationship in the present context, can be regarded as an implicit long term contractual
relation between them. And, as in the economics of all principal-agency relations, the
parent has instruments and authority in attempting to induce a desirable behaviour from a
child. I shall illustrate how the underlying economic model of this dissertation
nevertheless differs from the usual principal-agency framework at least in two aspects:
the parent�s objective function and tacit use of contractual instruments under the parent�s
disposal.
16
3.2 Pecuniary or Nonpecuniary Aspects
The standard story is that the principle can enforce a carrot (reward) and stick
(penalty) rule on the agent to induce a desirable outcome, a good behavior in the present
context.
In this regard, the nature of the reward and penalty in the parent-child relation is not
necessarily pecuniary. The standard principal-agency model essentially says that the
agent be rewarded with a bonus for a �good� outcome and penalized with a severe
payment reduction for a �bad� outcome, in such a way to maintain a sufficiently great
gap between the bonus and penalty in order to discourage the agent from taking a wrong
action that leads to a bad outcome. Such a bonus-penalty scheme is generally understood
to be a pecuniary measure. There can be a practical problem with the pecuniary measure
for parents in dealing with children, however. For example, it is arguable that
conditioning a child for monetary bonus for a good outcome can lead to a failure to foster
ethical or moral elements in shaping and maintaining a good life style in the long run, and
thus that concerned parents might not wish to use monetary means on a routine basis.
Also, the amount of bonus and penalty might practically have a ceiling and a floor
especially in the context of an intra-family relation and thus a range in which pecuniary
measure can be used may be quite limited.
In the parent-child context, the parent can give bonus allowances to reward the child�s
good outcome, such as a high school grade. But, there are many ways other than
17
monetary allowance that the parent can give to the child, even if a pecuniary action is
extended to include purchasable entities such as gifts. For example, it may take the form
of nonmarketable gift, such as a home made gift, an extra vacation day, and other means
showing the parent�s affection. Sometimes, verbal compliments can mean more than a
monetary allowance, and of course a show of affection.
A simple pecuniary penalty is a reduction in the child�s weekly allowance or in
monetizable gift. Instead of pecuniary means, parents have recourse to a wide range of
nonpecuniary means to discipline and reward children. An example of a nonpecuniary
penalty can be an extra household duty, or taking away household privileges, although a
classic textbook example may be �scolding� or �spanking�. These nonpecuniary
instruments of interaction with children are important in the context of the family
relationship that is not inherently monetary. If one accepts a thesis that a significant
content of a typical parent-child relationship has nonpecuniary contents, one realizes that
the implicit social contract between parents and children typically has elements that are
not necessarily monetizable or market mediated. A mechanism to enforce implicit
contractual terms must necessarily then depend on some form on internal nonpecuniary
means. A parent can take actions that directly affect a child�s utility, and in practice the
history of these nonpecuniary actions taken by parents will possibly come to play
important elements in the formation of the parent-child relationship.
Technically, in terms of the principal-agency modeling, the availability of
nonpecuniary means relaxes the so-called bankruptcy constraint and enable the principal
a wider range of Draconian penalties. Nevertheless, in the parent-child relationship, the
parent may not wish to take a nonpecuniary Draconian measure for the same reason that
18
pecuniary Draconian measure may not be credible. In addition, the parents� objective in
practice may not be just their own utility but may also take into account the child�s
utility.
Demanding parent perhaps use sharper demarcations in the use of rewards and
penalties in dealing with a child. But, it is not a priori clear whether demandingness per
se should translate into nonpecuniary or pecuniary measure of the reward-penalty
scheme. In terms of responsiveness, one may be tempted to associate a more responsive
parenting style with nonpecuniary means of reward and penalty. Again, however, it is
perhaps an empirical question to see if this indeed holds up. After all, an authoritative
parent, the parenting type usually considered desirable, can be both responsive and
demanding. It should also be noted that the parent can use the mixture of pecuniary and
nonpecuniary means in the implicit principal-agency relation with the child. For
example, the parent may use monetary bonus to reward, but make recourse to
nonpecuniary means for a penalty. It is worthy to note in this regard that Weinberg
(2000) reports the choice of pecuniary or nonpecuniary means depends on the parents�
income classes. More specifically, he reports that more affluent parents tend to rely more
on pecuniary rewards and penalty than less affluent parents. The ready availability of
both pecuniary and nonpecuniary means should be reckoned within any principal-agency
applications towards the analysis of a parent-child relationship. In other words, an
implicit social contract between a parent and a child contains both pecuniary and
nonpecuniary aspects as terms of contractual enforcement and agreement.
19
3.3 Altruistic Parents
It has been known both from economics and from development psychology4 that
parents have altruistic motives towards their children and behave accordingly5.
Economists� favorite subject is parents� altruistic bequest motive as they arise in
intertemporal macro economics and public finance. But, there are many altruistic
objectives and actions that occur daily between parents and children. For children who
are taking substances parental altruism could take several forms. An example of parental
altruism is �discussing the consequences of using and choosing not to use tobacco,
alcohol and other drugs�. Another example is �expose the child to media influences on
tobacco, alcohol and other drugs�. These are the ways a parent could protect the child
from substance use and show empathy with the child�s actions.
I take a standard economic approach to model the parent�s altruistic motives in terms
of the parent�s utility. Namely, the parent�s utility depends on the child�s utility6. This is
to say that the parent is happier when the child is happier, all other things equal. Letting
the parent�s utility be U, and the child�s utility V, I can specify in the most general
fashion the parent�s utility as U(• , V(• )) where • signifies a vector of consumption
4 Batson (1987, 1991), Rushton (1980), Hoffman (1976), Krebs (1975), Dovidio & Penner (2001), Schroeder, Penner, Dovidio & Piliavin (1995) and Monroe (1996) are references from development psychology. Abel (1998), Cardia & Michel (2004), Kurz (1997), Bernheim & Stark (1988) and Andreoni (1989) are references from economics. 5 Altruism refers to genuinely motivated helping actions that may or may not involve significant risk to self. Such altruism might be a manifestation of a sense of self-identity and expectation associated with social context and super-ego or of an innate empathic capacity that reacts immediately and instinctively to the other when imperiled.
6 The specification adopted here is not the only way that economists may postulate the parental altruism. An alternative specification of the parental altruism is to postulate that the child�s outcome directly enters the parent�s utility function. Such a specification will also occur, if the child directly provides utility to the parent�s utility.
20
bundles and outcomes that may affect utility valuations. An altruistic parent will have its
U monotone increasing in the child�s utility V. I specify further that the parent�s utility is
of a generalized Cobb-Douglas type, namely U(• , V(• )) = Uα(• ) V1- α(• ) where 0 ≤ α
≤ 1. Further, assuming that I can normalize the utility scale to be positive numbers, I
take the logarithmic transform to the form,
α log U(•) + (1 � α) log V(•).
In the logarithmic specification, it is transparent that a lower α corresponds to a more
altruistic parents.
Of the four stereotypes of parenting, authoritative, authoritarian, permissive, and
disengaged, the authoritative parent type is by definition most prominently altruistic.
But, in terms of technical definition of altruism that I use in this section, it is not a priori
decidable which type should or should not be more altruistic. It is tempting to say that
the disengaged type perhaps places the lowest weight on the child�s utility. But, the
disengaged parent can paradoxically claim that the child are left on its own as a result of
giving a high weight to the child�s utility. Namely, if the child is happy doing what it
chooses to do, then that is all that the parent should care, and the child does not need
neither reward nor penalty7 in the case of a disengaged parent.
On the other hand, the permissive parent can also claim that the child is given all the
weight in the parent�s logarithmic utility. After all the permissive parents might be
7 Technically, this line of a disengaged parents is summed up by saying that the parents completely succumb to the child�s incentives. See the discussion on the child�s incentive compatibility later in this section.
21
letting the child to maximize its own utility not the parent�s own utility. The
authoritarian types may be consistent with either an altruistic type or non altruistic type.
Thus, except possibly the authoritative type, the technical definition of altruism in terms
of the coefficient α does not in itself immediately help us to classify the parent types.
The specification of the parent�s logarithmic utility is meant to provide an intellectual
framework that describes how the economic environment of the parent-child relationship
does not fit the standard principal-agent model. Which type of parenting style is more
altruistic is an empirical question, which I hope to illustrate in my data analysis. .
Authoritative parenting, which balances clear, high parental demands with emotional
responsiveness and recognition of child autonomy, is one of the most consistent family
predictors of competence from early childhood through adolescence. Authoritarian
parents are highly demanding and directive, but not responsive. "They are obedience- and
status-oriented, and expect their orders to be obeyed without explanation" (Baumrind,
1991, p. 62). These parents provide well-ordered and structured environments with
clearly stated rules.8
8 To remind, authoritative parenting, which balances clear, high parental demands with emotional responsiveness and recognition of child autonomy, is one of the most consistent family predictors of competence from early childhood through adolescence. Authoritarian parents are highly demanding and directive, but not responsive. "They are obedience- and status-oriented, and expect their orders to be obeyed without explanation" (Baumrind, 1991, p. 62). These parents provide well-ordered and structured environments with clearly stated rules.
22
3.4 Constrained Nash Bargaining
The form of the altruistic parent�s utility function, α log U(• ) + (1 � α) log V(• ),
enables us to interpret it as a product of a classic Nash bargaining program9. If the parent
and the child bargain according to the Nash�s axioms, their bargaining takes place as if
they are maximizing the weighted product of their utilities10. If their relative bargaining
power is α and 1 � α respectively11, then the maximand of the bargaining process is
Uα(• ) V1- α(• ). Since the bargaining process is neutral to monotone transform of their
utility functions, it is functionally equivalent to maximizing its logarithmic form α log
U(• ) + (1 � α) log V(• ). The upshot is that the postulated altruistic form tacitly
incorporates a bargaining process with asymmetric bargaining power between the parent
and the child.
Although the model initially starts out with the parent as the principal and the child as
the agent, the outcome of the model, once the parent is assumed altruistic, turns into an
implicit bargaining model between the parent and the child. To the extent that a parent-
child relation is regarded as an implicit social contract, the altruistic parent behaves as if
the contract is bargained between the parent and the child.
The implied bargaining model is a constrained Nash bargaining model in the sense
that bargaining instruments must be chosen from restricted sets available to the parent
9 Nash (1953) 10 Strictly speaking, the Nash product that the bargaining process maximizes is not α log U(• ) + (1 � α) log V(• ) but α log [U(• ) � U*] + (1 � α) log [V(• ) � V*] where [U(• ) � U*] > 0 and [V(• ) � V*] > 0 follow from the nondegnerate bargaining set and each bargainer�s participation constraint. (U*, V*) is called a threat point of the bargaining problem; its role is same as each party�s reservation utility level. 11 This is an extension of the original Nash bargaining model in which α = ½ is assumed. For an axiomatic extension of the Nash�s original symmetric treatment, see Kalai (1977) and its n-person application, see Svejnar (1982).
