SlideShare une entreprise Scribd logo
1  sur  19
The Effect of Working Hours on Health Care Expenditure in the United States
Year after year, the United States is becoming known as a country of workaholics. In
2015, employees in the U.S. reached a 40-year high for unused vacation days. Oxford University
calculated this phenomenon to equate approximately to 169 million days, or the $52.4 billion in
lost benefits (Kasperkevic 2015). Furthermore; for the United States being the most powerful and
influential world leader, we do not mandate employers to provide paid vacation time. Only one
in four Americans receive paid vacation or holiday, which trails well behind a bulk of the
developed world. I find this trend to be regressive in terms of societal norm; and intend to
support this through the discovery of adverse effects on health care costs related to individuals’
working hours (Mohn 2013). The goal of my study is to investigate, and possibly quantify, the
degree to which working hours affects one’s health (using health care expenditure as a proxy).
An Economist article titled Working Hours: Get a Life from 2013 took a look into the
relationship between hours worked per year and productivity. To help identify its readers with
the purpose of their article; C.W. and A.J.K.D. referenced essays from Bertand Russell (“In
Praise of Idleness” 1932) and John Mayard Keynes (“Economic possibilities for our
grandchildren” 1930). The objective of these essays was to emphasize a utopian lifestyle that
would arise from a rise in technology, and more importantly, a lessening demand for labor.
Despite a vast rise in technology since the 1930’s, we have yet to cultivate an economy in which
labor is not the driving force.
A simple regression between hours worked per year (2080 hours/year equaling 40
hours/week) and GDP per hour worked (productivity) shows a strong negative correlation
between these two variables confirming that the marginal productivity is a diminishing function
of working hours. A simple empirical example comes from comparison of Greek and German
workers. Despite the fact that Greeks are hard-working individuals averaging about 2,000
working hours per year, Germans who work a mere 1,400 hours each year achieve a productivity
measure that is 70% higher than those of Greek workers (C.W. 2013).
In addition to the negative impact of working hours has on productivity, a recent study
recent study published by Lancet Medical Journal showed health related risks of long working
hours. The aforementioned study displayed a positive relationship between working hours
greater than 40 per week and cardiovascular issues. The study found that people working just an
hour extra a week, between 41 and 48 hours per week, have a 10% higher risk of stroke or other
cardiac events; while people working 49 to 54 hours per week have a 27% higher risk of stroke
(Mohney 2015).
According to the above study, the health risk only rises with additional working hours.
Individuals who worked 55 hours a week or greater were shown to be 33% more susceptible to
strokes, even after monitoring other risk factors such as smoking and alcohol consumption; and a
13% increased risk for coronary heart disease or heart attack (Mohney 2015). The statistical
inferences in this study are similar to what I hope to find in my research; relating to long working
hours and general adverse effects it has on an individual’s health.
The main research question of my study is to investigate the relationship between health
care expenditure and the number of hours individuals work per week. My research hypothesis is
that an increase in weekly working hours over a certain threshold positively correlates with
health care costs. This should be consistent with the findings of previous studies; demonstrating
that an increase in working hours has a negative impact on health, and thus, positively relating to
health care expenditure.
Literature Review
There is a substantial ongoing research focusing on the correlation between work and
health, but only few investigate this link for workers of the United States. A recent study
conducted using a sample of Korean workers illustrates the impact of gender on self-rated health,
in regards to hours worked per week (Seong-Sik 2015). The study consisted of men and women
similar in age distribution, who were asked to rate their health on the 5-point SRH scale (ranging
from “very good” to “very poor” health). Working hours were placed into ranges; including 20-
35h (less than standard working hours), 36-40h (standard working hours), 41-52h (overtime
within legally permitted working hours excluding weekend work), 53-68h (overtime within
legally permitted working hours including weekend work), and 69h or more (legally prohibited).
The results of this study showed women to have a higher proportion of poor health.
Approximately 75% of men and women worked more than standard working hours, but more
women worked longer than 5 days per week. This is likely due to women receiving significantly
lower wages than their male counterparts and needing additional income. Of the individuals in
the study, women were found to hold whiter collar jobs, and were also found to be less educated
than the participating males (which would inherently contribute to the gender wage differential).
Seong-Sik (2015) stated that working hours in the South Korea’s labor force are strongly driven
by factors of demand (the vast majority of individuals working over 40 hours per week). In
regards to women, the statistical findings were realistically supported by the idea of women
balancing their work and family lives. Attempting to support a family and take care of a family
would in doubt bear more burden on women compared to men. In sum, this study found that
higher levels of working hours contributed to poor health. In C.W.’s Working hours: Get a life
(2013); South Korea’s labor force is graphically shown to work an additional 400 hours per
person per year in 2012, compared to an approximate 1,000 working hours differential in 1990.
Workplace interventions have shown to be impactful to individual’s health and
consequently, productivity (Vander Klink, Blonk, Schene, Van Dijk 2001). Specifically,
workplace interventions in the forms of cognitive-behavioral and multi-model interventions were
found to have a significant impact on work-related stress. Outcomes that had the most significant
impact on work-related stress (deriving from workplace interventions) include: complaints,
psychological resources, responses, and perceived quality of work life. Although this study has
shown us the advantageous of taking a workplace intervention, I suspect the same advantage to a
greater magnitude will occur with a simple reduction in working hours of an individual. The
importance of this research is to exemplify a technique in order to counteract the negative effects
of our working culture.
Another study by Fein & Skinner (2015) looked at major occupational groups, along with
gender, to identify pathways through which work hours impact health (Fein & Skinner 2015).
The theoretical framework used in their research was based around the idea that working hours
increases work-life conflict, and hence leads to health outcomes. Previous research by Skinner
estimates the relationship strength between work-life-conflict and health to be between 0.23 and
0.40, depending on particular health outcomes. The relationship between work hours and health
in said study estimates a weaker relationship than work-life-conflict and health. It was shown
that the coefficient for working hours’ effect on health was 0.15 for psychological health, and
0.06 for physical health. The research suggests that work hours were only associated with
negative health outcomes when the hours did not fit or interfered with workers’ other
commitments or activities. For women, this effect was shown to be greater, similar to Seong-Sik
(2015). The average woman may struggle to balance a work-life and family obligations; to a
greater extent than their male counterparts (Fein 2015).
Bell, Otterbach, and Sousa-Poza (2012) have recently conducted a study to investigate
the impact of the difference between actual and desired work hours on self-perceived health
outcomes. Their purpose was to indicate the consequences of employment policies and shed light
onto the discrepancy between actual and desired work hours. This research shows that many
individuals are overemployed on average, working 4+ hours per week than desired. One notable
insight found was that overemployed employees were generally less satisfied with their own
health than unconstrained full-time workers. Several other significant inferences included
stronger negative coefficients related to overemployment (compared to unconstrained and
underemployed workers) in relation to health satisfaction and self-assessed health, over all
weekly working hour intervals; ranging from (Bell 2012).
In another study, Nagashima, et. al (2007) accepted that there was a statistically
significant relationship between working hours and an individual’s health. Thus, their research
aimed at identifying a threshold in which number of hours worked per month had a profound
impact on mental and physical fatigue. They developed their results through the distribution of
questionnaires to 843 (720 used in the study) male factory workers, which I suspect could have
influenced their findings. Nagashima’s results stated that working greater than 260 hours per
month would begin to have an adverse impact on the Self-Rating Depression Scale (SDS).
Measures involved in this scale include decreased vitality, general fatigue, physical disorders,
irritability, decreased willingness to work, anxiety, depressive feelings, and chronic tiredness; in
which general fatigue and tiredness, depressive feelings, irritability and decreased vitality were
most prevalent. In order to minimize these effects, males working in the sample factory should
work less than 260 hours per month (Nagashima 2007).
Conceptual Framework
There are also a number of additional implications of high working hours on health care
expenditure pertaining to the theory of demand for health capital. Basic economics tells us that
individuals use their income to maximize a preference function (utility). In regards to demand for
health capital, the preference function is as follows: Ut[utility]=U[Ht (health), Xt (other
commodities)], in a given period t. Since we know humans are not immortal, it is so that every
individual has a depreciation factor on their health. The health function for an individual in a
future period is dependent on his/her current health, minus the given depreciation consistent with
aging/risk averseness, plus investment into health-benefiting activities.
The Grossman Model states that individuals have the choice to invest in either medical
care (their health) or other market goods and services. Beyond this trade-of between health and
market goods is how any person spends their time. The time constraint of this individual is as
follows: TT (total time) = TW (time at work)+ TH (time spent on health) + TX (time spent on
other commodities) + TL (time lost due to illness).
My research is focused around two effects that will result from the aforementioned
model, which are highly related to the workplace culture in the United States; the first of which
is the wage effect. This effect states that the higher an individual’s wage the less incentive he/she
has to invest time into their health. Such as, an hour of working out would be more costly to a
CEO of a large corporation compared to a low wage, general labor employee. Secondly, the
Grossman Model also suggests the hours effect. This effect states that the more time you spend
working, the less time you have to invest elsewhere (specifically into time spent on health). For
example, a full-time worker will have less time available for consumption than a part-time
worker, and thus, will likely invest less time into their health.
I believe that Americans spend too much time at work; thus suffer additional time lost
from illness or sickness. Furthermore, U.S. obesity rates continue to rise year over year along
with diabetes, while physical inactivity stays level. Despite the lackluster investment of time
spent on health, the United States still pays second most in the world on health care.
This phenomenon is a product of the U.S. labor market, which encourages workers to
seek extraordinary working hours. Therefore, I intend my research to empirically prove the
positive relationship between working hours and health care expenditure (as a proxy to represent
individuals’ health). This would exemplify my belief in how our society has shifted to dedicate
more time towards work, consequently devaluing time spent on health and increasing the cost of
health care in the United States. This research could potentially make significant inferences into
the cost-benefit analysis individuals’ use when evaluating their demand for health capital;
incentivizing people to invest more time into better-health producing activities, rather than
working more hours simply to incur additional medical expenses.
Data & Descriptive Statistics
The data used to create this report are derived from the Current Population Survey (CPS).
The CPS is one of the oldest, largest surveys jointly sponsored by the U.S. Census Bureau and
the U.S. Bureau of Labor Statistics. Data composing the Current Population Survey are gathered
on a monthly basis from citizens of the United States. The sample size of this expansive survey is
approximately 90,000 individuals over the age of fifteen; excluding persons in the Armed Forces
along with persons institutionalized in prisons, hospitals, nursing homes, and other such
institutions. Individuals in the sample are contacted year-over-year, with a 90% response rate.
The government is willing to sponsor this survey because of its immense use throughout
research regarding the United States. Beyond the benefits of making this information available
public for research, the U.S. government derives a number of macroeconomic variables from this
data, including various unemployment rates (U1-U6) and the labor force participation rate.
Furthermore, the CPS is a valuable investment for the government in its ability to give
information regarding broad, demographic information; such as age, race, education, income,
marital status, and so on. Having such a large sample size, the Current Population Survey serves
as the most important survey in capturing statistics unbiased across the U.S. population.
Table 1 displays the proportion
of individuals pertaining to each health
insurance classification. Those with
health care coverage make up 86.24%
of the United States’ population, while
only 13.76% do not have any form of
health insurance. Excess observations
not accounted for in this distribution were due to survey participants not supplying an answer, or
possibly having a recent change in coverage. Despite having 80.09% of the population being
privately insured, the remaining 6.15% of insurance being provided by the government; the
United States still experiences difficulty with “crowding out” of the private-insurance industry.
Health care services in the United States have experienced sharper inclines in pricing
than seen throughout any other industry, with the exception of higher education institutions. The
proportion of individuals holding private insurance increased 4.2% in 2014; this trend will
eventually put downwards pressure on healthcare prices in the U.S. market. However, there are
numerous forces that drive said prices upwards including such as implementation of technology,
advanced surgical methods, demand for assistive living, and so on. Exacerbating the issue is an
expanding portion of elderly people in this country, who are the primary demographic for health
care services.
In its entirety, the Current Population Survey consists of 90,430 observations; however,
to better serve my research, I excluded all individuals who did not provide an input for average
weekly hours worked. Dropping said data will allow me to more accurately identify the effect of
working hours on health care expenditure. Removing said data leaves 9,563 observations for
analyses and overall, a more accurate portrayal of the findings. Part time workers have been
classified as those who work 27 hours per week or less. 27 working hours per week was set as
the threshold between full and part time employment because 28 working hours per week is the
lowest point at which a worker in the United States can qualify as being fully employed by an
employer. Full time workers are defined as those who regularly work 28 or more hours per week;
while those who regularly work 45 hours per week or greater are considered overemployed.
The figure to the left
shows that the vast majority
of individuals (67.4%) in this
