BMI AND WAGES

Rafaël

Schut

BMI and Wages

Rafaël Schut (11755229)

WG03 | mw. M.L.S. Rijvordt

Introduction to Econometrics

January, 2018

Content

Inhoud1 Introduction. 32 Theoretical framework. 42.1 Human capital theory. 42.1.1 Theory. 42.1.2 Model 42.2 Productivity theory. 52.2.1 Theory. 5* VERGEET GEEN CONCEPTUAL MODEL.. 52.2.2 Model 52.3 Social interaction theory. 72.3.1 Theory. 72.3.2 Model 72.4 Health and medication theory. 82.4.1 Theory. 83 Methods. 93.1 Information about the

dataset 93.1.1 Adjusting the dataset 93.1.2 Operationalization of

the variables. 93.2 Adjusting the models. 113.3 Evaluating the models. 126 References. 13

1 Introduction

Aside from the typical qualifications and work experience, various other

physical factors influence one’s ability to acquire better earning jobs and/or

higher salaries. Research has been conducted, observing 3000 employed US

citizens. Of every person, many variables such as their race, height,

bodyweight, religion and income is given. In this paper, the relation between

one’s height and bodyweight to income and employment is investigated. The aim

of the research is to confirm, through data obtained from an empirical experiment,

that job-applicants are not solely judged on their abilities relevant to work,

but on their physical appearance as well. The focus of this paper will be the relation between one’s physical

appearance and their job and income. This paper makes a clear distinction between

men and women; the effect of physical appearance on the job and income will be

calculated separately for each gender. As height and bodyweight are major

aspects of the physical appearance, the research question investigates to what

extent the height and bodyweight influence one’s income and employment. The

topic itself is of great importance, as it confirms the presence of

(unconscious) bias by employers. If this bias of employers happens to be true, having knowledge about this will allow

the job-applicants to prepare themselves better. Through adaptation,

job-applicants can correspond to the image of the ideal person for the job.

The research question will be approached through an analysis of 3000

observations. First of all, the hypothesis of the research will be stated. Secondly,

the results of all observations will be processed in a computer program, known

as EViews. With the use of the

program EViews a regression model

will be constructed. The regression model describes the relation between two

variables, in this case the height and bodyweight in relation to the income and

employment. The regression model will provide insight as to what the relation

between the variables may potentially be. Based on the acquired numbers, a conclusion

will be drawn accordingly.The report is structured into four different parts. First of all, a theoretical

framework will be presented serving as a lens to evaluate the research

question. Secondly, the methods of research will be discussed carefully. In

particular, the results of the observations, which may potentially contain

improper data and outliers, will be discussed thoroughly in order to prevent

false data analysis. Thirdly, the results of the research will be analysed with

the help of EViews building a

foundation for the discussion of the topic. The report will be concluded with a

discussion about the topic and the presence of any unexpected results.

2 Theoretical framework

2.1 Conceptual model

2.2 Human capital theory

2.2.1 Theory

Human capital is the accumulation of abilities,

knowledge and experiences of individuals that can be utilized to produce

economic value. Human capital can be gained through cumulative job experience,

and lost through unemployment or simply the passing of time (Berndt, 1991). Individuals are able to invest in human

capital through formal education; however, firms share this ability through the

possibility of on-the-job training (Berndt, 1991). The broadening of one’s

human capital leads to a higher productivity (Becker 1994). According to the

human capital theory, employees are paid on basis of their productivity. As the

human capital becomes greater, the productivity rises, leading to higher income

for the employee and turnover (per person) for the firm. Therefore, it is

attractive for firms to stimulate and guide the employee in developing the

employees’ human capital, which in turn increases the firm’s capital.

2.2.2 Model

In this paper the

human capital earnings function (HCEF)

constructed in 1974 by Mincer, is

used as the model of the human capital theory:

(1)

The variables are defined as follows: represents earnings, S represents years of completed education, X represents the number of years an

individual has worked since completing schooling, and e is a statistical

residual.

The

logarithm of earnings is the composition of a linear and quadratic equation:

the linear education term and a quadratic function of “years of potential

experience”. Mincer proposed, in the absence of explicit data on experience,

the use of “potential experience”: the years and individual with age A could

have worked, under the assumption that he started school at age 6, took exactly

S years of schooling to finish S years of schooling and started his career

immediately after finishing his schooling: (Card, 1999).