23
and child including not only the standard participation constraints for the parent and the
child, but also child�s incentive compatibility constraint. In other words, the Nash
bargaining outcome reached in this paradigm is constrained efficient, subject to the
child�s incentive compatibility constraint.
3.5 Child�s Incentive Compatibility Constraint
Summing up the discussion in this chapter, my basic framework is to specify that the
parent and child behave as if they maximizes α log U(• ) + (1 � α) log V(• ) subject to the
parent�s participation constraint U(• ) ≥ U*, the child�s participation constraint, V(• ) ≥
V*, and the child�s incentive compatibility constraint, as well as any other technological,
institutional and legal constraints that each party may face. The critical element that
makes this model different from the textbook case of Nash bargaining is the presence of
the child�s incentive compatibility constraint. And, the child�s incentive compatibility
constraint is what makes the economic aspect of the present investigation a hybrid of the
standard bargaining framework and the standard principal-agent model.
What makes an incentive compatibility constraint important in the principal-agent
model is that the principal can observe the outcome of the agent�s action, but cannot
observe directly the agent�s action. This is because the outcome (y) is a joint
consequence of the state of nature (s) and the agent�s action (x), that is, there is a function
y = f(x, s). If the principal can observe z and s, the principal can invert the function and
infer x accurately. But, if the principal cannot ascertain s, then it is not possible to infer
the agent�s action x from the outcome y. The principal share the common knowledge
with the agent that a better outcome y is more likely if the agent implement a proper x
24
given s. To the extent that the state of nature occurs stochastically, the principal can
only make a statistical inference of the agent�s x given y. In other words, the agent has
a room to shirk in effort to implement a proper x and may still get by. Such
informational asymmetry regarding the agent�s action coupled with unobservable state of
nature occurs generally in a typical real-world contractual environment, and in particular
in the parent-child situation. It is thus relevant for me to comment what aspect of the
parent-child environment is observable or monitorable to the parent.
There are aspects of child actions that are not observable to parents, and thus the
parent cannot base their actions on child�s actions, but only on the observable outcome of
the child. For example, the parent can observe the child�s school grade, but cannot tell
what actions the child took to deserve that school grade. If it is a good grade, was it just
by a good luck even though the child was not putting much effort? If it is a bad grade, is
it because the child did not make effort or the child tried but exogenous events, such as
temporary headache or psychological infirmity might have happened beyond the child�s
control? In either case, the parent does not have an economical way to monitor the
child�s diligence12. In the case of smoking or drinking, the parent can find out whether
the child is smoking or not, but may have a difficult time to discover what caused the
child to smoke. Were there some socio-psychological causes including a pressure from a
wrong peer group? Parents perhaps will have difficulty knowing how hard the child
exerted effort to quit or in what way the child may be making effort to stay away from a
substance use. More generally, the parents know the child�s outcome, say an accident,
12 In this sense, the situation is completely analogous to the management-worker relationship in the agency theory of production. Management observe an output amount, but does not observe the worker�s effort input to produce that output amount, and the output is subject to random production factors.
25
but they cannot tell the extent to which the child exercises self-control and due diligence
to abstain from high-risk behaviours. In this dissertation, I use the term �behaviour� to
mean an observable outcome of the child behaviour, such as a school grade or substance
use. But, the child�s actions or effort remain unobservable to the parent.
3.6 Empirical Comparative Statics
In principle, the economic framework described above implies a comparative statics
on the coefficient α, to trace the effect of the parenting style on the child�s action, and
probabilistic child�s behaviour. But, the comparative static exercise is difficult without a
set of regularity conditions regarding the child�s incentive compatibility constraint and
also without making radically simplifying assumptions on the child�s and parent�s utility
functions13. There are enough richness and thus accompanying comparative static
complications, even if the social contract between the parent and child is enforceable.
With the presence of the child�s incentive compatibility constraint, additional regularity
conditions are needed to deal with the child�s action space. The standard way to treat the
incentive compatibility is to grope for a set of regularity conditions that enables the so-
called first-order approach14. Deriving the qualitative comparative statics of parenting
style is thus difficult, and is not an aim of this dissertation. Instead, by using the data that
enables me to identify parenting styles exogenously, I run regressions to see how
13 The standard economics of principal and agency postulates that the principal is risk neutral and the agent risk averse. But, in the case of the parent-child relationship, the child can often assume a more aggressive risk-taking posture than the parent. Or, the parents are often more conservative in risk taking than children. 14 See Jewitt (1988), Pavoni & Arphad (2004) and Rogerson (1985) for the standard as well as the latest treatment of the incentive compatibility constraints in the principal-agency model.
26
different parenting styles affect the likelihood of child�s substance use. Given the nature
of the data set, I examine the likelihood of the child to engage in the substance use as
well as the likelihood of the extent of the substance use. The former is same as asking
whether a child under a given parenting style smokes or not, for example, without asking
how often or how much smoking the child does. The latter, instead, can ask whether a
child already smoking tends to smoke more or less, under a different parenting style. The
central contribution of this dissertation is in the examination whether there is an
empirically discernible relationship between the parenting style and substance use. In
other words, I try to conduct an empirical qualitative response econometrics of the
parent-child relationship. I hope that such empirical findings would give an impetus to
further sharpen the model specifications that may yield a qualitative comparative statics
within the context of the constraint bargaining model described in this section.
27
CHAPTER 4: CLASSIFICATION SCHEME
4.1 Classification schemes
The aim of this chapter is to develop a classification scheme of parenting styles.
These kinds of classification schemes are used in social sciences where lists are used and
then given such �scientific labels�15 as classifications. A scientific classification first
defines the domain and then specifies mutually exclusive and exhaustive categories
within that domain. The validity of a classification is assessed by its utility through one or
more of the stated purposes. Thus in classifying we are attempting to mark the perimeter
of a scientific territory. Though such a classification is different from a taxonomy, which
is based on a deep theory that explains the domain of concern.
A classification system should essentially have some properties to be considered a
valid classification system: A classification system should be exhaustive where the same
domain should include all the possible categories that can be possibly encountered.
Additionally the classification system should have mutually exclusive dimensions and
this is especially relevant when the categories are very similar in nature. Thus there
should be no fuzzy categories or overlapping in dimensions. Next each time the
classification system is used it should give the same result if the same individual is being
classified. A classification system should not give different results or term the person
15 Values in action (VIA) Classifications of Strengths, Peterson and Seligman (draft, January 4, 2003), Values in Action Institute 2002.
28
differently on different occasions. Also a classification system should be a representation
of the �real world� that is, it should accurately capture the truth.
In the case of parenting styles there are twin domains of demandingness and
responsiveness. These two domains comprise of different categories which could
reasonably be classified under a common domain. In order to check whether these
categories are consistent as part of a common factor, a factor analysis is performed in
later chapters on the two dimensions of demandingness and responsiveness. In the case of
the demanding factor all those questions which are rated as high on the punitive scale are
classified under one domain. In the case of responsiveness those questions which can be
listed as higher on the scales of warmth are classified under a common category. These
questions are listed in the table below. The original classification used in the development
psychology literature given by Baumrind (1991) can be listed as follows with the given
categories and the standard classification. One of the drawbacks with the Baumrind
categories is that the data does not always support the cause and effect of these
relationships.
4.2 Parenting Styles (Baumrind Classification)
In the Baumrind classification a major goal of parenting is to encourage and foster a
high level of interdependence. There also needs to be a balance between independence
and connectedness which should be maintained by parents. There are two broad
dimensions of parenting: demandingness -- making maturity demands and being willing
to supervise and responsiveness -- showing sensitivity and support.
29
Demandingness
Responsiveness Low High
Low Disengaged Authoritarian
High Permissive Authoritative
Figure 4.1 Parenting Styles Dimensions
The parenting styles emerge along the two dimensions and they can also be viewed
along with the child outcomes for the different parenting styles. In the case of
Authoritative Parenting the parents are warm but firm, they set standards consistent with
needs and capabilities, place high value on development of autonomy, assume ultimate
responsibility for child�s behaviour and discuss and negotiate with the child. The children
are lively, happy, self-confident, less gender-typed, task persistent, have social maturity,
high self-esteem, internalized moral standards and superior academic achievement. One
explanation forwarded for this parenting style being effective is that the control that is
fair is more likely to be complied with, parents also provide models of caring concern and
confidence and their demands are reasonable.
In the case of Authoritarian Parenting the parents place the highest value on
obedience and conformity, favor punitive and absolute disciplinary measures, believe that
kids should accept without questioning, place importance on restricting autonomy and the
children are anxious, withdrawn, unhappy, react with hostility when frustrated in peer
30
situations and the boys have high rates of anger and defiance, the girls are dependent,
lack in exploration, retreat from challenge.
In the case of Permissive Parenting the parents are overly tolerant, are very accepting,
place high value on personal freedom for children, see themselves as resources that child
can choose to use and the children are less mature, have trouble controlling impulses, are
overly demanding and dependent on adults, have less persistence on tasks, conform to
peers.
In the case of Disengaged Parenting parents mimimize the time and energy devoted to
children, know very little about their children, are overly concerned with the stresses in
own lives don�t converse with their children, don�t consider their opinions and are
considered �parent-centered�. The children in this case have problems with attachment,
cognitive development, play and social and emotional skills and exhibit delinquent
behaviour, also experiment with drugs and alcohol.
The questionnaires used in the Children of NLSY-79 survey are used to elicit
responses from the children themselves through self-reported inventories on substance
use. The parental control questions are extracted from the questionnaires administered to
the parents since those are the questionnaires which collect information on the likelihood
and frequency of parental responses to various child behaviours.16 There are issues
concerning the reliability of these responses and response quality.17 Its also recognized
that this method of data collection is non-experimental in nature and this kind of a
research has different implications as given in Johnson (2001). 16 Attachment and Control in Family and Mentoring Contexts as Determinants of Adolescent Adjustment to College , Journal of Family Psychology, Vol. 14, Soucy and Larose. 17 Reliability of Responses in Questionnaire Research with Children, Borgers and Hox.
31
Additionally the survey is longitudinal in nature and the subtype is a panel study since
the same children are being studied successively at various points. There are issues
involved in establishing causality with reference to the dependent variable as we have to
go backwards in time to find the movement of the independent variables that could
explain current differences in the dependent variable.