sample are full time workers.
Part time workers represent
17.2% of this data set. As
shown in the figure,
overemployed individuals
constitute a modest
percentage of workers
(15.4%). Although
overemployed individuals
are generally considered full
time workers, separate
classification is necessary in order to produce information related to the adverse effects of
strenuous labor driven from working an extraordinary number of hours.
Descriptive statistics in Table 2 are provided for the key independent variables. The set of
control variables includes, age, race, gender, disabilities, years of education, number of kids in a
family under six, weekly earnings, type of health insurance, and working hours. Demographic
information revealed in the descriptive statistics show a majority (64.6%) of respondents being
white, followed by 9.8% black and the remaining 16.2% of the population being Hispanic (while
9.2% are unspecified). The average age of those surveyed was 40 years old; with a minimum age
of 15 years and maximum age 64 years. The gender distribution in the analytical sample is split
evenly between male and females. Only 3.7% of sample observations reported having various
forms of disability. 57.3% of observations in the sample classified as married individuals.
Average years of education and works hours per week are 14.1 years and 36.5 hours,
respectively. Gross weekly earnings average $575.46 per week.
In Table 3, I have broken down the Descriptive statistics by working classification
(identifying sample size, mean, and standard deviation; vertically). There are several trends that
can be recognized within the different categories of insured/uninsured individuals. First, is the
trend of age to increase usual working hours; likely as a result of the need to save for retirement,
satisfy medical expenses, reassess financial position, and so on.
The next two variables look at race’s effect on working classification. Looking at the
white population, it can be seen that they consist of 66.26%, 61.45%, and 76.6% of part time, full
time, and overemployed; respectively. Black individuals on the other hand, display values of
8.9%, 11.06%, and 5.5% in the same respect. From this data, it can be inferred that the white
population is most likely to be overemployed, next to be part time, and least of the three to be
full time workers. The black population makes up a much less significant portion of the
population than whites, hence the lower percentages for worker classification. Information from
this data set tells us that most blacks are full time workers, followed by part time, and are least
likely to be overemployed.
Males display a trend of having working much more so than women. While men are
37.7%, 49.7%, and 67.7% to work part time, full time, and be overemployed; while women
comprise the remainder of each category. While this is the case now, women have been
progressing towards greater working hours over the years; evening this scale over time.
Those with a disability are significantly less likely to work full time and obviously less so
to be overemployed. A good percentage of individuals with disabilities are unemployed;
however, those observations have been removed from my test. Marital status has a strong impact
on working classification as well; increasingly likely to work a greater number of hours per
week. Education also trends upwards with working hours; as they are more encouraged due to
higher wages and greater value to a company.
Gross weekly earnings are a data set that obviously trends upwards with worker
classification. Interestingly, the gap between overemployment and full employment is greater
than that of the full time and part time differential; implying people are paid exponentially more,
on average, if they work more.
The trend in insurance status tells us that as one works more hours, they are more likely
to have private insurance; and definitely less so to be on government insurance as they earn too
much to qualify. Lastly, giving us a glimpse at my research question are descriptive statistics for
health care expenditure; in regards to working classification. Those who work a greater number
of hours, as suggested, will incur greater medical expenses. Each progression in working
classification brings a greater change to health care costs, despite a larger difference in working
hours from part to full time employment, compared to full time and overemployment.
As mentioned previously in this paper; observations with higher incomes and working
hours have implications on the time individuals spend on health activities. Thus, the higher these
inputs are, on average, the less time these individuals spend on health related activities will be as
well. This in turn, will result in higher healthcare costs. The following page contains quadratic
regressions for both working hours’ and weekly earnings’ effects on health care expenditure,
perfectly exemplifying these theories.
From the above visualizations, it can be seen that working hours and weekly earnings
have distinct quadratic relationships with health-care expenditure. The weekly earnings
relationship with health care expenditure consists of a narrower confidence interval over the
whole range of studied earnings, meaning weekly earnings has a more consistent relationship
with health care expenditure than working hours. Both positive associations are compatible with
our theoretical expectations coming from the Grossman Model. The regression analysis should
shed light whether these trends hold after controlling for a rich set of various individual level
factors.
Research Design & Results
My research hypothesis states there is a quadratic relationship between working hours and health
care costs. Working less hours per week increases the likelihood of poverty leading to poorer
health conditions; contrarily, working more hours dissuades individuals from spending time on
health related activities, while increasing disposable income (inevitably increasing money spent
on one’s health) simultaneously.
In the above empirical model, Yi is annual healthcare expenditure, which can be regarded
as a proxy for an individual’s health status, Hi is individuals’ usual weekly working hours and Xi
represents a set of control variables such as. There are a number of other factors that influence
the health care expenditure of an individual outside of my control and main independent
variables, which compose 𝜀 𝜀, the fit of the model not captured by the variables I have selected to
include.
I have chosen health care expenditure (Yi), calculated as the sum of premiums and out-of-
pocket medical expenses, to be my dependent variable testing my hypothesis. Self-perceived
health classification of individuals was an alternative to my favored dependent variable. Despite
it being a relevant variable for my test, I suspected the range of values was not wide enough to
provide meaningful interpretation. Furthermore, portraying my intended outcome in quantifiable
U.S. dollars as presented with health care expenditure will deliver a greater, more accurate
impact than factor data. Additionally, I transformed the dependent variable by taking the natural
log of it. As a result, parameters of the independent variables can be interpreted in the semi-
elasticity form, a percent change in health care expenditure due to a one unit change in an
independent variable.
Average working hours per week (Hi) is my main independent variable that thoroughly
answers my research question; the effect of working hours on health care expenditure. Although,
since my hypothesis suggests that working hours has a quadratic relationship with health care
costs, I have included a separate squared data set of those variables, in addition to the original
data set for usual working hours per week. The most inherently relevant control variable (Xi) I
had conjectured to include was average age of survey participants. No matter how much money
or time we invest into our health, its status will still depreciate, hence why age is an obvious
factor affecting health care expenditure of individuals. The older an individual is, greater health
care expenses can be expected. Furthermore, values for the age data set were found to be
consistently correlated with the other independent variables; most notably found in working
hours per week (both in linear and quadratic form) and wage.
Another independent variable I have configured is the categorical status of individuals’
health insurance; whereas those with private insurance have a value set equal to one and other
persons’ (holding government insurance or uninsured) are set equal to zero. Through the use of t-
table comparisons between costs of health care of the three different categories, I concluded
having private insurance significantly increases an individual’s expected health care expenditure.
Moreover, the regression’s results will give enlightening inferences into the magnitude and
significance of the disproportioned state of the health insurance industry (referring to the
“crowding out” theory)
On average, people with private insurance spend $2,217 per year on healthcare,
compared with $487 and $301 per year for those with government insurance and those
uninsured; respectively. This is the case as a result of the “crowding out” of the health insurance
industry; where individuals only purchase insurance if they have a forceful need to do so. Such a
large difference between costs of medical care among these categories, along with the results of
the t-tests indicate that privately held insurance has a strong, positive, and significant impact on
health care costs.
Health classification, as perceived by the survey participants, has also been introduced
into my model. This variable contains data ranging from one to five; one being indicative of
excellent health, while a value of five implies poor health conditions. Thus, the relationship
between health care classification and health care expenditure should result in a positive
relationship; meaning a greater value (indicating poor health) will cause greater health care costs.
Since my model is theoretically based around the health status of individuals in regards to
working hours, it is only reasonable to include a variable directly associated with their health
status as well. Despite the possibility of misperceived health care statuses, I found this health-
related variable to be the most suitable for the purpose of my research.
Wage is a data set I calculated by dividing gross weekly earnings by average usual hours
worked per week. While this variable may not be completely accurate to the person’s wage, it is
expressive of how wage is determined in such a context; consequently affecting the health care
costs of individuals. The consumer choice theory states that an increase in income increases
consumption of certain goods (normal goods); including those related to the person’s health
(research shows that healthcare can be considered as a normal good). Thus, this variable should
have a positive relationship with health care expenditure.
I have included wage as a control variable in my model because it enables me to
distinguish the income effect (greater disposable income to be broadly distributed) from the
hour’s effect. The hour’s effect conversely states that working more hours will limit an
individual’s excess time, therefore reducing demand for spending time on health related
activities. An important theory supporting my model is that the United States is unique to a
culture that supports working an abundance of hours per week. A consequence of this social
construction is the trend in the health of U.S. citizens; as can be seen in increasing obesity rates,
rises in mental health related illnesses, stress, and so on.
Lastly, education is another important variable to control for; because it has been proven
to have a strong, positive correlation with the health of an individual. Educated people are more
likely to partake in less risky behavior that will negatively impact their health, along with taking
proactive measures to combat illness or improve their health status. However, since my
dependent variable is related to health care expenditure instead of individuals’ actual health, the
coefficient may be suppressed.
Since educated people are more likely to be proactive in regards to their health, the effect
on health care expenditure contains negative pressure; whereas taking precautions to your health
will prevent future ailments, thus, health care costs. On the other hand, those with upper level
education on average, have private health insurance. As mentioned previously, this puts
significant upwards pressure on health care expenditure.
Below is a chart exhibiting the degrees of correlation between the variables used in my
test. As you can see, there are no extraordinarily high levels of correlation between any of the
variables, with the exception of working hours per week and working hours per week squared.
The most consistently elevated degrees of correlation lie with age, as previously stated. As one
gets older, they are more likely to work, have consistency of sustaining themselves above the
need for government insurance, while having demand for health insurance, consistent changes in
health classification, and have greater experience; thus higher wages.
Another high degree of correlation exists between education, wage, and private
insurance. A correlation of 16.08% between private insurance and wage is predictable. As one
becomes an increasingly impactful member of a company or industry, their benefits are to
increase as well (including health insurance packages). Wage and education were found to have
a relatively strong degree of correlation (26.94%). Education has been shown time and time
again to have a very strong, highly significant impact on individuals’ wage. Lastly, the
correlation between education and private insurance is 26.29%. As stated above; educated
persons are more likely to maintain good health and less likely to qualify for government health
insurance, due to having greater wages on average. Therefore, as a result of safeguarding against
risk; education has a relatively strong degree of correlation with private insurance status. While
some degrees of correlation presented are higher than desirable, they should not be problematic
in the analysis of my regression.
The degree of correlation between age and working hours was found to be 24.26% Thus,
as Americans get older; they are more likely to work greater than average working hours. A
familiar economic interpretation of this phenomenon is that nearing retirement age, individuals
become enlightened in regards to their less than desirable life savings. The Social Security
Administration often deludes individuals into thinking they have input an appropriate amount of
their allotted income towards sustaining a modest standard of living throughout retirement.
However, as retirement nears, these individuals discover their financial needs outweigh social
security benefits provided for them and decide to take on greater working hours in their later
years of employment.
The fit of my model captured 13.15% of the trend occurring in the percent change of
health care expenditure, with a total of 9,562 observations. This goodness-of-fit figure implies
that influential forces not included in this test account for 86.85% of changes in the dependent
variable. The probability that the variables included in my model have a no relationship with
percent change in health care cost is practically non-existent, given an F-score of 259.39. As
previously stated, all coefficients can be interpreted in percent changes to annual healthcare
expenditure, calculated as the sum of out-of-pocket costs and cost of premiums on health
insurance.
My main independent variable, usual working hours per week squared, has a positive
coefficient equal to .0178. This can be interpreted as a one hour increase in working hours
producing a 1.78% rise in annual healthcare costs. This variable has a t-score of 3.66, reaching
well beyond a 95% confidence level; indicating a reliable regression estimate. By way of the t-
score, the confidence interval for working hours per week squared is .00828-.02741; understood
in the same manner as its coefficient. This is a relatively loose range of possible values
(consistent with the t-score) on a 95% confidence interval, both ends of which still contain
positive figures.
The point of interest in my model, working hours per week squared, was not found to be
significant (t-score of -.24). As well, the coefficient calculated in my regression was -.0000179;
interpreted algebraically as a one hour increase in usual working hours will cause healthcare
costs, in terms of percentages, to decrease by -.00145(working hours per week)2 .This
relationship is extremely feeble and contains a negative trajectory, in opposition of my