The use of the log transformation knows various benefits. Firstly, the

distribution of log hourly earnings closely resembles the normal distribution,

making the transformation useful for data analysis. In addition, the

transformation allows the success of the HCEF, as the distribution is closely

approximated by the standard HCEF equation (Card, 1999). Another benefit is

that it can be mathematically proven that the coefficients in the equation of a

log function indicate percentage increases. Therefore the interpretation of the

meaning behind the coefficients is easier. Lastly, the log transformation is

known for its resistance against the impact of extreme outliers.

2.3 Productivity theory

2.3.1 Theory

Perhaps one of

the major objectives of a firm is to yield high productivity. Research displays

a positive correlation between obesity and presenteeism, that is, a reduced

productivity on the job. In particular obesity

has been shown to obstruct the employee’s productivity in physically-demanding jobs

(Baum and Ford, 2004). Health

effects on productivity are concentrated among the most obese workers with BMIs

of 35.0 and greater, suggesting that employers should consider workplace

interventions targeting obesity. Even modest weight loss could result in

hundreds of dollars of improved productivity costs per worker each year (Gates,

Succop, Brehm, Gillespie & Sommers, 2008). Furthermore,

as firms do not gain any turnover from absent employees, the absenteeism

related to obesity causes foregone earnings to the firm.

2.3.2 Model

In a report (Han, Norton & Stearns, 2009), the estimation model considers

the log hourly wages as dependent variables, which are modeled as a function of

weight status, age, other covariates and individual fixed effects, resulting in

the main equations:

ln() = h , ) (2)

Where suffixes i and t stand for individual and time, respectively. BMI in two dummy

variables for overweight and obese and with normal weight or underweight as the

reference groups. Categorical variables are chosen to measure incremental

effects of moving up one clinical weight classification on the probability of

employment and log wages. This specification does not impose a constant

marginal effect of a one-unit increase in BMI on the probability of employment

or log wages across the entire BMI range. Recent research displays that being

overweight adversely affects the employment across all race-gender subgroups

except black women and men1.

Therefore, SkillType is a vector of dummy variables representing interpersonal

skills required in each occupation. and denote other control variables at the

individual and state levels, respectively. M stands for time-invariant

individual fixed effects and e stands for independently identically distributed

error terms. Individual-level covariates include years of work experience. Furthermore,

the wage equation is modified by including interactions of BMI groups and age

and a three-way interaction of BMI groups, age, and characteristics of

interpersonal skills required in each occupation. The modifications allow

estimating any changes in the effect of BMI groups on hourly wages with age and

characteristics of interpersonal skills in each occupation.

2.4 Health and medication theory

2.4.1 Theory

It has been shown that “obesity is associated with a 36 percent increase in

inpatient and outpatient spending and a 77 percent increase in medications.” As

a result, employers, who provide health insurance, are less inclined to pay

high wages as the labor costs of overweight and obese workers are high (Sturm,

2002). Moreover,

research stated that body weight increases health expenditures more steadily

for women than men (Wee et al., 2005). If employers are to (partly) cover health

costs for the employees, the greater expenses could possibly dissuade employers

from hiring overweight women in comparison to

overweight men.

2.4.2 Model

The model consists of individuals’ wages that are related to BMI and other

variables (X is a vector of variables that affect wages), at time t, according

to equation:

+ ²

(1)where is the

residual. Exogeneity of BMI assures that OLS estimates of ? can be interpreted as a consistent

estimate of the true effect of BMI on wages. However, it can be reasoned that

BMI is endogenous, as BMI could be influenced by non-genetic factors such as

individual choices and environment and the possibility that obesity may be

influenced by wages. Endogeneity of BMI instead leads to inconsistent estimates

of the effect of weight on wages. In an attempt to

identify the possible sources of endogeneity, the residual in (1) is decomposed

into a genetic component (G), a non-genetic component (NG) and a residual (?), independent and identically

distributed random variables over individuals and time (De Sousa, 2012):

(2)

3 Methods

3.1 Information about the dataset

The National Longitudinal Survey of Youth 1979 (NLSY79) was a panel survey

conducted by the Bureau of Labor Statistics in the United States. In this

paper, the research will be done with the help of the dataset from NLSY79. The

dataset that is used to investigate the research question, dates from the

period of 1979 to 1994. All data has been conducted using US citizens. In the

beginning 3,003 males and 3,108 females participated in the sample; however,

for this research question, only 3000 observations will be used. In the

sixteen-year period, every two years the participants were interviewed. For

every participant, the education, training, employment, marital status,

fertility, health, child care and assets and income were held up-to-date.

Information of every participant is divided into four different categories. The

first category conducts personal variables of the participant are given: sex,

ethnicity, age (in 2002), educational background, marital status and faith. Secondly,

the respondent is asked about their family background, including years of

schooling of both parents; the possession of library cards in the family when

the respondent was 14; as well as the number of siblings and financial status

of the family (poverty). Lastly work-related information is asked from the

respondent, earnings, work experience, category of employment; place of

residence and pay set by collective bargaining.