In the case of the domains of demandingness and responsiveness these categories
given below have been extracted from three questions in the mother supplement
questionnaire. While one question is a parental response to child behaviour another
follow up question is a response to grades and finally there is a question which asks the
frequency of the actions taken in response.
Domain of Demandingness has several sub components such as Grounda which is
grounding the child, Spanka which is spanking the child, Allowa which is taking away
the child�s allowance, Momt which is Mom contacting the teacher about low grades,
Moml which is Mom limiting the child�s non-school activities, Momp which is Mom
punishing the child, Spankt which is the number of times spanked the child, groundt
which is number of times grounded the child, Allowt which is number of times took away
the child�s allowance.
Domain of Responsiveness has several sub �components such as Talka which is
Talking to the child, Momt which is Mom Talking to the child, Moms which is Mom
helping child with schoolwork, Mome which is Mom keeping a closer eye on child�s
activities, Momk which is Mom lecturing the child for low grades, Affect which is the
number of times Mom showed physical affection towards the child.
32
4.3 Psychological Models
There are different psychological models in the literature. The ecological model and
the risk-taking propensity model are reviewed briefly here but only the ecological model
will be tested empirically. In the dissertation Bronfrenbrenner�s (1979) Model of the
Ecology of Human Development is adapted to the parent-child association where the
child is the centre of the system and is surrounded by the Microsystem, which includes
parenting factors, and the neighbourhood factors, which constitute the Mesosystem. The
ecological paradigm began with Lewin�s Behaviour = f (Person, Environment) model
where humans are active and shape the environments in which they live. In the context of
adolescent psychological development there are individual factors such as the child�s
own propensity to consume substances and then the microsystem and then the
mesosystem and these are the factors influencing the child�s behaviour. Thus the
adolescent while growing up in the household has a core behaviour and then a peripheral
component which is constantly adapting to the environment.
Figure 4.2 Ecological Paradigm
Neighbourhood factors (Mesosystem)
Parenting factors (Microsystem)
Child factors (Individual)
33
This model of risk-taking behaviour is adapted from Irwin & Millstein (1986) and the
definition of risk-taking inherent in psychosocial development is that risk-taking is a
result of an interaction between the biopsychosocial processes of adolescence and the
environment. The development psychology literature does indicate that some risk-taking
is necessary in the natural developmental process, but extreme forms of risk-taking have
severe consequences. The underlying strand of thinking indicates that young children do
not have an adequate understanding of the long-term consequences of their actions and
therefore may take actions that are potentially destructive in nature.
34
Figure 4.3 Theoretical Model of Risk-Taking Propensity
SELF-PERCEPTION OF CHILD BEHAVIOURAL PROBLEM INDEX Self-Esteem Self-Reliance
PERCEPTIONS OF SOCIAL ENVIRONMENT
RISK-PERCEPTION
RISK-TAKING BEHAVIOUR
35
Figure 4.4: Schematic Representation (PARENT-CHILD MODEL)
PARENTAL STYLES AUTHORITATIVE
AUTHORITARIAN PERMISSIVE DISENGAGED
INDIVIDUAL CHILD FACTORS FAMILY FACTORS
CHILD
OUTCOMES:
ALCOHOL
CONSUMPTION
SMOKING
CHILD BEHAVIOURAL OUTCOMES
CHILD MEASURES BEHAVIOURAL PROBLEM INDEX
36
CHAPTER 5: EMPIRICAL MODEL
5.1 Introduction
The empirical model which is being used in this dissertation is a discrete choice
probit model. Such models are effective when qualitative choices are being made. The
goal of this research is to provide more precision and more definitiveness in
understanding the role of parents in the development of children.18 The precision is
increased by testing a conceptual framework that emphasizes the associations between
parenting dimensions and key domains of child functioning. The paths of this framework
were derived from the substantial empirical and theoretical literatures finding links
between parenting and child development. These literatures were interpreted to be
compelling enough to recommend moving towards greater precision in understanding the
associations, and to recommend the hypothesized paths of the model.
Principal component analysis involves a mathematical procedure that transforms a
number of correlated variables into a smaller number of uncorrelated variables called
principal components. The first principal component accounts for as much of the
remaining variability. The main aim of conducting a principal component analysis is to
discover or reduce the dimensionality of the dataset and to identify new meaningful
underlying variables.
18 Parental Support, Psychological Control and Behavioural Control: Assessing Relevance Across Time, Method and Culture, Barber, Stolz, Olsen and Maugham, Monographs Series of the Society for Research in Child Development.
37
Principal component analysis and factor analysis are two of the most common
methods used to investigate the correlation structure. PCA decomposes the correlation (or
variance-covariance) into eigenvalues and orthogonal eigenvectors which explain the
variation and interrelationship of the variables in the data. Factor analysis also explores
the correlation structure of the data.
The main theme of principal component analysis is to reduce the dimensionality of a
dataset that may contain many highly correlated variables, while retaining as many as
possible of the variations in the data set. This also makes the interpretation of the data
easier.
5.2 Factor Analysis
Two factors emerge in parenting style factor analysis. These factors are
demandingness and responsiveness. Demandingness refers to the demands made on the
child by the parent, the expectations for maturity in behaviour, the discipline and
supervision which is provided. Responsiveness refers to the encouragement of
individuality, self-assertion and regulation and being responsiveness to needs and
demands. 19
There are two kinds of factor analysis- exploratory and confirmatory factor analysis.
Exploratory factor analysis is used to explore data to determine the number or the nature
of factors that account for the covariation between variables when the researcher does not
have sufficient evidence to form a hypothesis about the number of factors underlying the
19 Journal of Early Adolescence 11(1) p 56-95 1991, Diana Baumrind.
38
data. Thus, exploratory factor analysis is generally thought of as more of a theory-
generating procedure as opposed to a theory-testing procedure.20
Factor analysis is also involved with questions of validity and in the process of
finding out whether the identified factors are correlated, exploratory factor analysis
answers the question of construct validity and if the scores on the test measure what the
test purports to measure.
The general objective of factor analytic research is to determine the nature of the
underlying factors and develop an understanding of their relationships to the surface
attributes. An important fundamental point about factor analytic research is that its
virtually impossible to achieve this objective in a single study. In the case of the
problems being addressed within much of the research in psychology one of the
preliminary steps involves identifying the domain and population of interest. The domain
is the phenomenon which we are interested and could be defined in very broad or very
narrow terms. In the present case there are twin domains of demandingness and
responsiveness.
Additionally there is the population of interest, and in this case it is the sample of 10-
14 year old children being studied.21 Once the domain and population of interest are
established, the researcher selects from the domain the variables which are to be
measured. These variables are referred to as �surface attributes�, where a surface attribute
is one of the many attributes of people that can be observed and measured. A set of
surface attributes measured in a given study is refered to as a battery of surface attributes.
20 Applied Multivariate statistics for the social sciences, Mahwah, J. Stevens. 21 Exploratory Factor Analysis by Ledyard Tucker and Robert Mac Callum, 1997.
39
Factor analysis provides a set of techniques designed to identify order and structure in
data by providing a meaningful explanation for the observed variation and covariation in
surface attributes. There are also internal attributes where an internal attribute is an
unobservable characteristic of people on which people differ in extent or degree. Internal
attributes cannot be measured directly but the effects are reflected when one obtains
measures on surface attributes and they are referred to as factor or latent variables. We
can further distinguish between two types of internal attributes, one type is a common
factor which is defined as an internal attribute which affects more than one of the surface
attributes in the selected set or battery. The second type of internal attributes is specific
factors, each of which influence only one of the surface attributes in any given battery.
The figure below depicts the relationship between surface attributes and three types of
factors defined above. The figure shows four surface attributes on the right. These are the
various variables being combined under the two domains. On the left are three types of
factors, the two internal attributes are represented as influencing more than one of the
surface attributes.
There are also two common factors which are intercorrelated. Then there are four
specific factors which also represent internal attributes. Each specific factor influences
only a single surface attribute. Then at the end are the measurement factors, arising from
unsystematic events.
40
(Internal attributes) Surface Attributes
Figure 5.1: Schematic representation of common factor theory
41
5.3 Estimated Probit Model
We estimate the following reduced form specification as a probit model for the different
forms of substance use by the child. The likelihood of participating in different forms of
substance use is unobservable to the parent and Pij* is the unobservable variable which is
the child�s latent propensity to smoke (consume alcohol). Therefore we define the
observed variable Pij as:
Pij=1 if Pij*>0
Pij=0 otherwise
Prob (Pij=1) = 1-Φ (-αStylei + Xi′β)
where Φ represents the cumulative normal distribution. X is a vector of: explanatory
variables that includes all child-specific characteristics, mother-specific characteristics
and family background variables. They include mother�s substance use over the lifetime
as well as mother�s substance use during pregnancy.
5.3.1 MODEL 1: (With Authoritative, Authoritarian, Permissive Parenting Styles as RHS
variables)
We assume there is a four-style world and the styles are mutually exclusive and
exhaustive and they are exogenously determined. Therefore we construct a Dummy for
Authrve where Authrve=1 in the case of Authoritative parenting style and Authrve=0 in
the case of the other parenting styles. And similarly in the case of Authoritarian,
Permissive and Disengaged parenting style. When estimating we run the regression by
excluding one Dummy variable at one time.
42
Prob (Pij=1) = 1-Φ (-αStylei + Xi′β)
where Style i =Authve, Authran, Permve
The HOME questions in the Mother Supplement questionnaire do not directly ask
the mother what action she would take if the child engages in substance abuse. The
question reads as follows �Sometimes children get so angry at their parents that they say
things like �I hate you� or swear in a temper tantrum. Please check which action(s) you
would take if this happened�
Therefore the question in the Mother Supplement only asks what the mother would
do if the child misbehaves or throws a tantrum. There is also a need to control for other
behavioural problems as those could be potential sources of endogeneity so we control
for those problems using the behavioural problem index. The behavioural problem
index22 is based on responses from the mother to 28 questions in the Mother Supplement
which deal with specific behaviours that children age four and over may exhibit in the
previous three months. The standard score used in this paper sums across the subscores
created according to the following domains:(1) antisocial behaviour, (2)
anxiousness/depression,(3) headstrongness, (4) hyperactivity (5) immature, (6)
dependency and (7) peer conflict/social withdrawal. The standard score of BPI is scaled
from 70 to 140 and higher scores represent a greater level of behavioural problems.
This dissertation uses the measure of lifetime substance use i.e. if the child smoked or
drank alcohol in his entire life prior to the time of the interview.