hypothesis. Moreover, the confidence interval ranges from -.0001326 to .0001035; showing both
a potential positive or negative relationship. Overall, it can be said that the quadratic relationship
between working hours and health care expenditure is weak, displays a negative trend (in
conflict with expectations), and has an 80.9% probability of being unreliable; which is extended
by the inconsistent sign of the confidence interval. Therefore, I conclude that this model failed to
reject the null hypothesis of working hours having a linear impact on health care expenditure.
Age is the next independent variable I would like to discuss; with a coefficient of positive
.0224, marginally stronger than that of working hours per week. This can be interpreted as each
year being added to your life increasing cost of health care by 2.244%. Given a t-score of 16.87,
age has a 95% confidence interval from .0002507-.0003102. Being highly significant, it is
improbable for this variable to not have a real world effect on health care expenditure.
Simply from a glimpse of the t-test comparisons of each type of health insurance status, it
can be inferred that this difference is significant. Indeed, in my model, private insurance status
versus those government insured and uninsured has a t-score of 24.38; the most reliable in my
test. This dependent variable has a coefficient of 1.161, understood as private health insurance,
on average, will result in a 116.12% increase in medical expenses per year; over those with
uninsured and with government insurance. Indicative of its t-score is a narrow confidence
interval between 1.068 and 1.254, making its devastating impact on health care expenditure
apparent.
Individual's’ wages in my regression analysis, as shown in the table above, supplies us
with a coefficient of .005338; taken to mean each dollar added to your wage will result in an
estimated .5338% increase in health care expenditure. An output t-score of 5.54 for this factor
provides us with a .00345-.00723 confidence interval. Inclusion of this variable allows for
comparison between the impacts of the income effect and hours effect; comprehended through
wage and working hours per week.
Education is another important variable in my regression; because of the characteristics
educated persons have in regards to health care. As described previously, those with higher
levels of education are more likely to gain higher wages Moreover, those with greater education
are likely more conscious of their health; thus, are further inclined to utilize health care available
to them.
Lastly, I would like to discuss the effect of health classification, as perceived by
contributors of the CPS. This independent variable is important because it directly relates to the
health of individuals, rather than the cost of their health. The regression delivers a coefficient of
.0974 and a t-score of 5.62 (confidence interval from .0634-.1314). After determining the
variable’s significance (well above that of a 95% confidence level), it can be assumed that an
increase of one on the classification scale (1 being in excellent health, 5 being in poor health)
will increase health care expenses by 9.74%. Health classification was found to have the second
strongest relationship with health care expenditure behind that of being a holder of private
insurance.
The alternative to this variable was an indication of a person’s disability status. Results
for the same regression with these variables being substituted yield a coefficient of .2352 and t-
score of 2.93. While the magnitude of its coefficient is much greater than health classification,
and reasonably so, it has a less reliable t-score. This is a result of the wide range of health care
costs respective to various disabilities; whereas health classification will likely weigh their
disability status, along with health care costs.
Although the results of my primary model failed to show a quadratic relationship
between working hours and health care costs, I was able to support my hypothesis via
substituting my dependent variable, percent change in health care expenditure, with an indicator
of good health. Derived from the health classification variable, good health encapsulates values
one through three (indicating good health), set equal to one; and values of four and five set equal
to zero. The results of this model’s regression resulted in a goodness of fit of 29.31% (compared
to a prior 13.15%) and a high level significance for the overall test. Dependent variables follow
the same trend as in the previous model despite having an opposite sign; whereas those with
good health will typically incur lesser health care expenditures.
Despite a few variables now lacking significance on a 95% confidence level, there are a
number of inferences that can be made regarding the regression’s output. Firstly, education is
shown to have a negative and significant relationship with good health. While there are many
arguments contradicting this result; educated individuals could simply be more inclined to
perceive themselves as less healthy, than their less-educated counterparts. Secondly, and most
importantly, the quadratic variable for working hours now displays a significant, negative
relationship with a person’s health (as stated in my research hypothesis). While the linear
working hours per week remains to have a positive impact on good health. Therefore, given its
parabolic form, there is a point at which working hours will begin to negatively impact the health
of individuals. Taking the derivative of my regression output (0.0017- 2(0.0000165)(working
hours)= ẟ good health/ ẟ working hours) holding the control variables constant, I was able to
determine that the threshold for working hours changing from a health-increasing activity to a
health-diminishing activity lies at 51.5 hours per week.
Discussions & Conclusions
After rigorously testing of my research model, I conclude that usual working hours per
week does not have a quadratic relationship with health care expenditure (as the sum of
insurance premiums and out-of-pocket costs). The independent variable testing for this quadratic
relationship, displayed a negative relationship with health care expenditure, while my hypothesis
states a positive relationship should exist. Despite these contradictory results, the variable was
found to be insignificant (t=-.03, P>|t|=.765). However, the working hours’ variable testing for a
linear relationship with health care expenses in my regression was discovered to have a positive
and significant (t=3.54, P|t|=.000) relationship with my dependent variable. Additionally, this
linear relationship states that a one hour increase in working hours per week will cause a 1.72%
increase in health care expense. While this is enlightening information, it is diluted by exogenous
variables; responsible for 86.70% of changes in the health care costs.
The motivation of this paper was to obtain evidence that being working beyond ordinary
hours per week would have negative impacts on one’s health. In my primary regression, this
relationship was exemplified through the dependent variable health care expenditure; where
those with higher medical expenses will more than likely be less healthy. Although my main
research model did not yield the results I had hoped, a supplementary regression ran under the
same premise supported my hypothesis. In this second regression, my main dependent variable
health care expenditure is substituted with categorical variable for those who classify themselves
as having good health; without making any adaptations to the independent data sets. Through
this complementary regression model, it was found that working hours per week contained both
a positive, linear relationship and a negative, quadratic relationship with good health. As a result
of having a decent income and standard of living; working improves health up until a certain
point. At this threshold, defined at 51.5 working hours per week, working begins to have a
deteriorating effect on a person’s health. This is an interesting phenomenon that occurs; which
support the main rationale of my research hypothesis. I created one more model based from my
primary regression; using a categorical variable for those working greater than 51.5 hours per
week in place of my main independent variable. The results tell us that those who are
overworked, on average, incur 32% more medical expenses.
Although my holistic research hypothesis was found to be empirically unsupported, I
conclude the theory behind my motivation to be valid. Simply by viewing U.S. consumption
trends, it can be determined that demand for goods and services is steadily on the rise. In other
words, citizens are experiencing greater a greater desire for newly developed goods and services;
especially in the realm of technology. Granted this may be an apparent observation; it does not
recognize what pressure this puts on citizens to afford such luxuries. Due to individuals having
little control over their wages in the short term, funds for this climbing consumption are
maintained through additional working hours.
Research similar to my own referenced in the introduction of my paper identify a number
of criteria that influence health in regards to labor; including work interventions, work-life
conflict, working significantly more hours than desirable, and one that blatantly assumes a
relationship between working hours and health of individuals. This last study mentioned, by
Nagashima, strives to recognize the point in which working hours begins to have a profound
impact on health; interpreted in a comparable form to the threshold I calculated at 51.5 working
hours per week. Nagashima’s study was conducted in a factory setting in Japan; which
determined their threshold to be at 60 working hours per week. This figure is significantly above
my own; showing a more profound effect on health (rather than health care expenditure), cultural
differences, and demographic differences due to the study’s factory setting.
Because my research conveys working hours’ effect on health care costs rather than
health itself, the relationship should differ in interpretation than other studies; however, is still
relevant to the reinforcing theory. Supported by the Grossman Model, my results reveal a trade-
off consumers must make concerning working hours and health care consumption. Those who
work low hours per week will have poorer health and lower medical expenses on average; due to
lower income, thus lesser demand for health care even when suffering from an illness.
Individuals who work more will have higher income allowing for greater consumption in health
care, but also increasing risk factors related to work-life conflict. Accepting 51.5 working hours
per week as the point of inflection with health care expenditure, it should be determined that
working above this threshold will negatively impact the health of individuals; and thus, people
should be discouraged from doing so.
The United States is a country that thrives upon an honest day’s work. However, those
seem to be getting longer as society grows more advanced. According to the Economic Policy
Institute, “The average worker worked 1,868 hours in 2007, an increase of 181 hours from the
1979 work year of 1,687 hours,” a 10.7% increase over 38 years. Furthermore, at 22%, the
increase in annual working hours was greater for the lower-fifth percentile (U.S. wage
distribution) than the middle-fifth percentile of earners (10.9%). While those with more lucrative
wage experienced moderate annual wage growth from 1979-2007, the bottom 60% of Americans
derived annual wage growth as much from increasing working hours, as they did from real
hourly wage increases. Subsequently, inputting the results of my regression in with this trend
displays a problematic phenomenon occurring in the lower class in regards to American culture
and heavily inflated costs of health care.
The “crowding out” of U.S. health insurance is causing prices for the service to increase
exponentially, due to healthy individuals bearing the risk of being uninsured. Therefore, the
statistical significance of working hours on health care costs found in my regression; combined
with the escalating health care prices and lackluster lower/middle class wage growth, causes poor
conditions for many Americans. Testing for elasticity between working hours and health care
expenses, holding the other variables in my regression constant, tells us that a 1% increase in
working hours will increase health care expenditure by .375%. Given a 22% increase in annual
working hours between 1979-2007; the calculation states working hours has caused an average
8.25% incline in health care costs for the lower-fifth class of workers, over that same 38 year
time span. I would argue that the trend in Unites States’ working class culture puts upwards
pressure on health care inflation as well. Dedicating more time towards work, and less towards
other utilities for time (specifically health related activities), individuals are relying upon
prescription medication and procedures reactive to health conditions; rather than being proactive
in regards to health status.
The Affordable Care Act (ACA) entering healthier, risk-tolerant individuals to the
markets should in time allow health insurance agencies to reassess risk of the average policy-
holders. Improving the distribution of health in the health insurance market will eventually
alleviate positive pressure on prices caused by a concentration of high-risk individuals. There are
a large number of factors driving up health care prices; including supply of doctors,
technological use in facilities, and increasing demand for medical care. However, normalization
of the health insurance market, accomplished via the ACA, theoretically has a beneficial impact
on affordability of health care; especially amongst lower-earning individuals.
The previous paragraph addresses the most current progression made towards
normalization of the health insurance industry. Regardless of the long-term impact of ACA on
insurance pricing, the problem is still eminent amongst society; whereas few people are
concerned with trends in working hours. This is likely so do to working hours being a choice for
individuals, allowing for greater levels of consumption; often times necessary for sustaining a
preferable standard of living. Due to a conflict of interest, we cannot expect society to resolve the
issue independently. Thus, employers should be held accountable for monitoring weekly
working hours, in order to cultivate a healthier workforce; in order to benefit both the worker and
the company in which they work for. According to Maslow’s hierarchy of needs, improving the
physiological conditions in workers increases their motivation; which in turn will convert into
productivity, and then into greater profits for the company. Expending the endowment point of
working hours’ effect on good health calculated earlier; employers should limit individuals’
weekly working hours to no more than 51.5. A superior policy implementation would be for
consumers to elect a greater involvement in preventing poor health and associated conditions;
rather than relying on pharmaceutical drugs and unnecessary medical procedures while partaking
in risky behavior.
Despite achieving respectable results in regards to answering my research question, there
are still a number of limitations I encountered while doing this research. One limitation was the
time constraint of completing this study. Although this Senior Seminar course enabled adequate
time to fully investigate our research questions, there is much more thought that could have been
put into my research. Another limitation of this study is the goodness of fit, equal to 13.30%. It is
apparent other factors are affecting health care expenditure; inclusion of which would benefit
both the efficacy and insights of the research. These exogenous variables would include
information such as disability status, race, medical conditions (consisting of asthma, diabetes,
obesity, and other such factors), number of drinks per week, smoking status, marital status,
number of children in a family, family income, and many other factors influencing the risk
pertaining to an individual.
To conclude, my paper provides empirical evidence that working hours has a positive
correlation with health care expenditure; and also has a concave-down relationship with good
health. Although there are many other studies related to the trends in U.S. working hours, my
research presents tangible figures for relevant factors’ influences on health care costs and
demographic trends in health care expenditure. Furthermore, through a supplementary
regression, I was able to calculate the break-even point of working hours’ affect towards good
health; which can be used as a benchmark for the optimal point of working hours’ per week.