It is noteworthy that some samples have been discontinued as they did not

correspond to the aim of having results representative for US citizens.

3.2 Operationalization of the variables

As the aim for this paper is to provide reliable results and confirm potential

relations, certain dummy variables, control variables and adjustments are

introduced.

Firstly,

the sex of the respondent is of great importance, as in previous research from

Blau and Kahn (1994) a wage gap between males and females was found. The respondent’s

sex is processed as a dummy control variable: females receive the number 0 and

men receive the number 1.

In

addition, it is of great importance to note that the BMI is an index

calculated on basis of the bodyweight and height of the respondent. The BMI serves

as an indication to what weight group the person is in, ranging from underweight

to obesity. A BMI under 18.5 is classified as underweight and a normal weight

is a BMI ranging from 18.5 to 24.9. Any value greater than 25 is considered

overweight and any value above 30 is classified as obesity. In this paper, the BMI serves as a dummy variable. If the

BMI of the respondent takes a value classified as either overweight or obese,

the value will be 0. For a BMI value ranging between underweight and normal

weight, the value is 1. In order to estimate the

effect of BMI on income, it is of great significance to bear in mind that not

every respondent has the same profession. Consequently, the variable SkillType

will be regarded as a dummy variable. The variable will take the value of 0 for

white-collar work and 1 for blue and pink-collar workers. Lastly,

due to the positive correlation between work experience and wages (Becker,

1994), the control variable of work experience is introduced.

3.2 Overview of variables

In order to keep a proper overview of the dataset, the relevant variables are

listed in the table below, with their abbreviations and definitions.

Variable

Abbreviation

Definition

BODY-MASS INDEX

BODYWEIGHT

BMI

BW

A value based on one’s

height and bodyweight, indicating if a person is underweight, overweight,

“normal” weight or obese

The respondent’s weight measured in pounds (lbs)

HEIGHT

The respondent’s height

measured in feet and inches

INCOME

The respondent’s income

measured in USD ($) per hour

SEX

A variable indicating

whether the respondent is male or female

SKILLTYPE

ST

The category of profession

of the respondent

WORK EXPERIENCE

The respondent’s accumulated

working experience in years

NATURAL LOG OF INCOME

NLI

The natural logarithm of

income

Table

1: Variables used in the dataset and their abbreviations as well as definitions

3.3 Evaluating the models

The aim of this research is to evaluate whether there is any correlation

between BMI and level of income. As three theory models will be discussed, a

comparison of the results from the regression analysis will be drawn.

3.3.1 Regression

In order to

compare the different models, the multiple regression analysis will be applied

to every model. The three estimated equations have been created in such a way

that allows OLS to be applied as estimator for the coefficients. Statistical

properties provide information about the estimated equations and their

reliability. In this paper, the , p-values and t-statistics of the coefficients will provide

insight as to which model yields the most reliability.

6 References

Becker, G. S. (1994). Human capital revisited. In Human Capital: A

Theoretical and Empirical Analysis with Special Reference to Education (3rd

Edition) (pp. 15-28). The University of Chicago Press.

Berndt, E. R.

(1991). The practice of econometrics: classic and contemporary. Reading:

Addison-Wesley.

Card, D. (1999). The causal effect of education on earnings. Handbook of

labor economics, 3, 1801-1863.

De Sousa, S. (2012). Does size matter? A propensity score approach to the

effect of BMI on labour market outcomes. EUI Florence.

Gates, D. M., Succop, P., Brehm, B. J., Gillespie, G. L., & Sommers, B. D.

(2008). Obesity and presenteeism: the impact of body mass index on workplace

productivity. Journal of Occupational and Environmental

Medicine, 50(1), 39-45.

Han, E., Norton, E. C., & Stearns, S. C. (2009). Weight and wages: fat

versus lean paychecks. Health economics, 18(5), 535-548.

Manning, A. (2000). Movin’on up: Interpreting the earnings–experience

profile. Bulletin of Economic Research, 52(4), 261-295.

Wee, C. C., Phillips, R. S., Legedza, A. T., Davis, R. B., Soukup, J. R.,

Colditz, G. A., & Hamel, M. B. (2005). Health care expenditures associated

with overweight and obesity among US adults: importance of age and race. American

Journal of Public Health, 95(1), 159-165.

1 In particular, professions that require interpersonal

skills, with presumably more social interactions, the negative relationship

between BMI and wages is amplified (Han, Norton & Stearns, 2009).