22 Details of the behavioural problem index are given in the NLSY79 Child Handbook.
43
5.3.2 MODEL II: (With demandingness and responsiveness as RHS variables)
Prob (Pij=1) = 1-Φ (-αDemandgi + Responsvi + Xi′β)
An alternative model specification is to assume there is a two-factor world and the
factors are also mutually exclusive and exhaustive and they are exogenously determined.
Therefore we construct a Dummy for Demandingness where Demandingness=1 and
Demandingness=0 in the case of the other parenting styles.
44
CHAPTER 6: DATA DESCRIPTION AND RESULTS
6.1 Sample Construction
The National Longitudinal Surveys of Youth 1979 (NLSY79) began as a sample of
12,686 individuals born between 1957 and 1964. There are three subsamples, a cross
section sample representative of the population of this birth cohort living in the US as of
the beginning of 1979, oversamples of blacks, Hispanics and economically disadvantaged
whites and a military sample. The NLSY79 surveys use complex sampling i.e. multi-
level clustered sample design.23 Beginning in 1986 information was collected biennially
on the children of the female respondents. The interviews for the NLSY79 surveys have
been conducted annually between 1979 and 1994 and biennially after that. The combined
mother-child (NLSYMC) represents a unique dataset with panel data on both mothers
and their children. The data for this dissertation uses the 1986-1998 waves of the
NLSYMC. The poor white oversample was dropped after 1990 and the military sample
was dropped in 1984. So the data used for this dissertation include the cross section and
the Black and Hispanic oversamples. Since the women in the original sample are close to
the end of their childbearing years, the children in this sample are representative of the
children ever born to women in the 1957 and 1964 who were living in the US at the end
23 The implications of the cluster sample for estimation of race effects is discussed by Eberweing, Olsen and Reagan (2003).
45
of 1978. The sample is representative of each race, ethnicity group, but contains
oversamples of blacks and Hispanic minorities. Since 1986 assessments have been
administered to the children of NLSY79 mothers to measure cognitive ability,
temperament, motor and social development, behaviour problems, perceived competence,
and the quality of their home environment. Therefore to construct the dataset for this
study the data from the mother�s records from the NLSY79 were combined with the child
records in the Child survey conducted since 1986. The combined data helps in exploring
the interrelationships between family and environmental factors, maternal behaviour, and
child and young adult development.
The base sample consisted of the children aged 10-14 who were interviewed between
1986-98. In 1986 there were 294 children, there were 851 children interviewed in 1988,
1,158 interviewed in 1990, 1,700 interviewed in 1992, 1,955 interviewed in 1994, 1,951
interviewed in 1996 and 1,996 interviewed in 1998. There were 10,007 observations on
all the children interviewed in the overall sample (including multiple observations for
each child), of these there were 3499 observations for Black children and 2250
observations were Hispanic and 4258 White. The sample selection criterion are described
in detail in Table 6.1 and in the first selection criterion only those children were selected
aged 10-14 who completed the self-administered questionnaire. This questionnaire was
given to all children aged 10-14 as of the December of the interview years 1994-98 so
this sample deletion deleted all those children who did not complete the questionnaire.
The sample was further narrowed to include those children whose mothers were also
administered the Mother supplement which includes HOME questions and then
subsequent sample deletions to get the final sample for smoking and for alcohol.
46
Numbers of observations and Reasons for deletion from sample
N
Black
Hispanic
Non Black
Non Hispanic
Number of observations on Children Respondents of NLSY79 aged 10-14 interviewed in 1998
10007 3499 2250
4258
After deletion as children did not answer SA questions on substance use
7553 2713 1698 3142
After deletion as children�s residence is not with mother
7441 2640 1662 3139
After deletion as child did not answer SA questions on getting an allowance
4030 1367 930 1733
After deletion as mother reactions to the child questions are not missing
3399 1105 757 1537
After deletion as explanatory variables missing like race, highest grade completed and family income
2237
691 478 1068
After deletion as child did not answer questions on smoking (Smoking sample)
1049 672 466 1053
After deletion as child did not answer questions on alcohol (Alcohol sample)
1069 682 467 1054
Table 6.1 Sample deletion for different groups
47
Descriptive Statistics
Variable
Variable definitions
SMOKE Smoke=1 if child ever smoked in lifetime
ALCOHOL Alcohol=1 if child consumed alcohol in lifetime
AUTHRVE Authrve=1 if the parenting style is Authoritative
AUTHRAN Authran=1 if the parenting style is Authoritarian
DISENGG Disengg=1 if the parenting style is Disengaged
PERMV Permv=1 if the parenting style is Permissive
ALL All=1 if child gets an allowance
CWORK Cwork=1 if child works for pay
CHILDAGE Age of child at time of interview (in years)
BLACK Black=if child is Black
HISPANIC Hispanic=1 if child is Hispanic
MALE Male=1 if child is Male
MOMSMK Momsmk=1 if mother smoked atleast 100 cigarettes in lifetime
MOMSP Momsp=1 if mother smoked during pregnancy
BPI Total standard score scaled from 70-145
Table 6.2 Variable Definitions
48
Variable
Means
SMOKE 0.1296473 (0.3360752)
ALCOHOL 0.1169317 (0.3214893)
AUTHORITATIVE 0.1693171 (0.3752073)
AUTHORITARIAN 0.0608045 (0.2390832)
DISENGAGED 0.4883068 (0.500972)
PERMISSIVE 0.2815716 (0.449976)
ALLOWANCE 0.7970065 (0.4024159)
CWORK 0.4097287 (0.4920138)
CHILDAGE 12.06564 (1.178243)
BLACK 0.2843779 (0.4513288)
HISPANIC 0.202058 (0.4017232)
MALE 0.4976614 (0.5002286)
MOMSMK 0.4480823 (0.49753)
MOMSP 0.2815716 (0.44976)
BPI 107.7109 (13.9016)
# of total observations for smoking
1049
# of total observations for alcohol
1069
Table 6.3 Descriptive Statistics
49
Variables BPI Cwork Childage Allowance
White 107.9737 (13.5534) [-0.1941]
0.4462659 (0.4975576)
[-0.0609]
12.01214 (1.196226) [-1.2161]
0.7751634 (0.355716) [-2.8245]*
Black 107.1713 (13.06311) [0.5374]
0.3148148 (0.4655209)
[3.1875]*
12.10378 (1.147461) [-0.5324]
0.8010757 (0.3988201)
[-1.2161]
Variables Authoritative Authoritarian Disengaged Permissive
Hispanic 0.2105 (0.4083546)
[-1.8377]
0.0455373 (13.5534) [-2.9930]*
0.4735883 (0.4997573)
[-1.7036]
0.3205829 (0.3713161) [5.4348]*
Black 0.1342593 (0.3417226)
[0.8998]
0.0509259 (0.2203572)
[-0.3164]
0.4675926 (0.5001077)
[0.1493]
0.3472222 (0.477198) [-0.7057]
* significance at 5% level
Table 6.4 T-test for difference in means across different racial groups
6.2 NLSY-79 Dataset description and properties
The Home Observation Measurement of the Environment � Short Form (HOME-SF) is
the primary measure of the quality of a child�s home environment included in the
NLSY79 Child survey. The HOME-SF is divided into four parts and the fourth part (Part
D) is for children 10 and over. The HOME-SF is a modification of the HOME inventory,
a unique observational measure of the quality of the cognitive stimulation and emotional
support provided by a child�s family. The HOME-SF is about half as long as the HOME
inventory. The Child Supplement and the Mother Supplement which are administered to
the children are part of the HOME administered questionnaires. In the Child Supplement
there are questions answered by the child on smoking and alcohol consumption.
50
In the Mother Supplement there are questions answered by the mother on the kind of
responses which the mother would take in response to different actions of the child.
The measure used to sum across the different child responses and interpret behaviour
is the Behavioural Problem Index. The Behaviour Problem Index was created to measure
the frequency, range, and type of childhood behaviour problems for children ages 4 and
over, the items being derived from child behaviour scales. The Behaviour Problem total
score is based on responses from the mother to 28 questions in the Mother Supplement.
These mother-report questions ask about specific behaviours that children age four and
above may have exhibited in the previous three months. For the overall Behaviour
Problems scale and the set of six subscales, responses to the individual items are
dichotomized and summed to produce and index for each child. Higher scores represent a
greater level of behaviour problems. Factor analysis was used to determine the six
subscores alluded to above according to the following domains: (1) antisocial behaviour,
(2) anxiousness/depression, (3) headstrongness, (4) hyperactivity, (5) immaturity, (6)
dependency, (7) peer conflict/social withdrawal.
There are two subscales which measure the internalizing or externalizing behaviour.
All of the above scores and subscores are available for all age-eligible children who were
assessed biennially between 1986 and 2000. For all the above except the last three (the
non-dichotomous external, internal and total scores which were not recoded), overall as
well as �same-gender� normed scores created based on the data from the 1981 National
Health Interview Survey.
The average behavioural problem index for the children is 107. This figure is close to
the mean for the NLSYChild sample of 108 for children aged 10 and over but higher than
51
the national mean of 100.24 The mother of 98.7% of the population is employed and
60.3% are married with a spouse present in the family. Among the different substances
consumed by mothers alcohol is the highest at 75.5% with smoking at 44.9%. The
corresponding national estimates are 18.1% for alcohol and 62.2% for females smoking
in their lifetime. Among the substances consumed by mothers at pregnancy alcohol has
the highest figure.
This dissertation only uses smoke and alcohol as the dependent variables due to the
small samples in the case of marijuana and cocaine. The summary statistics given in
Table 6.2 indicate that smoking has the highest percentage of children at 13% followed
by alcohol at 11%. These figures are lower than national estimates for children between
12-17 but the sample size of the national survey is much larger at 25,357 children. But
the explanation for this is that the children in the sample here are younger. The statistics
from the National Institute of Drug Abuse indicate 37.1% of children smoked cigarettes
in their lifetime. For alcohol the figures were 42.9%. The percentage of children who are
working in the sample is about 40%.
When these estimates are compared with those for pregnant women between ages 15-
44 in the NHSDA (National Household Survey on Drug Abuse) survey they are
consistent as those estimates indicate 17.6% of pregnant women smoke cigarettes
compared to our estimate of 28%.
For alcohol their figures indicate 13.8% compared to our estimate of 45%. The racial
composition indicates that roughly 29% of the population is Black and 21% is Hispanic.