Contenu connexe

Tendances

The Pursuit of Happiness - Australian Doctor - July 3 2015
The Pursuit of Happiness - Australian Doctor - July 3 2015The Pursuit of Happiness - Australian Doctor - July 3 2015
The Pursuit of Happiness - Australian Doctor - July 3 2015
John Kron
 
The prevalence and correlates of low back pain in adults
The prevalence and correlates of low back pain in adultsThe prevalence and correlates of low back pain in adults
The prevalence and correlates of low back pain in adults
Younis I Munshi
 
Educational level, sex and church affiliation on health seeking
Educational level, sex and church affiliation on health seeking Educational level, sex and church affiliation on health seeking
Educational level, sex and church affiliation on health seeking
Alexander Decker
 

Tendances (14)

Health and Wellbeing after Deportation: The Roles of Socio-Demographic Variab...
Health and Wellbeing after Deportation: The Roles of Socio-Demographic Variab...Health and Wellbeing after Deportation: The Roles of Socio-Demographic Variab...
Health and Wellbeing after Deportation: The Roles of Socio-Demographic Variab...
 
marital status_work_&_ subjective health
marital status_work_&_ subjective healthmarital status_work_&_ subjective health
marital status_work_&_ subjective health
 
Reducing Tobacco Use Among Adolescents Using Social Cognitive Theory and Soci...
Reducing Tobacco Use Among Adolescents Using Social Cognitive Theory and Soci...Reducing Tobacco Use Among Adolescents Using Social Cognitive Theory and Soci...
Reducing Tobacco Use Among Adolescents Using Social Cognitive Theory and Soci...
 
What Makes for a Healthy Workplace
What Makes for a Healthy WorkplaceWhat Makes for a Healthy Workplace
What Makes for a Healthy Workplace
 
The Pursuit of Happiness - Australian Doctor - July 3 2015
The Pursuit of Happiness - Australian Doctor - July 3 2015The Pursuit of Happiness - Australian Doctor - July 3 2015
The Pursuit of Happiness - Australian Doctor - July 3 2015
 
Factors associated with persistent disease in women with asthma
Factors associated with persistent disease in women with asthmaFactors associated with persistent disease in women with asthma
Factors associated with persistent disease in women with asthma
 
Worklife balance of women doctors in coimbatore
Worklife balance of women doctors in coimbatoreWorklife balance of women doctors in coimbatore
Worklife balance of women doctors in coimbatore
 
Corporate Health and Wellness: Combating Stress in the Workplace
Corporate Health and Wellness: Combating Stress in the WorkplaceCorporate Health and Wellness: Combating Stress in the Workplace
Corporate Health and Wellness: Combating Stress in the Workplace
 
The prevalence and correlates of low back pain in adults
The prevalence and correlates of low back pain in adultsThe prevalence and correlates of low back pain in adults
The prevalence and correlates of low back pain in adults
 
DNP Capstone Project Sample
DNP Capstone Project SampleDNP Capstone Project Sample
DNP Capstone Project Sample
 
Convención NAOS The Leadership
Convención NAOS The LeadershipConvención NAOS The Leadership
Convención NAOS The Leadership
 
MedicalResearch.com Top Interviews September 5 2015
MedicalResearch.com Top Interviews September 5 2015MedicalResearch.com Top Interviews September 5 2015
MedicalResearch.com Top Interviews September 5 2015
 
Managment of stress among unikl mestech students (Research)
Managment of stress among unikl mestech students (Research)Managment of stress among unikl mestech students (Research)
Managment of stress among unikl mestech students (Research)
 
Educational level, sex and church affiliation on health seeking
Educational level, sex and church affiliation on health seeking Educational level, sex and church affiliation on health seeking
Educational level, sex and church affiliation on health seeking
 

En vedette

Portafolio dagnóstico
Portafolio dagnósticoPortafolio dagnóstico
Portafolio dagnóstico
alexpapag
 

En vedette (14)

Tics
TicsTics
Tics
 
Blosuors
BlosuorsBlosuors
Blosuors
 
Said cv
Said cvSaid cv
Said cv
 
Resumen clases
Resumen clasesResumen clases
Resumen clases
 
Breiner moreno
Breiner morenoBreiner moreno
Breiner moreno
 
Ecology Ottawa Research Report
Ecology Ottawa Research ReportEcology Ottawa Research Report
Ecology Ottawa Research Report
 
Los climas toni
Los climas toniLos climas toni
Los climas toni
 
Portafolio dagnóstico
Portafolio dagnósticoPortafolio dagnóstico
Portafolio dagnóstico
 
Trabajo espartero joaquin y javier
Trabajo espartero joaquin y javierTrabajo espartero joaquin y javier
Trabajo espartero joaquin y javier
 
RIO DON-
RIO DON- RIO DON-
RIO DON-
 
La españa del XIX (I)
La españa del XIX (I)La españa del XIX (I)
La españa del XIX (I)
 
BCCDC Watershed Metagenomics Project: Viral Biomarkers 2013
BCCDC Watershed Metagenomics Project: Viral Biomarkers 2013BCCDC Watershed Metagenomics Project: Viral Biomarkers 2013
BCCDC Watershed Metagenomics Project: Viral Biomarkers 2013
 
chemical storage
chemical storagechemical storage
chemical storage
 
Fauna de la Antártida
Fauna de la AntártidaFauna de la Antártida
Fauna de la Antártida
 

Similaire à The Effect of Working Hours on Health Care Expenditure in the United States

Raul Quiroz - Research Paper Business Communications
Raul Quiroz - Research Paper Business CommunicationsRaul Quiroz - Research Paper Business Communications
Raul Quiroz - Research Paper Business Communications
Raul Quiroz
 
Exploring occupational balance in adults in Sweden
Exploring occupational balance in adults in SwedenExploring occupational balance in adults in Sweden
Exploring occupational balance in adults in Sweden
Grupo OT5
 
Technical And Business Of Entrepreneurship
Technical And Business Of EntrepreneurshipTechnical And Business Of Entrepreneurship
Technical And Business Of Entrepreneurship
Diane Allen
 
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docxRESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
ronak56
 
The Impact of Burnout syndrome on Nurse Workers .docx
The Impact of Burnout syndrome on Nurse Workers               .docxThe Impact of Burnout syndrome on Nurse Workers               .docx
The Impact of Burnout syndrome on Nurse Workers .docx
rtodd33
 
The Impact of Burnout syndrome on Nurse Workers .docx
The Impact of Burnout syndrome on Nurse Workers               .docxThe Impact of Burnout syndrome on Nurse Workers               .docx
The Impact of Burnout syndrome on Nurse Workers .docx
arnoldmeredith47041
 
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY .docx
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY               .docxRunning head PHYSICAL ACTIVITY AND SELF-EFFICACY               .docx
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY .docx
charisellington63520
 
Predictors of sickness absence and presenteeism: Does the pattern differ by a...
Predictors of sickness absence and presenteeism: Does the pattern differ by a...Predictors of sickness absence and presenteeism: Does the pattern differ by a...
Predictors of sickness absence and presenteeism: Does the pattern differ by a...
Palkansaajien tutkimuslaitos
 
Stress and Healthcare Workers Productivity at Lexington Medical
Stress and Healthcare Workers Productivity at Lexington Medical Stress and Healthcare Workers Productivity at Lexington Medical
Stress and Healthcare Workers Productivity at Lexington Medical
blazelaj2
 
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docxActivity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
daniahendric
 
Respond to the Main post bellow, in one or more of the follo
Respond to the Main post bellow, in one or more of the folloRespond to the Main post bellow, in one or more of the follo
Respond to the Main post bellow, in one or more of the follo
mickietanger
 
Organizational Behavior - Work Life Balance
Organizational Behavior - Work Life Balance Organizational Behavior - Work Life Balance
Organizational Behavior - Work Life Balance
shanelle_sumitra
 
Quantitative Analysis Template !Instructions When analyzing.docx
Quantitative Analysis Template !Instructions When analyzing.docxQuantitative Analysis Template !Instructions When analyzing.docx
Quantitative Analysis Template !Instructions When analyzing.docx
amrit47
 
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
ijtsrd
 

Similaire à The Effect of Working Hours on Health Care Expenditure in the United States (20)

Raul Quiroz - Research Paper Business Communications
Raul Quiroz - Research Paper Business CommunicationsRaul Quiroz - Research Paper Business Communications
Raul Quiroz - Research Paper Business Communications
 
Exploring occupational balance in adults in Sweden
Exploring occupational balance in adults in SwedenExploring occupational balance in adults in Sweden
Exploring occupational balance in adults in Sweden
 
Technical And Business Of Entrepreneurship
Technical And Business Of EntrepreneurshipTechnical And Business Of Entrepreneurship
Technical And Business Of Entrepreneurship
 
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docxRESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
RESEARCH ARTICLE Open AccessExperiences of reduced work ho.docx
 
The Impact of Burnout syndrome on Nurse Workers .docx
The Impact of Burnout syndrome on Nurse Workers               .docxThe Impact of Burnout syndrome on Nurse Workers               .docx
The Impact of Burnout syndrome on Nurse Workers .docx
 
The Impact of Burnout syndrome on Nurse Workers .docx
The Impact of Burnout syndrome on Nurse Workers               .docxThe Impact of Burnout syndrome on Nurse Workers               .docx
The Impact of Burnout syndrome on Nurse Workers .docx
 
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY .docx
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY               .docxRunning head PHYSICAL ACTIVITY AND SELF-EFFICACY               .docx
Running head PHYSICAL ACTIVITY AND SELF-EFFICACY .docx
 
Predictors of sickness absence and presenteeism: Does the pattern differ by a...
Predictors of sickness absence and presenteeism: Does the pattern differ by a...Predictors of sickness absence and presenteeism: Does the pattern differ by a...
Predictors of sickness absence and presenteeism: Does the pattern differ by a...
 
Stress and Healthcare Workers Productivity at Lexington Medical
Stress and Healthcare Workers Productivity at Lexington Medical Stress and Healthcare Workers Productivity at Lexington Medical
Stress and Healthcare Workers Productivity at Lexington Medical
 
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docxActivity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
Activity Week 2 SWOT PowerPointDue Week 2 and worth 200 points.docx
 
2017 SHRM SIOP Science of HR - Employee Well-being
2017 SHRM SIOP Science of HR - Employee Well-being2017 SHRM SIOP Science of HR - Employee Well-being
2017 SHRM SIOP Science of HR - Employee Well-being
 
Respond to the Main post bellow, in one or more of the follo
Respond to the Main post bellow, in one or more of the folloRespond to the Main post bellow, in one or more of the follo
Respond to the Main post bellow, in one or more of the follo
 
The Changing Organization of Work and the Safety and Health of Working People
The Changing Organization of Work and the Safety and Health of Working PeopleThe Changing Organization of Work and the Safety and Health of Working People
The Changing Organization of Work and the Safety and Health of Working People
 
A Corporate Wellness Program And Nursing Home Employees Health
A Corporate Wellness Program And Nursing Home Employees  HealthA Corporate Wellness Program And Nursing Home Employees  Health
A Corporate Wellness Program And Nursing Home Employees Health
 
Organizational Behavior - Work Life Balance
Organizational Behavior - Work Life Balance Organizational Behavior - Work Life Balance
Organizational Behavior - Work Life Balance
 
Unemployment and self-assessed health: Evidence from panel data
Unemployment and self-assessed health: Evidence from panel dataUnemployment and self-assessed health: Evidence from panel data
Unemployment and self-assessed health: Evidence from panel data
 
Gender Difference in Response to Preventative Health Care
Gender Difference in Response to Preventative Health CareGender Difference in Response to Preventative Health Care
Gender Difference in Response to Preventative Health Care
 
Quantitative Analysis Template !Instructions When analyzing.docx
Quantitative Analysis Template !Instructions When analyzing.docxQuantitative Analysis Template !Instructions When analyzing.docx
Quantitative Analysis Template !Instructions When analyzing.docx
 
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
Knowledge, Attitude and Practices of Expectant Mothers on Physical Activity A...
 