24 The national mean for BPI is available from the National Health Interview Survey 1981.
52
While there is expected to be significant under-reporting in substance abuse by
young children the advantage of examining the age group from 10-14 is that parental
control may be thought to have a greater impact on this age group than in the case of
older children. Also the relative importance of peer pressure may be lower in this age
group but these hypotheses need to be tested in future analyses.
The Parenting Style classification uses the following 3 questions from the HOME (D)
section in the Mother Supplement Questionnaire of the NLSY-79 Mother-Child dataset
for 10-14 year old children. The first question is �Sometimes children get so angry at
their parents that they say things like �I hate you� or swear in a temper tantrum please
check which action(s) you would take if this happened�. The possible responses are
ground, spank, talk with child, give him or her household chore, ignore it, send to room
for more than one hour, take away his/her allowance, take away TV, phone or other
privileges, other, put child in a short �time out�. The second question is worded as
follows �If your child brought home a report card with grades lower than expected, how
likely would you be to do any of the following�. The responses are �contact his or her
teacher or principal?, �lecture the child?�, �keep a closer eye on child�s activities?�,
�punish the child?�, �talk with the child?, �wait and see if child improves on his/her
own?�, � tell child to spend more time on schoolwork?�, �spend more time helping child
with schoolwork?�, �limit or reduce child�s non-school activities (play,sports,clubs,etc)�.
The responses are scaled on a scale of �very unlikely� to �highly likely�. The third
question is worded as follows �Sometimes kids mind pretty well and sometimes they
don�t. Sometimes they do things that make you feel good. How many times in the past
week have you��The responses are �had to spank your child?�, �grounded him/her?�,
53
�taken away TV or other privileges?�, �praised child for doing something worthwhile?�,
�taken away his/her allowance?�, �shown child physical affection (kiss, hug, stroke hair,
etc)?�, �sent child to his/her room?�, �told another adult (spouse, friend, co-worker,
visitor, relative) something positive about the child?�. In the case of different parenting
styles the means indicate that Disengaged parents are the highest in the sample at 49%
and Authoritarian least at 7%. Disengaged parents are neither demanding nor responsive,
Authoritarian parents are highly demanding but not responsive. There are approximately
61% of parents who change their parenting styles for the same child across different
years. The estimates from Table 6.3 suggest that 14% of parents change their parenting
styles across different children in the same year. From the available estimates we can
draw the conclusion that the hypothesis of switching can be rejected strongly. On
examining the matrix more closely we can see that if we examine the diagonal elements
we can see that out of those parents who remained the same there were no Authoritative
parents who remained the same. Only 1% Authoritarian parent remained the same while
11% Permissive parents remained the same. There were 28% of Disengaged parents who
remained the same.
This dataset offers an opportunity to ask this question about parenting styles changing
across time and across children due to the longitudinal nature of the dataset and also due
to the multiple observations for each child.
54
To From
Authoritative Authoritarian Permissive Disengaged
Total
Authoritative
0 0 .133 .867 100
Authoritarian
0
.077 .077 84.6 100
Permissive
0 0 .458 .542 100
Disengaged 0 .030 .121 .848
100
Total
0 .02 .20 .78 100
Table 6.5 Proportion switching parenting styles across different years for the same child
To From
Authoritative Authoritarian Permissive Disengaged
Total
Authoritative .857 0 .143 0 100
Authoritarian .143
.714 0 .143 100
Permissive
.077 0 .808 .115 100
Disengaged
.019 .019 .038 .925 100
Total
.16 .06 .25 .53 100
Table 6.6: Switching parenting styles across different children in the same year
6.3 Description of variables and Coding scheme
The variables which are used in the empirical analysis subsequently can be described
here and the scale is described or characterized constituting how the different items can
be reasonably clustered under one common scale. The factor analysis was done using two
55
different scales. The first parenting style inventory was constructed with the two
subscales of demandingness and responsiveness and these are constructed from the
seminal Baumrind (1989) framework. There is an extension of this is in the Maccoby and
Martin (1983) framework where there are three scales i.e. warmth, restrict and ignored.
Warmth is analogous to responsiveness and restrict is analogous to being demanding
while ignored falls into a separate category which is being uninvolved.
The scaling was done to generate the different cronbach alpha values to check for
internal consistency. Scaling using these items gave a cronbach alpha of 0.4605 for
demandingness with 9 items included under this scale and 0.4277 for responsiveness with
9 items included under this scale. While the scale alphas could be as high as 0.80
potentially there isn�t a �thumb rule� as such about the threshold value for alpha which is
acceptable. That is partly a matter of judgment and partly a matter of subjective choice
about what is the level of internal consistency needed among the items. There is also a
possibility that items that have restricted variability or items which are skewed may fail
to correlate well with other items, even if they are measuring the same underlying
concept. Thus I did the scaling repeatedly to check or even delete those items which
maybe bringing down the value of the alpha and the scale might be better off without one
or more of these potentially troublesome items.
Then checking the inter-item correlations to find the highest correlations it was found
that using a cutoff of 0.5 there were a selected set of variables where the correlation was
found to be high. Using this threshold value there were 3 new scales which could be
constructed with 2 items under each scale. The number of items under each scale is not a
forced classification, its a chance that there are an equal number of items under each
56
scale. Then I scaled again using the different items under three new scales of warmth,
restrict and ignored. Under warmth scaling using the two items included gave a cronbach
alpha of 0.46 and for restrict a cronbach alpha of 0.56.
Scaling involves the assignment of objects to numbers according to a rule. Scaling is
different from a response scale, where scales assign numbers according to a common
rule. Scaling is used in this context to see how well these questions �hang together� and
in some instances to score all the responses to generate a single number that represents
the overall construct. Thus a scale refers to a set of items and each item on a scale has a
scale value. There are there major types of uni-dimensional scaling methods. Thus they
differ considerably in how they arrive at scale values for different items. Scale
construction involves the creation of empirical measures for theoretical constructs and
these measures mostly consist of several items. The process of measurement involves the
assignment of numbers to empirical realizations of the variables of interest. In Thurstone
and Guttman scales, each item represents degrees of the variables of interest, such as the
difficulty of an item. In Likert and Semantic Differential Scales, each item represents
different degrees of the variables of interest. The differences between the scales affect the
computation of reliability. Thus in the construction of scales many items have been used
to develop the scales. 25The concepts of directiveness have been used in the construction
of scales.
25 An improved Directiveness Scale, Australian Journal of Psychology, J. J. Ray and F.H.Lovejoy.
57
6.4 Scale characteristics and cronbach alpha results
The scaling was used as a method to recognize the patterns of inter-item correlations
which exist among different items within the variables which could all be potentially
used under one scale.
There are theoretical justifications which exist about the scale construction. It is infact
true that two measures of reliability can be used as dependent variables.
Table 6.5 below gives the different parenting style inventories and the items which
are included under each scale. The Cronbach alpha for these items in the Responsiveness
Subscale is 0.46.
Responsiveness subscale for parenting style inventory-1
Item definitions
Talk=1
=1 if Talk with the child; 0 otherwise
Momt=1
Mom talk with the child
Moms=1 Mom help child with the schoolwork
Momg=1 Mom talk with child for lower grades
Mome=1 Mom keep a closer eye on child�s activities
Moml=1 Mom lecture child for lower grades
Table 6.7 Responsiveness subscale for the parenting style inventory
In the case of the demandingness subscale there are few items which are classified
under the scale. These items have been constructed from 3 different questions in the
Mother Supplement. Some of the items are responses to the first question which is with
regard to the response of the Mother to the child�s behaviour. The other items have been
58
constructed from the next question which is a Likert scale about the responses of the
mother to the performance of the child in school. The third question is a frequency
question about the number of times the parent �actually� took one of the actions which
they said they would be likely to take in the first question. The first item is allow, which
is taking away the child�s allowance. The next scale is spank, which is spanking the child.
Another variable is ground, which is grounding the child. Another variable is Momp,
which is the Mother punishing the child. Another variable is Momt which is Mother
contacting the teacher about low grades. The next variable is Moml, which is Mother
limiting the child�s non-school activities. Another variable is Allowt, which is the number
of times they took away the child�s allowance. Finally, the variable is Spankt which is the
number of times the child was spanked.
Demandingness subscale for parenting style inventory-1
Item definitions
Allow Take away child�s allowance
Spank Spank the child Ground Ground the child Momp Mom punish the child Momt Mom contact teacher about low grades Moml Mom limit the child�s non-school activities Allowt
Number of times took away the child�s allowance
Spankt
Number of times spanked the child
Table 6.8 Demandingness subscale for parenting style inventory
6.5 Reliability and Consistency
It is noted that two measures of reliability were used as dependent variables (Borgers,
1997) and the first indicator is the reliability measured at the scale level, for which
59
Cronbach alpha was selected. The second indicator is measured at item level, being the
item-test correlation. Cronbach alpha as well as the item-test correlation cannot be
defined for individual children; they are only defined for groups of children. In order to
compare the reliability between several age categories, years of education categories,
groups of children had to be constructed. However the groups could not be constructed
by combining all three child characteristics. The result would be too many groups with
too few children in the different groups. Therefore, three separate groupings were
constructed were constructed, based on different parental response categories. For each
group of children the Cronbach alpha was computed for every scale in the questionnaire
they answered. Similarly, for every group of children item-rest correlations were
computed for each question.
This brings us to questionnaire characteristics and question design to elicit responses
from children. There could be positive effects such as the position of questions in the
questionnaire, and the existence of complex constructions in the question. Questions that
ask for a numeric quantity could have a negative effect. Also questions that invoke some
opinions have a positive effect on Cronbach�s alpha but a negative effect on item-rest
correlations. Questions that ask for attributes and capacities have a positive effect on
item-rest correlations. The number of words in the introductory text have a positive effect
and the more words used in the introductory text the more reliable the responses will be.
That is also true for the comprehensive readability index which has a positive effect on
both measures, the same as questions with a given reference period in the questionnaire
and sensitive questions.
60
There are overall issues concerning the reliability of responses and both child
characteristics and question characteristics affect the reliability of responses of children
in self-administered questionnaire research such as in the NLSY Mother-Child
Supplements. These effects of the child characteristics could work in different directions.
Younger children produce less reliable responses than older children and girls give more
consistent responses than boys. This is in support of the hypothesis that reliability
increases with the cognitive level and data quality increases with cognitive level (Borgers
& Hox, 1999). Younger children are sensitive to social desirability and young children
would be reluctant to express their own thoughts and feelings being afraid to suggest
something wrong (Maccoby & Maccoby, 1954).
Also ambiguous response scales could decrease the reliability of response and
ambiguous words in questions should be avoided as far as possible and it should be
recognized that the information about perspectives, attitudes and behaviour of children
should be collected from the children themselves.
61
CHAPTER 7: EMPIRICAL RESULTS
7.1 Results
The empirical model which is being used in this dissertation is a discrete choice
probit model. Such models are effective when qualitative choices are being made. The
goal of this research is to provide more precision and more definitiveness in
understanding the role of parents in the development of children.26 The precision is
increased by testing a conceptual framework that emphasizes the associations between
parenting dimensions and key domains of child functioning. The paths of this framework
were derived from the substantial empirical and theoretical literatures finding links
between parenting and child development. These literatures were interpreted to be
compelling enough to recommend moving towards greater precision in understanding the
associations, and to recommend the hypothesized paths of the model across time.
Behavioural Problem Index
The results are given here for all the different types of regression analysis. The
empirical specification was tested on the data and the following results hold for the linear
regression model. This model was selected in the case of the behavioural problem index.
The results showed that parenting style is significant as an explanatory variable in
determining child behaviour controlling for random effects [see Table 7.1].
26 Parental Support, Psychological Control and Behavioural Control: Assessing Relevance Across Time, Method and Culture, Barber, Stolz, Olsen and Maugham, Monographs Series of the Society for Research in Child Development.
62
The coefficient on authoritative parenting style is 5.78 and the coefficient on
authoritarian parenting style is 4.44 and similarly the coefficient on disengaged parenting
style is 3.39 and all the parenting styles come out to be highly significant.
Random Effects OLS Analysis of BPI Dependent Variable: BPI
Variable Coefficient SE-error CONSTANT 99.20 4.10 AUTHORITATIVE 5.78* 1.24 AUTHORITARIAN 4 0.44* 1.80 DISENGAGED 3.39* 0.96 ALLOWANCE -0.35 0.99 CWORK 0.43 0.84 CHILDAGE 0.21 0.32 BLACK 0.26 1.05 HISPANIC 0.04 1.16 MALE 2.89* 1.12 MOMSMK 2.86* 0.88 MOMSP 0.71 1.23 # of observations 1069 # of children 954
Table 7.1 Random effects OLS model for BPI
The regression reported in the first column is estimated by OLS with standard errors
which are robust to the presence of heteroskedasticity. The reported estimate omits the
lagged endogenous variable since inclusion severely reduces the number of observations.
Smoking
The outcomes include drug and alcohol outcomes and behavioural outcomes. In this
research since substance use outcomes were measured repeatedly across a panel i.e. a
series of repeat observations on the same person, the goal is often to examine the effects
63
of different treatments and/or predictors on usage levels can be aggregated to provide a
single outcome per subject, for example an average substance use. In these cases,
standard statistical procedures can be applied. However, these approaches are limited
because they ignore changes across time or they only consider within subject change that
is linear. Finally, from a statistical point of view these approaches are inefficient. The
development of more general statistical methods for longitudinal data analysis has been
an active area of statistical research. There are several features that make Random-Effects
Regression Models especially useful in longitudinal research.
The results are given here for all the different types of regression analysis. The
empirical specification was tested on the data and the following results hold for the linear
regression model.
Random Effects Probit analysis of Smoking Dependent Variable: Smoking
Table 7.2 Random effects Probit model for Smoking
Variable Coefficient SE-error CONSTANT -9.74 3.68 AUTHORITATIVE 0.31 0.33 AUTHORITARIAN 0.44 0.52 DISENGAGED 0.19 0.24 ALLOWANCE 0.11 0.25 CWORK 0.58 0.28 CHILDAGE 0.56 0.21 BLACK -1.13 0.49 HISPANIC -0.23 0.27 MALE 0.08 0.20 MOMSMK 0.69 0.36 MOMSP 0.27 0.28 # of observations 1049 # of children 940
64
In order to examine the relationship between parenting style and other child outcomes
such as smoking and alcohol consumption an alternative model i.e. a discrete choice
probability model was selected. Additionally, in the case of smoking and alcohol
consumption as dependent variables since there are repeat observations on the same
individual I have used a random effects probit model. This model takes account of child-
specific effects. Such a specification is typical of panel data equations. For the main
results from the random effects probit analysis [see Tables 7.2].
Random Effects Tobit Model for Smoking Dependent Variable: Csmok Variable Coefficient
SE-error CONSTANT .0.46 2.03 AUTHORITATIVE 0.57* 0.28 AUTHORITARIAN 0.55* 0.53 DISENGAGED -0.43 0.28 CWORK 0.60* 0.15 CHILDAGE -0.002 0.009 BLACK -0.42 0.30 HISPANIC 0.29 0.31 MOMSMK -0.03 0.41 MOMSP 0.12 0.42 # of observations 1049 # of children 940
Table 7.3 Tobit model for Smoking intensity
65
Random Effects Probit Analysis of Smoking consumption (With interactions of parenting style and race) Dependent Variable: Smoking Variable Coefficient SE-error CONSTANT -9.90 3.69 AUTHORITATIVE -0.19 0.58 AUTHORITATIVE*BLACK -11.62 804949.8 AUTHORITATIVE*HISPANIC 0.79 0.96 AUTHORITARIAN 0.26 0.41 AUTHORITARIAN*BLACK 0.51 1.09 DISENGAGED -0.005 0.29 DISENGAGED*BLACK 1.35 1.05 DISENGAGED*HISPANIC 0.23 0.57 ALLOWANCE 0.13 0.26 CWORK 0.60 0.29 CHILDAGE 0.57 0.22 BLACK -2.07 1.13 HISPANIC -0.41 0.47 MALE 0.09 0.21 MOMSMK 0.29 0.37 MOMSP 0.29 0.29 # of observations 1049 # of children 940
Table 7.4 Random Effects Probit Analysis of Smoking
Alcohol Consumption
As we see from Table 7.5 below, the authoritative parenting style does not affect
probability, nor does allowance. The coefficient on authoritative style is 0.02. No other
factors were significant.
66
Table 7.5 Random Effects Probit Analysis of Alcohol
Random Effects Tobit Model for Alcohol Dependent Variable: Cdrnk Variable Coefficient SE-error CONSTANT 232.4 79.44 AUTHORITATIVE -9.98 9.85 AUTHORITARIAN 18.29 26.27 DISENGAGED -21.66 16.81 CWORK -19.40* 5.61 CHILDAGE -.31 .26 BLACK -16.90 11.45 HISPANIC -4.67 15.91 MOMSMK -22.65 19.46 MOMSP 16.44 20.55 # of observations 1069 # of children 954
Table 7.6 Tobit model for alcohol intensity
Random Effects Probit Analysis of Alcohol Consumption Dependent Variable: Alcohol
Variable Coefficient SE-error CONSTANT -4.46 0.60 AUTHORITATIVE 0.02 0.17 AUTHORITARIAN 0.18 0.23 DISENGAGED 0.17 0.12 ALLOWANCE -0.05 0.13 CWORK 0.44 0.10 CHILDAGE 0.23 0.04 BLACK -0.06 0.12 HISPANIC 0.26 0.13 MALE -0.05 0.10 #of observations 1069 # of children 954
67
Referring to the Table 7.6, please note that in the case of the tobit model for the
number of drinks consumed the results are not highly significant in the case of child-
specific and mother-specific effects.
Random Effects Probit Analysis of Alcohol consumption (With Interactions of parenting style and race) Dependent Variable: Alcohol Variable Coefficient SE-error CONSTANT -4.45 0.61 AUTHORITATIVE 0.20 0.28 AUTHORITATIVE*BLACK -0.08 0.51 AUTHORITARIAN -0.006 0.43 AUTHORITARIAN*BLACK 0.18 0.16 AUTHORITARIAN*HISPANIC -0.05 0.34 DISENGAGED -0.007 0.28 DISENGAGED*BLACK 0.48 0.10 DISENGAGED*HISPANIC 0.23 0.04 ALLOWANCE -0.01 0.29 CWORK 0.20 0.22 CHILDAGE -0.05 01.10 BLACK -0.05 0.10 HISPANIC 0.11 0.13 MALE -0.01 0.14 MOMSMK 0.11 0.13 MOMSP -0.01 0.61 # of observations 1069 # of children 954
Table 7.7 Random Effects Probit Analysis of Alcohol
The Table 7.7 shows the random effects probit models with interactions of
parenting style and race. In the case of the number of cigarettes smoked and number of
alcoholic drinks consumed a tobit model was selected which accounts for the censoring
values which arise in these responses due the sample selection which has already taken
place in the previous question which asked if the respondents consume substances or not.
68
The respondents are only taken to the follow-up question if they have answered �yes� to
the previous question. Thus the results for the tobit model show that when the number of
cigarettes are regressed on the explanatory variables, in the case of the child-specific
effects the results come out to be significant. In the case of mother-specific effects the
results are also significant.
7.2 Correlations
Recall that it was expected that smoking and behavioural problem index will be
positively related. To analyze the data, a Pearson Product Moment correlation was
calculated. Results revealed that analyzing the data for girls and boys separately for girls
the correlation coefficient is 0.15 for smoking. For the full sample the correlation
between smoking and the behavioural problem index is higher than for alcohol and the
behavioural problem index [see Table 7.12]. For girls the correlation coefficient is 0.15
for smoking while for boys the corresponding figure is 0.14. In the case of alcohol, the
figure is 0.11 for girls and 0.06 for boys. The figures for correlations between substance
use and the externalizing behavioural problem index and reveals that for the full sample
again smoking has a higher correlation than alcohol [see Table 7.13]. In the case of girls,
the correlation is 0.14 for smoking and 0.16 for boys. In the case of alcohol, the figures
are 0.09 for girls and 0.05 for boys the consumption of alcohol and are significant.
69
Correlations between substance use and BPI
BPI�Substance
use
Sample Size
(Full)
SampleSize
(Boys)
Sample Size
(Girls)
Pearson Correlation coefficient
(Full)
Pearson Correlation coefficient
(Boys)
Pearson Correlation coefficient
(Girls)
BPI-Smoke 1615 804 811 0.14524 0.14386 0.15056
BPI-Alcohol 1611 807 811 0.09073 0.06553 0.11936
Table 7.8 Correlations between substance use and BPI
Correlations between substance use and BPI-Externalizing
BPIe�Substance use
Sample Size
(Full)
Sample Size (Boys)
Sample Size
(Girls)
Pearson Correlation coefficient (Full)
Pearson Correlation coefficient (Boys)
Pearson Correlation coefficient (Girls)
BPIe-Smoke
1198 598 600 0.15250 0.16334 0.14588
BPIe-Alcohol
1197 597 600 0.05920 0.06553 0.09842
Table 7.9 Correlations between substance use and BPI-Externalizing
7.3 Discussion and Interpretation
Thus the results of the dissertation have revealed that the effect of parenting style is
significant in the NLSY-Child sample for 10-14 year old children depending on the child
outcome being investigated. In the case of smoking and alcohol consumption the results
are not significant but in the case of behaviour i.e. the behavioural problem index the
results are highly significant. In the case of the behavioural problem index the linear
70
regression model is used but in the case of smoking and alcohol consumption the discrete
probability models are used. This study controlled for all family background factors
including parental substance use. In the case of parental substance use the results show
that the coefficients are significant. Thus in this study the importance of Parenting style
is highlighted and Parenting style is constructed as an index from several questionnaires
responses. Therefore this is a study which brings out the importance of Parent-Child
interactions from the Sociology and Psychology literature and uses the methodology and
framework of Economics to model these relationships.