Worksite Wellness
Worksite WellnessWorksite Wellness
Worksite Wellness
 

The Effect of Working Hours on Health Care Expenditure in the United States

  • 1. The Effect of Working Hours on Health Care Expenditure in the United States Year after year, the United States is becoming known as a country of workaholics. In 2015, employees in the U.S. reached a 40-year high for unused vacation days. Oxford University calculated this phenomenon to equate approximately to 169 million days, or the $52.4 billion in lost benefits (Kasperkevic 2015). Furthermore; for the United States being the most powerful and influential world leader, we do not mandate employers to provide paid vacation time. Only one in four Americans receive paid vacation or holiday, which trails well behind a bulk of the developed world. I find this trend to be regressive in terms of societal norm; and intend to support this through the discovery of adverse effects on health care costs related to individuals’ working hours (Mohn 2013). The goal of my study is to investigate, and possibly quantify, the degree to which working hours affects one’s health (using health care expenditure as a proxy). An Economist article titled Working Hours: Get a Life from 2013 took a look into the relationship between hours worked per year and productivity. To help identify its readers with the purpose of their article; C.W. and A.J.K.D. referenced essays from Bertand Russell (“In Praise of Idleness” 1932) and John Mayard Keynes (“Economic possibilities for our grandchildren” 1930). The objective of these essays was to emphasize a utopian lifestyle that would arise from a rise in technology, and more importantly, a lessening demand for labor. Despite a vast rise in technology since the 1930’s, we have yet to cultivate an economy in which labor is not the driving force. A simple regression between hours worked per year (2080 hours/year equaling 40 hours/week) and GDP per hour worked (productivity) shows a strong negative correlation between these two variables confirming that the marginal productivity is a diminishing function of working hours. A simple empirical example comes from comparison of Greek and German workers. Despite the fact that Greeks are hard-working individuals averaging about 2,000 working hours per year, Germans who work a mere 1,400 hours each year achieve a productivity measure that is 70% higher than those of Greek workers (C.W. 2013). In addition to the negative impact of working hours has on productivity, a recent study recent study published by Lancet Medical Journal showed health related risks of long working hours. The aforementioned study displayed a positive relationship between working hours greater than 40 per week and cardiovascular issues. The study found that people working just an hour extra a week, between 41 and 48 hours per week, have a 10% higher risk of stroke or other cardiac events; while people working 49 to 54 hours per week have a 27% higher risk of stroke (Mohney 2015). According to the above study, the health risk only rises with additional working hours. Individuals who worked 55 hours a week or greater were shown to be 33% more susceptible to strokes, even after monitoring other risk factors such as smoking and alcohol consumption; and a
  • 2. 13% increased risk for coronary heart disease or heart attack (Mohney 2015). The statistical inferences in this study are similar to what I hope to find in my research; relating to long working hours and general adverse effects it has on an individual’s health. The main research question of my study is to investigate the relationship between health care expenditure and the number of hours individuals work per week. My research hypothesis is that an increase in weekly working hours over a certain threshold positively correlates with health care costs. This should be consistent with the findings of previous studies; demonstrating that an increase in working hours has a negative impact on health, and thus, positively relating to health care expenditure. Literature Review There is a substantial ongoing research focusing on the correlation between work and health, but only few investigate this link for workers of the United States. A recent study conducted using a sample of Korean workers illustrates the impact of gender on self-rated health, in regards to hours worked per week (Seong-Sik 2015). The study consisted of men and women similar in age distribution, who were asked to rate their health on the 5-point SRH scale (ranging from “very good” to “very poor” health). Working hours were placed into ranges; including 20- 35h (less than standard working hours), 36-40h (standard working hours), 41-52h (overtime within legally permitted working hours excluding weekend work), 53-68h (overtime within legally permitted working hours including weekend work), and 69h or more (legally prohibited). The results of this study showed women to have a higher proportion of poor health. Approximately 75% of men and women worked more than standard working hours, but more women worked longer than 5 days per week. This is likely due to women receiving significantly lower wages than their male counterparts and needing additional income. Of the individuals in the study, women were found to hold whiter collar jobs, and were also found to be less educated than the participating males (which would inherently contribute to the gender wage differential). Seong-Sik (2015) stated that working hours in the South Korea’s labor force are strongly driven by factors of demand (the vast majority of individuals working over 40 hours per week). In regards to women, the statistical findings were realistically supported by the idea of women balancing their work and family lives. Attempting to support a family and take care of a family would in doubt bear more burden on women compared to men. In sum, this study found that higher levels of working hours contributed to poor health. In C.W.’s Working hours: Get a life (2013); South Korea’s labor force is graphically shown to work an additional 400 hours per person per year in 2012, compared to an approximate 1,000 working hours differential in 1990. Workplace interventions have shown to be impactful to individual’s health and consequently, productivity (Vander Klink, Blonk, Schene, Van Dijk 2001). Specifically, workplace interventions in the forms of cognitive-behavioral and multi-model interventions were found to have a significant impact on work-related stress. Outcomes that had the most significant impact on work-related stress (deriving from workplace interventions) include: complaints,
  • 3. psychological resources, responses, and perceived quality of work life. Although this study has shown us the advantageous of taking a workplace intervention, I suspect the same advantage to a greater magnitude will occur with a simple reduction in working hours of an individual. The importance of this research is to exemplify a technique in order to counteract the negative effects of our working culture. Another study by Fein & Skinner (2015) looked at major occupational groups, along with gender, to identify pathways through which work hours impact health (Fein & Skinner 2015). The theoretical framework used in their research was based around the idea that working hours increases work-life conflict, and hence leads to health outcomes. Previous research by Skinner estimates the relationship strength between work-life-conflict and health to be between 0.23 and 0.40, depending on particular health outcomes. The relationship between work hours and health in said study estimates a weaker relationship than work-life-conflict and health. It was shown that the coefficient for working hours’ effect on health was 0.15 for psychological health, and 0.06 for physical health. The research suggests that work hours were only associated with negative health outcomes when the hours did not fit or interfered with workers’ other commitments or activities. For women, this effect was shown to be greater, similar to Seong-Sik (2015). The average woman may struggle to balance a work-life and family obligations; to a greater extent than their male counterparts (Fein 2015). Bell, Otterbach, and Sousa-Poza (2012) have recently conducted a study to investigate the impact of the difference between actual and desired work hours on self-perceived health outcomes. Their purpose was to indicate the consequences of employment policies and shed light onto the discrepancy between actual and desired work hours. This research shows that many individuals are overemployed on average, working 4+ hours per week than desired. One notable insight found was that overemployed employees were generally less satisfied with their own health than unconstrained full-time workers. Several other significant inferences included stronger negative coefficients related to overemployment (compared to unconstrained and underemployed workers) in relation to health satisfaction and self-assessed health, over all weekly working hour intervals; ranging from (Bell 2012). In another study, Nagashima, et. al (2007) accepted that there was a statistically significant relationship between working hours and an individual’s health. Thus, their research aimed at identifying a threshold in which number of hours worked per month had a profound impact on mental and physical fatigue. They developed their results through the distribution of questionnaires to 843 (720 used in the study) male factory workers, which I suspect could have influenced their findings. Nagashima’s results stated that working greater than 260 hours per month would begin to have an adverse impact on the Self-Rating Depression Scale (SDS). Measures involved in this scale include decreased vitality, general fatigue, physical disorders, irritability, decreased willingness to work, anxiety, depressive feelings, and chronic tiredness; in which general fatigue and tiredness, depressive feelings, irritability and decreased vitality were
  • 4. most prevalent. In order to minimize these effects, males working in the sample factory should work less than 260 hours per month (Nagashima 2007). Conceptual Framework There are also a number of additional implications of high working hours on health care expenditure pertaining to the theory of demand for health capital. Basic economics tells us that individuals use their income to maximize a preference function (utility). In regards to demand for health capital, the preference function is as follows: Ut[utility]=U[Ht (health), Xt (other commodities)], in a given period t. Since we know humans are not immortal, it is so that every individual has a depreciation factor on their health. The health function for an individual in a future period is dependent on his/her current health, minus the given depreciation consistent with aging/risk averseness, plus investment into health-benefiting activities. The Grossman Model states that individuals have the choice to invest in either medical care (their health) or other market goods and services. Beyond this trade-of between health and market goods is how any person spends their time. The time constraint of this individual is as follows: TT (total time) = TW (time at work)+ TH (time spent on health) + TX (time spent on other commodities) + TL (time lost due to illness). My research is focused around two effects that will result from the aforementioned model, which are highly related to the workplace culture in the United States; the first of which is the wage effect. This effect states that the higher an individual’s wage the less incentive he/she has to invest time into their health. Such as, an hour of working out would be more costly to a CEO of a large corporation compared to a low wage, general labor employee. Secondly, the Grossman Model also suggests the hours effect. This effect states that the more time you spend working, the less time you have to invest elsewhere (specifically into time spent on health). For example, a full-time worker will have less time available for consumption than a part-time worker, and thus, will likely invest less time into their health. I believe that Americans spend too much time at work; thus suffer additional time lost from illness or sickness. Furthermore, U.S. obesity rates continue to rise year over year along with diabetes, while physical inactivity stays level. Despite the lackluster investment of time spent on health, the United States still pays second most in the world on health care. This phenomenon is a product of the U.S. labor market, which encourages workers to seek extraordinary working hours. Therefore, I intend my research to empirically prove the positive relationship between working hours and health care expenditure (as a proxy to represent individuals’ health). This would exemplify my belief in how our society has shifted to dedicate more time towards work, consequently devaluing time spent on health and increasing the cost of health care in the United States. This research could potentially make significant inferences into the cost-benefit analysis individuals’ use when evaluating their demand for health capital;
  • 5. incentivizing people to invest more time into better-health producing activities, rather than working more hours simply to incur additional medical expenses. Data & Descriptive Statistics The data used to create this report are derived from the Current Population Survey (CPS). The CPS is one of the oldest, largest surveys jointly sponsored by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics. Data composing the Current Population Survey are gathered on a monthly basis from citizens of the United States. The sample size of this expansive survey is approximately 90,000 individuals over the age of fifteen; excluding persons in the Armed Forces along with persons institutionalized in prisons, hospitals, nursing homes, and other such institutions. Individuals in the sample are contacted year-over-year, with a 90% response rate. The government is willing to sponsor this survey because of its immense use throughout research regarding the United States. Beyond the benefits of making this information available public for research, the U.S. government derives a number of macroeconomic variables from this data, including various unemployment rates (U1-U6) and the labor force participation rate. Furthermore, the CPS is a valuable investment for the government in its ability to give information regarding broad, demographic information; such as age, race, education, income, marital status, and so on. Having such a large sample size, the Current Population Survey serves as the most important survey in capturing statistics unbiased across the U.S. population. Table 1 displays the proportion of individuals pertaining to each health insurance classification. Those with health care coverage make up 86.24% of the United States’ population, while only 13.76% do not have any form of health insurance. Excess observations not accounted for in this distribution were due to survey participants not supplying an answer, or possibly having a recent change in coverage. Despite having 80.09% of the population being privately insured, the remaining 6.15% of insurance being provided by the government; the United States still experiences difficulty with “crowding out” of the private-insurance industry. Health care services in the United States have experienced sharper inclines in pricing than seen throughout any other industry, with the exception of higher education institutions. The proportion of individuals holding private insurance increased 4.2% in 2014; this trend will eventually put downwards pressure on healthcare prices in the U.S. market. However, there are numerous forces that drive said prices upwards including such as implementation of technology, advanced surgical methods, demand for assistive living, and so on. Exacerbating the issue is an expanding portion of elderly people in this country, who are the primary demographic for health care services.
  • 6. In its entirety, the Current Population Survey consists of 90,430 observations; however, to better serve my research, I excluded all individuals who did not provide an input for average weekly hours worked. Dropping said data will allow me to more accurately identify the effect of working hours on health care expenditure. Removing said data leaves 9,563 observations for analyses and overall, a more accurate portrayal of the findings. Part time workers have been classified as those who work 27 hours per week or less. 27 working hours per week was set as the threshold between full and part time employment because 28 working hours per week is the lowest point at which a worker in the United States can qualify as being fully employed by an employer. Full time workers are defined as those who regularly work 28 or more hours per week; while those who regularly work 45 hours per week or greater are considered overemployed. The figure to the left shows that the vast majority of individuals (67.4%) in this sample are full time workers. Part time workers represent 17.2% of this data set. As shown in the figure, overemployed individuals constitute a modest percentage of workers (15.4%). Although overemployed individuals are generally considered full time workers, separate classification is necessary in order to produce information related to the adverse effects of strenuous labor driven from working an extraordinary number of hours.
  • 7. Descriptive statistics in Table 2 are provided for the key independent variables. The set of control variables includes, age, race, gender, disabilities, years of education, number of kids in a family under six, weekly earnings, type of health insurance, and working hours. Demographic information revealed in the descriptive statistics show a majority (64.6%) of respondents being white, followed by 9.8% black and the remaining 16.2% of the population being Hispanic (while 9.2% are unspecified). The average age of those surveyed was 40 years old; with a minimum age of 15 years and maximum age 64 years. The gender distribution in the analytical sample is split evenly between male and females. Only 3.7% of sample observations reported having various forms of disability. 57.3% of observations in the sample classified as married individuals. Average years of education and works hours per week are 14.1 years and 36.5 hours, respectively. Gross weekly earnings average $575.46 per week.
  • 8. In Table 3, I have broken down the Descriptive statistics by working classification (identifying sample size, mean, and standard deviation; vertically). There are several trends that can be recognized within the different categories of insured/uninsured individuals. First, is the trend of age to increase usual working hours; likely as a result of the need to save for retirement, satisfy medical expenses, reassess financial position, and so on. The next two variables look at race’s effect on working classification. Looking at the white population, it can be seen that they consist of 66.26%, 61.45%, and 76.6% of part time, full time, and overemployed; respectively. Black individuals on the other hand, display values of 8.9%, 11.06%, and 5.5% in the same respect. From this data, it can be inferred that the white population is most likely to be overemployed, next to be part time, and least of the three to be full time workers. The black population makes up a much less significant portion of the population than whites, hence the lower percentages for worker classification. Information from this data set tells us that most blacks are full time workers, followed by part time, and are least likely to be overemployed. Males display a trend of having working much more so than women. While men are 37.7%, 49.7%, and 67.7% to work part time, full time, and be overemployed; while women comprise the remainder of each category. While this is the case now, women have been progressing towards greater working hours over the years; evening this scale over time. Those with a disability are significantly less likely to work full time and obviously less so to be overemployed. A good percentage of individuals with disabilities are unemployed; however, those observations have been removed from my test. Marital status has a strong impact on working classification as well; increasingly likely to work a greater number of hours per week. Education also trends upwards with working hours; as they are more encouraged due to higher wages and greater value to a company. Gross weekly earnings are a data set that obviously trends upwards with worker classification. Interestingly, the gap between overemployment and full employment is greater than that of the full time and part time differential; implying people are paid exponentially more, on average, if they work more. The trend in insurance status tells us that as one works more hours, they are more likely to have private insurance; and definitely less so to be on government insurance as they earn too much to qualify. Lastly, giving us a glimpse at my research question are descriptive statistics for health care expenditure; in regards to working classification. Those who work a greater number of hours, as suggested, will incur greater medical expenses. Each progression in working classification brings a greater change to health care costs, despite a larger difference in working hours from part to full time employment, compared to full time and overemployment. As mentioned previously in this paper; observations with higher incomes and working hours have implications on the time individuals spend on health activities. Thus, the higher these
  • 9. inputs are, on average, the less time these individuals spend on health related activities will be as well. This in turn, will result in higher healthcare costs. The following page contains quadratic regressions for both working hours’ and weekly earnings’ effects on health care expenditure, perfectly exemplifying these theories. From the above visualizations, it can be seen that working hours and weekly earnings have distinct quadratic relationships with health-care expenditure. The weekly earnings relationship with health care expenditure consists of a narrower confidence interval over the whole range of studied earnings, meaning weekly earnings has a more consistent relationship with health care expenditure than working hours. Both positive associations are compatible with our theoretical expectations coming from the Grossman Model. The regression analysis should shed light whether these trends hold after controlling for a rich set of various individual level factors. Research Design & Results My research hypothesis states there is a quadratic relationship between working hours and health care costs. Working less hours per week increases the likelihood of poverty leading to poorer health conditions; contrarily, working more hours dissuades individuals from spending time on health related activities, while increasing disposable income (inevitably increasing money spent on one’s health) simultaneously. In the above empirical model, Yi is annual healthcare expenditure, which can be regarded as a proxy for an individual’s health status, Hi is individuals’ usual weekly working hours and Xi represents a set of control variables such as. There are a number of other factors that influence the health care expenditure of an individual outside of my control and main independent variables, which compose 𝜀 𝜀, the fit of the model not captured by the variables I have selected to include.
  • 10. I have chosen health care expenditure (Yi), calculated as the sum of premiums and out-of- pocket medical expenses, to be my dependent variable testing my hypothesis. Self-perceived health classification of individuals was an alternative to my favored dependent variable. Despite it being a relevant variable for my test, I suspected the range of values was not wide enough to provide meaningful interpretation. Furthermore, portraying my intended outcome in quantifiable U.S. dollars as presented with health care expenditure will deliver a greater, more accurate impact than factor data. Additionally, I transformed the dependent variable by taking the natural log of it. As a result, parameters of the independent variables can be interpreted in the semi- elasticity form, a percent change in health care expenditure due to a one unit change in an independent variable. Average working hours per week (Hi) is my main independent variable that thoroughly answers my research question; the effect of working hours on health care expenditure. Although, since my hypothesis suggests that working hours has a quadratic relationship with health care costs, I have included a separate squared data set of those variables, in addition to the original data set for usual working hours per week. The most inherently relevant control variable (Xi) I had conjectured to include was average age of survey participants. No matter how much money or time we invest into our health, its status will still depreciate, hence why age is an obvious factor affecting health care expenditure of individuals. The older an individual is, greater health care expenses can be expected. Furthermore, values for the age data set were found to be consistently correlated with the other independent variables; most notably found in working hours per week (both in linear and quadratic form) and wage. Another independent variable I have configured is the categorical status of individuals’ health insurance; whereas those with private insurance have a value set equal to one and other persons’ (holding government insurance or uninsured) are set equal to zero. Through the use of t- table comparisons between costs of health care of the three different categories, I concluded having private insurance significantly increases an individual’s expected health care expenditure. Moreover, the regression’s results will give enlightening inferences into the magnitude and significance of the disproportioned state of the health insurance industry (referring to the “crowding out” theory) On average, people with private insurance spend $2,217 per year on healthcare, compared with $487 and $301 per year for those with government insurance and those uninsured; respectively. This is the case as a result of the “crowding out” of the health insurance industry; where individuals only purchase insurance if they have a forceful need to do so. Such a large difference between costs of medical care among these categories, along with the results of the t-tests indicate that privately held insurance has a strong, positive, and significant impact on health care costs. Health classification, as perceived by the survey participants, has also been introduced into my model. This variable contains data ranging from one to five; one being indicative of
  • 11. excellent health, while a value of five implies poor health conditions. Thus, the relationship between health care classification and health care expenditure should result in a positive relationship; meaning a greater value (indicating poor health) will cause greater health care costs. Since my model is theoretically based around the health status of individuals in regards to working hours, it is only reasonable to include a variable directly associated with their health status as well. Despite the possibility of misperceived health care statuses, I found this health- related variable to be the most suitable for the purpose of my research. Wage is a data set I calculated by dividing gross weekly earnings by average usual hours worked per week. While this variable may not be completely accurate to the person’s wage, it is expressive of how wage is determined in such a context; consequently affecting the health care costs of individuals. The consumer choice theory states that an increase in income increases consumption of certain goods (normal goods); including those related to the person’s health (research shows that healthcare can be considered as a normal good). Thus, this variable should have a positive relationship with health care expenditure. I have included wage as a control variable in my model because it enables me to distinguish the income effect (greater disposable income to be broadly distributed) from the hour’s effect. The hour’s effect conversely states that working more hours will limit an individual’s excess time, therefore reducing demand for spending time on health related activities. An important theory supporting my model is that the United States is unique to a culture that supports working an abundance of hours per week. A consequence of this social construction is the trend in the health of U.S. citizens; as can be seen in increasing obesity rates, rises in mental health related illnesses, stress, and so on. Lastly, education is another important variable to control for; because it has been proven to have a strong, positive correlation with the health of an individual. Educated people are more likely to partake in less risky behavior that will negatively impact their health, along with taking proactive measures to combat illness or improve their health status. However, since my dependent variable is related to health care expenditure instead of individuals’ actual health, the coefficient may be suppressed. Since educated people are more likely to be proactive in regards to their health, the effect on health care expenditure contains negative pressure; whereas taking precautions to your health will prevent future ailments, thus, health care costs. On the other hand, those with upper level education on average, have private health insurance. As mentioned previously, this puts significant upwards pressure on health care expenditure. Below is a chart exhibiting the degrees of correlation between the variables used in my test. As you can see, there are no extraordinarily high levels of correlation between any of the variables, with the exception of working hours per week and working hours per week squared. The most consistently elevated degrees of correlation lie with age, as previously stated. As one
  • 12. gets older, they are more likely to work, have consistency of sustaining themselves above the need for government insurance, while having demand for health insurance, consistent changes in health classification, and have greater experience; thus higher wages. Another high degree of correlation exists between education, wage, and private insurance. A correlation of 16.08% between private insurance and wage is predictable. As one becomes an increasingly impactful member of a company or industry, their benefits are to increase as well (including health insurance packages). Wage and education were found to have a relatively strong degree of correlation (26.94%). Education has been shown time and time again to have a very strong, highly significant impact on individuals’ wage. Lastly, the correlation between education and private insurance is 26.29%. As stated above; educated persons are more likely to maintain good health and less likely to qualify for government health insurance, due to having greater wages on average. Therefore, as a result of safeguarding against risk; education has a relatively strong degree of correlation with private insurance status. While some degrees of correlation presented are higher than desirable, they should not be problematic in the analysis of my regression. The degree of correlation between age and working hours was found to be 24.26% Thus, as Americans get older; they are more likely to work greater than average working hours. A familiar economic interpretation of this phenomenon is that nearing retirement age, individuals become enlightened in regards to their less than desirable life savings. The Social Security Administration often deludes individuals into thinking they have input an appropriate amount of their allotted income towards sustaining a modest standard of living throughout retirement. However, as retirement nears, these individuals discover their financial needs outweigh social security benefits provided for them and decide to take on greater working hours in their later years of employment.
  • 13. The fit of my model captured 13.15% of the trend occurring in the percent change of health care expenditure, with a total of 9,562 observations. This goodness-of-fit figure implies that influential forces not included in this test account for 86.85% of changes in the dependent variable. The probability that the variables included in my model have a no relationship with percent change in health care cost is practically non-existent, given an F-score of 259.39. As previously stated, all coefficients can be interpreted in percent changes to annual healthcare expenditure, calculated as the sum of out-of-pocket costs and cost of premiums on health insurance. My main independent variable, usual working hours per week squared, has a positive coefficient equal to .0178. This can be interpreted as a one hour increase in working hours producing a 1.78% rise in annual healthcare costs. This variable has a t-score of 3.66, reaching well beyond a 95% confidence level; indicating a reliable regression estimate. By way of the t- score, the confidence interval for working hours per week squared is .00828-.02741; understood in the same manner as its coefficient. This is a relatively loose range of possible values (consistent with the t-score) on a 95% confidence interval, both ends of which still contain positive figures. The point of interest in my model, working hours per week squared, was not found to be significant (t-score of -.24). As well, the coefficient calculated in my regression was -.0000179; interpreted algebraically as a one hour increase in usual working hours will cause healthcare costs, in terms of percentages, to decrease by -.00145(working hours per week)2 .This relationship is extremely feeble and contains a negative trajectory, in opposition of my hypothesis. Moreover, the confidence interval ranges from -.0001326 to .0001035; showing both a potential positive or negative relationship. Overall, it can be said that the quadratic relationship between working hours and health care expenditure is weak, displays a negative trend (in conflict with expectations), and has an 80.9% probability of being unreliable; which is extended by the inconsistent sign of the confidence interval. Therefore, I conclude that this model failed to reject the null hypothesis of working hours having a linear impact on health care expenditure.