Parenting style is an independent variable influencing child outcomes, focusing here
on alcohol and smoking. While exploring this relationship there is a need to control for
all other influences which are simultaneously impacting the child outcomes. Parenting
style is also distant from parenting practices which are the actions parents can take.
Parenting style is a broader and comprehensive term which consists of various parenting
practices and additionally a broader spectrum of parental behaviours which define the
parenting style in these households.
In the switching results the pattern of results turned out this way because Disengaged
is pulling out from every other category and there is a very high percentage of
disengaged. Across years its highly consistent. This study enables us to understand the
importance of all explanatory factors in substance use by young children. These results
and studies are important in determining how policy makers could influence these
juvenile delinquent behaviours. These behaviours are potentially risky both for the
individual and also put the society at risk in general due to their impact through various
criminal activities. Thus the dynamics of intra-familial interactions is one more area
71
which is being exploited to get a better view of a healthy society which has healthy
children as well from the perspective of maintaining peace and order which needs the
youth to function in an orderly manner.
Thus there is a substantial interest in trying to find all the possible causal mechanisms
which can explain these behaviours and in the case of very young children the parental
control is much higher then parenting style can be explored as a logical explanation for
substance use in households. Its partially an explanation or an interest to model
comparative statics where current phenomenon of substance use can be explained.
Moreover its also to predict and forecast these behaviours and how such families could be
identified as possible homes for juvenile delinquency. This is important especially in the
current age group which is young enough to be identified and if possible corrected to
prevent the onset of substance use in later adolescent years. There are several studies
which examine the high-school population and there are surveys such as Monitoring the
Future Surveys and High School and Beyond Surveys which concentrate on older
adolescents. The interest in younger age groups is important because the process or
likelihood of substance use can be identified and moreover these youth would be entering
the labour market in the future and therefore can be identified early in the process.
7.8 Explanation
The main hypothesis centered around investigating the smoking and alcohol
outcomes and bpi for young children and the research question addressed here was how
significant is parenting style as an explanatory variable in determining child substance
use.
72
The results showed in the dissertation proved to be significant and important and
contribute to a deeper understanding of parenting style and parenting style variables
proved to be significantly affecting child substance use. This result can be supported by
the existing literature in child and family studies (Robinette, Fletcher and Wright, 2002)
and (Byrne, Haddock & Poston, 2002) which showed this kind of a relationship.
Several possible explanations could be forwarded for these pattern of results. One
explanation is that the small proportion of children who actually engage in these
behaviours skewed the results due to the small sample size. Also this could be because
peer effects are becoming stronger in this age group especially in the higher age groups
closer to 14 and above. This could be responsible for the declining influence of parenting
style, also when children go to high school there are additional neighborhood effects. The
other explanation goes further back into the questionnaire and the wording of the
questions, and possibly the questions were not selective enough in screening the children
who engaged in these deviant behaviours, or the children did not perhaps understand the
questions. In the case of the follow-up questions which were used to construct indexes of
the intensity of these behaviours measuring the number of cigarettes smoked and the
number of alcoholic drinks consumed, the parenting style variables come to be a
significant predictor for alcohol consumption when controlling for child-specific effects.
Though some of the results particularly in the case of smoking and alcohol consumption
in the discrete-choice models the results were not anticipated but it does not necessarily
point to a �bad data� problem or the hypothesis needs to be reformulated.
This could be because smoking and alcohol outcomes are inappropriate outcomes for
this age group and positive outcomes very rarely occur and thus it is hard to estimate
73
probit models. This could even mean that this is something unexpected and warrants for
further study and that the role of parenting style in the case of smoking and alcohol
consumption needs to be re-examined and while the results did not support the
hypothesis, the effect of parenting style is not necessarily unimportant or should be
neglected. In this sample and with this age-group of children between 10 and 14 and
with the questions which were asked of these children the results are in this direction and
this is especially clear when with the questions used on intensity of cigarettes used and
alcohol consumed i.e. the number of alcoholic drinks and the total number of cigarettes
consumed the results come out to be highly significant.
In the case of an alternative dependent variable which is child behaviour and not as
adverse as the substance use outcomes the results come to be highly significant as well.
Thus this proves that the results are highly driven by the selection of variables.
74
CHAPTER 8: CONCLUSION
8.1 Conclusion
Thus the results of the dissertation have revealed that the effect of parenting style is
significant in the NLSY-Child sample for 10-14 year old children depending on the child
outcome being investigated. In the case of smoking and alcohol consumption the results
are not highly significant but in the case of behaviour i.e. the behavioural problem index
the results are significant. In the case of the behavioural problem index the linear
regression model is used but in the case of smoking and alcohol consumption the discrete
probability models are used. This study controlled for all family background factors
including parental substance use. In the case of parental substance use the results show
that the coefficients are significant. Thus in this study the importance of Parenting style
is highlighted and Parenting style is constructed as an index from several questionnaires
responses. Therefore this is a study which brings out the importance of Parent-Child
interactions from the Sociology and Psychology literature and uses the methodology and
framework of Economics to model these relationships. Parenting style is an independent
variable influencing child outcomes, focusing here on alcohol and smoking. While
exploring this relationship there is a need to control for all other influences which are
simultaneously impacting the child outcomes. Parenting style is also distant from
parenting practices which are the actions parents can take.
75
Parenting style is a broader and comprehensive term which consists of various
parenting practices and additionally a broader spectrum of parental behaviours which
define the parenting style in these households.
8.2 Policy Recommendations
Thus this study contributes to the literature with a better understanding of the dynamics
of intra-familial interactions. These interactions enable the preservation of the social
fabric which underlies a healthy population. Thus there is a great interest in finding the
optimal parenting style for society which would not only mitigate the substance use by
young children in households but would lower crime and juvenile delinquency. This
study enables us to understand the importance of all explanatory factors in substance use
by young children. These results and studies are important in determining how policy
makers could influence these juvenile delinquent behaviours. These behaviours are
potentially risky both for the individual and also put the society at risk in general due to
their impact through various criminal activities. Thus the dynamics of intra-familial
interactions is one more area which is being exploited to get a better view of a healthy
society which has healthy children as well from the perspective of maintaining peace and
order which needs the youth to function in an orderly manner. Thus there is a substantial
interest in trying to find all the possible causal mechanisms which can explain these
behaviours and in the case of very young children the parental control is much higher
76
then parenting style can be explored as a logical explanation for substance use in
households.
Its partially an explanation or an interest to model comparative statics where current
phenomenon of substance use can be explained. Moreover its also to predict and forecast
these behaviours and how such families could be identified as possible homes for
juvenile delinquency.
This is important especially in the current age group which is young enough to be
identified and if possible corrected to prevent the onset of substance use in later
adolescent years. There are several studies which examine the high-school population
and there are surveys such as Monitoring the Future Surveys and High School and
Beyond Surveys which concentrate on older adolescents. The interest in younger age
groups is important because the process or likelihood of substance use can be identified
and moreover these youth would be entering the labour market in the future and therefore
can be identified early in the process.
77
Means and standard deviations under different parenting styles across Gender
Parenting Styles
Sample Size Girls)
Sample Size (Boys)
Mean (Girls)
Mean (Boys)
Standard deviation (Girls)
Standard deviation (Boys)
Authorve 821 822 0.23021 0.2494 0.4212 0.4329 Authoran 821 822 0.10597 0.0815 0.3080 0.2738 Permisve 821 822 0.23508 0.2567 0.4243 0.4371 Disengag 821 822 0.4287 0.4124 0.4952 0.4926
Table 8.1 Means and Standard Deviations under parenting styles across gender
78
Table 8.2 Cronbach alpha measures of coefficient
Item Means
Mean
Minimum
Maximum
Range
Variance
.4870 .0459 .9891 .9432 21.5513 0.1260
Item Variances Mean Minimum Maximum Range Variance .1302
.0108 .2456 .2349 22.8066 0.0070
Table 8.3 Means and Variances of Factor Analysis Items
Cronbach Alpha measures of coefficient Parenting Styles
Sample Size (Girls)
SampleSize (Boys)
Mean (Girls)
Mean (Boys)
Standard deviation (Girls)
Standard deviation(Boys)
Authorve
821 822 0.23021 0.2494 0.4212 0.4329
Authoran
821 822 0.10597 0.0815 0.3080 0.2738
Permisve
821 822 0.23508 0.2567 0.4243 0.4371
Disengag
821 822 0.4287 0.4124 0.4952 0.4926
79
Eige
nval
ues
Number1 2 3 4 5
.5
1
1.5
Figure 8.1: Eigenvalues plot
80
Figure 8.2 Trends in smoking cigarettes and consumption of alcohol for 12-17 year old respondents
Trends in smoking cigarettes and consumption of alcohol for 12-17 year old respondents
(1965-97)
0500
1000150020002500300035004000
1965
1968
1971
1974
1977
1980
1983
1986
1989
1992
1995
Years
Num
ber o
f ini
tiate
s (1,
000s
)
SmokeAlcohol
81
BIBLIOGRAPHY
Adams, Michelle Janssen. (1995). �Youth in Crisis: An Examination of Adverse Risk Factors Affecting Children�s Cognitive and Behavioural/Emotional Development, Children Ages 10-16�, Ph.D. Dissertation, The University of Texas at Dallas.
Akabayashi, Hideo. (1996). �On the role of incentives in the formation of human capital in the family� Ph.D. Dissertation, University of Chicago.
Albanese, Paul J. (1988). Psychological Foundations of Economic Behaviour edited with a foreword by Tibor Scitovsky.
Andreoni, James (1989). �Giving with Impure Altruism: Applications to Charity and Ricardian Equivalence�, Journal of Political Economy, 1989, 97 (6), 1447-1458. Armsden, Gay C. and Mark T.Greenberg. (1987) �The Inventory of Parent and Peer Attachment: Individual Differences and Their Relationship to Psychological Well-Being in Adolescence�, Journal of Youth and Adolescence, Vol. 16, No. 5, 427-455.
Batson, C. D. (1991) The altruism question: Towards a social social-psychological answer, Hillsdale, NJ: Erlbaum.
Baumrind, Diana (1989). �The influence of parenting style on adolescent competence and substance abuse�, Journal of Early Adolescence, 11(1), 56-95.
Becker, Gary S. (1991). A Treatise on the Family, Enlarged edition, Cambridge, MA: Harvard University Press.
Bernheim, D. & Stark, O. (1988). �Altruism within the family reconsidered: Do nice guys finish last�, 78, American Economic Review, 1034-45.
82
Bracken, Bruce A., Sherry Bunch, Timothy Z. Keith and Patricia B. Keith (2000). �Child and Adolescent Multidimensional Self-Concept: A Five-Instrument Factor Analysis�, Psychology in the Schools, Vol. 37 (6), 483-493. Brandstatter, Hermann and Werner Guth (1994). Essays on Economic Psychology, Springer-Verlag 1994. Brock, William A. and Steven N. Durlauf (2000).�Discrete Choice with Social Interactions�, Review of Economics Studies, 2001, 68, 235-260.
Bronfenbrenner, Urie (1986). �Ecology of the Family as a Context for Human Development: Research Perspectives� Development Psychology, Vol. 22, No.6, 723-42.
Cardia, Emanuela & Michel, Philippe, (2004). "Altruism, intergenerational transfers of time and bequests," Journal of Economic Dynamics and Control, Elsevier, vol. 28(8), pages 1681-1701. Carlson, Marcia Jeanne (1999). �Family Structure, Father Involvement and Adolescent Behavioural Outcomes�, Ph.D. Dissertation University of Michigan.
Courts, Frederick A. (1966). �Psychological Statistics: An Introduction�, Homewood: Dorsey Press, 1966.
Darling, Nancy and Teru Toyokawa �Construction and Validation of the Parenting Style Inventory 11 (PS 11)�, Department of Human Development and Family Studies. Dovidio, J. F., & Penner, L. A. (2001). Helping and altruism. In Fletcher, G. J. 0. & dark, M. S.(Eds.) International handbook of social psychology: Interpersonal processes (pp. 162-195) Oxford, UK: Blackwell. Eberweing, Randall Olsen and Patricia Reagan, (2003) �Intracluster Correlation and Complex Sampling: Do Geographic Data Lessen the Problem�.
Emerson,Patrick M. and Andre Portela (2001). �Bargaining over Sons and Daughters: Child Labor, School Attendance and Intra-Household Gender Bias in Brazil� Working paper No. 02-W13, May 2002.
Feldman, S. Shirley, and Glen R. Elliott, eds. (1990) �At the Threshold: The Developing Adolescent�. Cambridge, Mass.: Harvard University Press.
83
Finken, Laura Lei (1996). �A Developmental Extension of the Propensity-Event Theory to Adolescents� Reckless Behaviour�, Ph.D. Dissertation, University of Nebraska-Lincoln. Fudenberg, Dean and Jean Tirole (1984). �The Fat-Cat Effect, the Puppy-Dog Ploy, and the Lean and Hungry Look�, The American Economic Review, Vol. 74, Issue 2, 361-66 Gardner, Robert C. 2001. �Psychological Statistics using SPSS for Windows�.
Green, Leonard, and John H. Kagel, eds. (1987). Advances in Behavioural Economics. Vol. 1. Norwood, N.J.: Ablex Publishing.Green.
Gregory, Robert J. (2000). �Psychological Testing: History, Principles and Applications� Third Edition, Allyn &Bacon, Inc. Needham Heights MA USA.
Hao, Linxin, V. Joseph Hotz and Ginger Zhe Jin (2000). �Games Daughters and Parents Play: Teenage Childbearing, Parental Reputation and Strategic Transfers� NBER Working Paper 7670.
Hill, J.P. (1987). �Research on adolescents and their families: Past and Prospect�, In C.E. Irwin (Ed.) New Directions for Child Development, Vol.37, 13-31.
Hernstein, Richard J. and Charles Murray (1994). �The Bell Curve: Intelligence and Class Structure in American Life� The Free Press. 1994. Hirschi, Travis and Michael J. Hindelang (1977). �Intelligence and Delinquency: A Revisionist Review�, American Economic Review, Vol. 42 (August): 571-87.
Hoffman, Martin L. and Lois W. Hoffman (1964). �Review of Child Development Research�, Volume 1 Publisher: University of Chicago Press 1964. Holmstrom, Bengt (1979). �Moral Hazard and Observability�, The Bell Journal of Economics, Vol. 10, Issue 1, 74-91. Jessor, S.L., and Jessor, R. (1974). �Maternal Ideology and Adolescent Problem Behaviour� Development Psychology, 10, 246-54. Kalai, Ehud (1977) �Nonsymmetric Nash Solutions and Replications of 2-Person Bargaining�, International Journal of Game Theory, No. 6, pp. 129-133.
84
Kaufmann, Dagmar, Ellis Gesten, Raymond C. Santa Lucia, Octavio Salcedo, Gianna Rendina-Gobioff, and Ray Gadd (2000). �The Relationship between Parenting Style and Children�s Adjustment: The Parent�s Perspective�, Journal of Child and Family Studies, Vol. 9, No. 2, 231-45.
Kooreman, Peter and Adriaan Soetevent (2002). �A Discrete Choice Model with Social Interactions; an Analysis of High School Behaviour� 1-28, CCSO Working Papers 200401, University of Groningen, CCSO Centre for Economic Research. Krebs, D. L. (1975). Empathy and altruism. Journal of Personality and Social Psychology, 32, 1134-1146. Kung, M.S. and Albert D. Farrell (2000). �The Role of Parents and Peers in Early Adolescent Substance Use: An Examination of Mediation and Moderating Effects� Journal of Child and Family Studies, Vol. 9 No. 4, 509-528. Lane, J., Dean Gerstein, Lynn Huang and Douglas Wright (1997). �Risk and Protective Factors for Adolescent Drug Use: Findings from the 1997 National Household Survey on Drug Abuse� Department of Health and Human Services, Substance Abuse and Mental Health Services Administration.
Lambert, Richard (1983). �Long-term contracts and moral hazard� The Bell Journal of Economics, 1983, 441-452.
Larson, Reed W., Suzanne Wilson, B. Bradford Brown, Frank F. Furstenberg, Jr. and Suman Verma (2002). �Changes in Adolescents� Interpersonal Experiences: Are they being Prepared for Adult Relationships in the Twenty-First Century�, Journal of Research on Adolescence, 12(1), 31-68. Lipstick, Lewis P. and Leonard L. Mitnick (1991). Self-Regulatory Behaviour and Risk Taking: Causes and Consequences, Norwood, NJ: Ablex.
Lea, Stephen E.G., Paul Webley and Brian M.Young (1992). New Directions in Economic Psychology: Theory, Experiment and Application United Kingdom: Edward Elgar Publishing, 1992.
Maccoby, Eleanor E, and John A. Martin. (1983). �Socialization in the context of the family: Parent-Child Interaction� in Handbook of Child Psychology, Vol. 4 edited by E. M. Hetherington.
85
Mach, Traci and Catalina Amuedo-Dorantes (2001). �The Impact of Families on Juvenile Substance Use�, Working Paper, April 2001.
Murphy, Kevin R. and Charles O. Davidshofer (2001). �Psychological Testing, Principles and Applications�, Fifth Edition, New Jersey: Prentice Hall International.
Nash, J.F. (1953), �Two-person cooperative games�, Econometrica 21 128-140.
Pavoni, Nicola, and Arphad Abraham (2004). �First-Order Approach for Principal-Agent Models with Hidden Borrowing and Lending: The Two-person Case�, Society for Economic Dynamics: 2004 Meeting Papers, 572. Reagan, Patricia B. and Randall J. Olsen, (2000). �You Can Go Home Again: Evidence from Longitudinal Data�, Demography, Vol. 37, no. 13, 339-350.
Reagan, Patricia B. and Janet Currie (1998), �Distance to Hospital and Children�s access to care: Is being closer better, and for whom?� NBER Working Paper 6836.
Rogerson, Richard (1985). �Repeated Moral Hazard� Econometrica, Vol. 53(1), 69-76.
Rushton, J. P. (1980) Altruism, Socialization, and Society. Englewood Cliffs, New Jersey: Prentice-Hall. Santrock, John (1998). �Adolescence�, 3rd edition, Mc Graw Hill.
Seo, Gye Soon Kong (1998). �The Impact of Maternal Problem Drinking on Children�s Developmental Outcomes: Focus on Parenting as Mediator� Ph.D. Dissertation, The Ohio State University. Schneeweiss, Hans and Harald Mathes (1995) �Factor Analysis and Principal Components�, Journal of Multivariate Analysis, 55, 105-24.
Schroeder, D. A., Penner, L. A., Dovidio, J. F., & Piliavin, J. A. (1995). The psychology of helping and altruism: Problems and puzzles. New York: McGraw-Hill.
Shumacher, Jorg, Martin Eisemann, Andreas Hinz & Elmar Brahler �The Assessment of Perceived Parental Rearing and its Relationship with Life Satisfaction and Interpersonal Problems: A German General Population Study�.
86
Steinberg, Laurence, Nina S. Mounts and Susie D. Lamborn and Sanford M. Dornbusch (1991) �Authoritative Parenting and Adolescent Adjustment Across Varied Ecological Niches� Journal of Research on Adolescence, 1(1), 19-36.
Steinberg, Laurence (1989) �Adolescence�, 2nd edition, McGraw Hill.
Steinberg, Laurence, Sanford M. Dornbusch, B. Bradford Brown (1992) �Ethnic Differences in Adolescent Achievement�, American Psychologist, Vol. 47, No. 6, 723-729. Svejnar, Jan (1982). �On the theory of participatory firm�, Journal of Economic Theory, No. 27 pp. 313-330. Tefler, Jo Ann and Judy Lupart (2001) �Gender, Grade and Achievement Differences in Student Perceptions of Parental Support�, Presented at the Canadian Society for Studies in Education, Quebec.
Weinberg, Bruce (2000) �An Incentive Model of the effect of parental income on children� Journal of Political Economy, 109(2), 266-80.
Yin, Tao (2000) �The Relationship between Mother�s Alcohol Use and Child�s Well-Being� Ph.D. Dissertation, University of Maryland.