  • 14. Age is the next independent variable I would like to discuss; with a coefficient of positive .0224, marginally stronger than that of working hours per week. This can be interpreted as each year being added to your life increasing cost of health care by 2.244%. Given a t-score of 16.87, age has a 95% confidence interval from .0002507-.0003102. Being highly significant, it is improbable for this variable to not have a real world effect on health care expenditure. Simply from a glimpse of the t-test comparisons of each type of health insurance status, it can be inferred that this difference is significant. Indeed, in my model, private insurance status versus those government insured and uninsured has a t-score of 24.38; the most reliable in my test. This dependent variable has a coefficient of 1.161, understood as private health insurance, on average, will result in a 116.12% increase in medical expenses per year; over those with uninsured and with government insurance. Indicative of its t-score is a narrow confidence interval between 1.068 and 1.254, making its devastating impact on health care expenditure apparent. Individual's’ wages in my regression analysis, as shown in the table above, supplies us with a coefficient of .005338; taken to mean each dollar added to your wage will result in an estimated .5338% increase in health care expenditure. An output t-score of 5.54 for this factor provides us with a .00345-.00723 confidence interval. Inclusion of this variable allows for comparison between the impacts of the income effect and hours effect; comprehended through wage and working hours per week. Education is another important variable in my regression; because of the characteristics educated persons have in regards to health care. As described previously, those with higher levels of education are more likely to gain higher wages Moreover, those with greater education are likely more conscious of their health; thus, are further inclined to utilize health care available to them. Lastly, I would like to discuss the effect of health classification, as perceived by contributors of the CPS. This independent variable is important because it directly relates to the health of individuals, rather than the cost of their health. The regression delivers a coefficient of .0974 and a t-score of 5.62 (confidence interval from .0634-.1314). After determining the variable’s significance (well above that of a 95% confidence level), it can be assumed that an increase of one on the classification scale (1 being in excellent health, 5 being in poor health) will increase health care expenses by 9.74%. Health classification was found to have the second strongest relationship with health care expenditure behind that of being a holder of private insurance. The alternative to this variable was an indication of a person’s disability status. Results for the same regression with these variables being substituted yield a coefficient of .2352 and t- score of 2.93. While the magnitude of its coefficient is much greater than health classification, and reasonably so, it has a less reliable t-score. This is a result of the wide range of health care
  • 15. costs respective to various disabilities; whereas health classification will likely weigh their disability status, along with health care costs. Although the results of my primary model failed to show a quadratic relationship between working hours and health care costs, I was able to support my hypothesis via substituting my dependent variable, percent change in health care expenditure, with an indicator of good health. Derived from the health classification variable, good health encapsulates values one through three (indicating good health), set equal to one; and values of four and five set equal to zero. The results of this model’s regression resulted in a goodness of fit of 29.31% (compared to a prior 13.15%) and a high level significance for the overall test. Dependent variables follow the same trend as in the previous model despite having an opposite sign; whereas those with good health will typically incur lesser health care expenditures. Despite a few variables now lacking significance on a 95% confidence level, there are a number of inferences that can be made regarding the regression’s output. Firstly, education is shown to have a negative and significant relationship with good health. While there are many arguments contradicting this result; educated individuals could simply be more inclined to perceive themselves as less healthy, than their less-educated counterparts. Secondly, and most importantly, the quadratic variable for working hours now displays a significant, negative relationship with a person’s health (as stated in my research hypothesis). While the linear working hours per week remains to have a positive impact on good health. Therefore, given its parabolic form, there is a point at which working hours will begin to negatively impact the health of individuals. Taking the derivative of my regression output (0.0017- 2(0.0000165)(working hours)= ẟ good health/ ẟ working hours) holding the control variables constant, I was able to determine that the threshold for working hours changing from a health-increasing activity to a health-diminishing activity lies at 51.5 hours per week.
  • 16. Discussions & Conclusions After rigorously testing of my research model, I conclude that usual working hours per week does not have a quadratic relationship with health care expenditure (as the sum of insurance premiums and out-of-pocket costs). The independent variable testing for this quadratic relationship, displayed a negative relationship with health care expenditure, while my hypothesis states a positive relationship should exist. Despite these contradictory results, the variable was found to be insignificant (t=-.03, P>|t|=.765). However, the working hours’ variable testing for a linear relationship with health care expenses in my regression was discovered to have a positive and significant (t=3.54, P|t|=.000) relationship with my dependent variable. Additionally, this linear relationship states that a one hour increase in working hours per week will cause a 1.72% increase in health care expense. While this is enlightening information, it is diluted by exogenous variables; responsible for 86.70% of changes in the health care costs. The motivation of this paper was to obtain evidence that being working beyond ordinary hours per week would have negative impacts on one’s health. In my primary regression, this relationship was exemplified through the dependent variable health care expenditure; where those with higher medical expenses will more than likely be less healthy. Although my main research model did not yield the results I had hoped, a supplementary regression ran under the same premise supported my hypothesis. In this second regression, my main dependent variable health care expenditure is substituted with categorical variable for those who classify themselves as having good health; without making any adaptations to the independent data sets. Through this complementary regression model, it was found that working hours per week contained both a positive, linear relationship and a negative, quadratic relationship with good health. As a result of having a decent income and standard of living; working improves health up until a certain point. At this threshold, defined at 51.5 working hours per week, working begins to have a deteriorating effect on a person’s health. This is an interesting phenomenon that occurs; which support the main rationale of my research hypothesis. I created one more model based from my primary regression; using a categorical variable for those working greater than 51.5 hours per week in place of my main independent variable. The results tell us that those who are overworked, on average, incur 32% more medical expenses. Although my holistic research hypothesis was found to be empirically unsupported, I conclude the theory behind my motivation to be valid. Simply by viewing U.S. consumption trends, it can be determined that demand for goods and services is steadily on the rise. In other words, citizens are experiencing greater a greater desire for newly developed goods and services; especially in the realm of technology. Granted this may be an apparent observation; it does not recognize what pressure this puts on citizens to afford such luxuries. Due to individuals having little control over their wages in the short term, funds for this climbing consumption are maintained through additional working hours.
  • 17. Research similar to my own referenced in the introduction of my paper identify a number of criteria that influence health in regards to labor; including work interventions, work-life conflict, working significantly more hours than desirable, and one that blatantly assumes a relationship between working hours and health of individuals. This last study mentioned, by Nagashima, strives to recognize the point in which working hours begins to have a profound impact on health; interpreted in a comparable form to the threshold I calculated at 51.5 working hours per week. Nagashima’s study was conducted in a factory setting in Japan; which determined their threshold to be at 60 working hours per week. This figure is significantly above my own; showing a more profound effect on health (rather than health care expenditure), cultural differences, and demographic differences due to the study’s factory setting. Because my research conveys working hours’ effect on health care costs rather than health itself, the relationship should differ in interpretation than other studies; however, is still relevant to the reinforcing theory. Supported by the Grossman Model, my results reveal a trade- off consumers must make concerning working hours and health care consumption. Those who work low hours per week will have poorer health and lower medical expenses on average; due to lower income, thus lesser demand for health care even when suffering from an illness. Individuals who work more will have higher income allowing for greater consumption in health care, but also increasing risk factors related to work-life conflict. Accepting 51.5 working hours per week as the point of inflection with health care expenditure, it should be determined that working above this threshold will negatively impact the health of individuals; and thus, people should be discouraged from doing so. The United States is a country that thrives upon an honest day’s work. However, those seem to be getting longer as society grows more advanced. According to the Economic Policy Institute, “The average worker worked 1,868 hours in 2007, an increase of 181 hours from the 1979 work year of 1,687 hours,” a 10.7% increase over 38 years. Furthermore, at 22%, the increase in annual working hours was greater for the lower-fifth percentile (U.S. wage distribution) than the middle-fifth percentile of earners (10.9%). While those with more lucrative wage experienced moderate annual wage growth from 1979-2007, the bottom 60% of Americans derived annual wage growth as much from increasing working hours, as they did from real hourly wage increases. Subsequently, inputting the results of my regression in with this trend displays a problematic phenomenon occurring in the lower class in regards to American culture and heavily inflated costs of health care. The “crowding out” of U.S. health insurance is causing prices for the service to increase exponentially, due to healthy individuals bearing the risk of being uninsured. Therefore, the statistical significance of working hours on health care costs found in my regression; combined with the escalating health care prices and lackluster lower/middle class wage growth, causes poor conditions for many Americans. Testing for elasticity between working hours and health care expenses, holding the other variables in my regression constant, tells us that a 1% increase in working hours will increase health care expenditure by .375%. Given a 22% increase in annual
  • 18. working hours between 1979-2007; the calculation states working hours has caused an average 8.25% incline in health care costs for the lower-fifth class of workers, over that same 38 year time span. I would argue that the trend in Unites States’ working class culture puts upwards pressure on health care inflation as well. Dedicating more time towards work, and less towards other utilities for time (specifically health related activities), individuals are relying upon prescription medication and procedures reactive to health conditions; rather than being proactive in regards to health status. The Affordable Care Act (ACA) entering healthier, risk-tolerant individuals to the markets should in time allow health insurance agencies to reassess risk of the average policy- holders. Improving the distribution of health in the health insurance market will eventually alleviate positive pressure on prices caused by a concentration of high-risk individuals. There are a large number of factors driving up health care prices; including supply of doctors, technological use in facilities, and increasing demand for medical care. However, normalization of the health insurance market, accomplished via the ACA, theoretically has a beneficial impact on affordability of health care; especially amongst lower-earning individuals. The previous paragraph addresses the most current progression made towards normalization of the health insurance industry. Regardless of the long-term impact of ACA on insurance pricing, the problem is still eminent amongst society; whereas few people are concerned with trends in working hours. This is likely so do to working hours being a choice for individuals, allowing for greater levels of consumption; often times necessary for sustaining a preferable standard of living. Due to a conflict of interest, we cannot expect society to resolve the issue independently. Thus, employers should be held accountable for monitoring weekly working hours, in order to cultivate a healthier workforce; in order to benefit both the worker and the company in which they work for. According to Maslow’s hierarchy of needs, improving the physiological conditions in workers increases their motivation; which in turn will convert into productivity, and then into greater profits for the company. Expending the endowment point of working hours’ effect on good health calculated earlier; employers should limit individuals’ weekly working hours to no more than 51.5. A superior policy implementation would be for consumers to elect a greater involvement in preventing poor health and associated conditions; rather than relying on pharmaceutical drugs and unnecessary medical procedures while partaking in risky behavior. Despite achieving respectable results in regards to answering my research question, there are still a number of limitations I encountered while doing this research. One limitation was the time constraint of completing this study. Although this Senior Seminar course enabled adequate time to fully investigate our research questions, there is much more thought that could have been put into my research. Another limitation of this study is the goodness of fit, equal to 13.30%. It is apparent other factors are affecting health care expenditure; inclusion of which would benefit both the efficacy and insights of the research. These exogenous variables would include information such as disability status, race, medical conditions (consisting of asthma, diabetes,
  • 19. obesity, and other such factors), number of drinks per week, smoking status, marital status, number of children in a family, family income, and many other factors influencing the risk pertaining to an individual. To conclude, my paper provides empirical evidence that working hours has a positive correlation with health care expenditure; and also has a concave-down relationship with good health. Although there are many other studies related to the trends in U.S. working hours, my research presents tangible figures for relevant factors’ influences on health care costs and demographic trends in health care expenditure. Furthermore, through a supplementary regression, I was able to calculate the break-even point of working hours’ affect towards good health; which can be used as a benchmark for the optimal point of working hours’ per week.