Descriptive statistics for the variables in the data

Data in this essay has been categorized into three parts, the dependent variable
(exhaustion score), the independent variable (whether the employee works at home or
office) and control variable (age, tenure, gross wage, children, commute, gender,
relationship and volunteer). There are 5109 observations in the total, detailed
explanation of each variable has been provided in the appendix. 
This essay separates the sample into two groups based on 1) whether they are into
treatment or control group; 2) whether they are married; 3) whether they have the
children and 4) whether the commute time is higher than average or no. Table 1
reports the results. This essay has found telecommuting can reduce the exhaustion
score dramatically. Interestingly, this essay finds if the employee is married, their
exhaustion scores are lower than those who do not have. Additionally, there is slightly
different when taking the children effect into consideration based on the average
number of two subsamples. Furthermore, this essay also finds that gender effect may
cause fewer differences in the score since only 0.05 can be witnessed from the
average number. The commute is another factor that can influence the exhaustion
score, result suggesting that if the employee’s commute time is higher than average,
then he/she will have higher exhaustion score. 
Table 1 also expresses the consistent conclusion, suggesting that when employees
are chosen telecommuting, their exhaustion score will consequently decrease, since
the density of zero scores is higher than the group that employees work in the office.

Table 1. Exhaustion Score on Different Characteristics

YES NO
Panel A. Whether they are in treatment or control group
Observation Mean Observation Mean
Exhaustion Score 2886 6.628 2223 10.52
Panel B. Whether they are married
Exhaustion Score 858 8.35 4,251 8.31
Panel C. Whether they have the children
Exhaustion Score 1,365 6.91 3,744 8.83
 Panel D. whether the commute time is higher than average

Density
.3

.2

.1

0
Exhaustion Score 2,418 8.83 2,691 7.86

0 1

0 10 20 30 40 0 10 20 30 40
exhaustion score
Graphs by whether in treatment group

Table 2 presents the correlation coefficient among each variable at 5% significant

level. This essay finds the relationship between the exhaustion score and expgroup is
negative and significant, suggesting that working at home can reduce the sense of
exhaustion. Apart from this relationship, this essay also finds the positive relationship
between age and children on exhaustion.
Table 2. Correlation Matrix

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Ln(exhaustion) (1) 1
expgroup (2) -0.2388* 1
Age (3) 0.0363* 0.1124* 1
Tenure (4) -0.00720 0.0552* 0.5898* 1
Gross Wage (5) 0.0142 -0.0442* 0.4022* 0.5503* 1
Children (6) 0.0425* -0.2236* 0.4085* 0.2990* 0.2099* 1
Commute (7) -0.0484* -0.1253* -0.0215 0.0903* 0.1384* 0.0357* 1
Men (8) -0.0330* -0.0489* 0.0152 -0.2180* -0.1044* -0.1055* -0.2795* 1
Married (9) -0.1005* -0.0964* 0.4117* 0.3651* 0.3180* 0.7440* 0.0291* -0.1323* 1
Volunteer (10) -0.1172* -0.0452* 0.0121 -0.0672* -0.0830* 0.1052* 0.00550 -0.0609* 0.1198* 1
Hypothesis Design 

This essay proposes the hypothesis as:

H1. Telecommuting can reduce the employee’s exhaustion 
Table 1 reports the descriptive statistic by separate the sample into two groups,
considering whether the employees are telecommuting or work in the office. Overall,
all variables are significantly differenced at 1% level between the treatment and
control group. If the employees are telecommuting, then they are in the treatment
group, and the rest is in the control group. We find a significant difference for the
dependent variable, suggest that telecommuting working environment can reduce the
exhaustion score. In particular, if the employee works at home, the exhaustion score
will reduce 3.89.
Furthermore, statistics of the control variables also indicate the differences. For
example, employees who choose to work at home in this experiment are older, having
longer tenure, and less gross wage. In addition, they are also having the family
(children) and takes a long time for working. Hence, the negative motivation for
office working will be generated when combining these characteristics together.

Table 3. Variable Difference

Treatment Group Control Group Difference
Mean Mean t-statistic
Exhaustion Score 6.63 10.52 17.80***
Age 25.14 24.35 -8.09***
Tenure 30.95 28.36 -3.95***
Gross Wage 3.08 3.16 3.16***
Children 0.09 0.26 16.39***
Commute 103.45 119.17 9.03***
Men 0.46 0.51 3.50***
Married 0.23 0.32 6.92***
Volunteer 0.86 0.89 3.24***

Observation 2,886 2,223

Regression and Analysis Testing

In this section, we go further in analysing how different variables have impacted

the dependent variable exhaustion. Two models will be formed, the first equation is
only affected by numerical independent variables, expressed as follows:
Model 1:
Exhaustion=β 0 + β 1 Ages+ β 2 tenure+ β 3 gross wage+ β 4 commute+ μ

Result for the regression analysis can be found in the Appendix. It can be seen
that the results of treatment group using model 1 obtained R-Squared value is 0.0304
meaning that the accuracy of the model is 3.04% and with F-Test is 22.57. This model
is positively influenced by the age and tenure variables. The value of exhaustion level
in the condition of the independent variable 0 is 0.4625984. Meanwhile, the control
group using model 1 obtained R-Squared is 0.0218 meaning that the accurancy is
2.18%

Model 2:
Exhaustion=β 0 + β 1 Ages+ β 2 tenure+ β 3 gross wage+ β 4 children+ β5 commute+ β 6 men+ β7 married + β8 vol

Model 2 has a higher R-Squared value where 11.97% of variation is explained by

this model. The increase in R-Squared compared to model 1 is due to the variables
age, grosswage, children, and age. Because of the higher accurancy value, Model 2 is
a better regression model to describe the level of employee exhaustion in the
treatment data. Based on the F value <0.05, it can be interpreted that all independent
variables simultaneously have a significant influence on the dependent variable. In
addition, the control group using Model 2 is also obtained the higher R-Squared at
5.75%. Thus, the value of F value 0.05
The level of exhaustion of workers in the treatment group is lower than the
control group, this means that there is a positive impact of treatment carried out.
Based on the result of the regression analysis of variables that affect the level of
employee exhaustion in the treatment group positively are age, children, and man.
Whereas in the control group the variables that positively affect employee exhaustion
are tenure, grosswage, children, bedroom, and commute.
Blibliography

Bloom, N., Liang, J., Roberts, J. and Ying, Z. (2013). Does Working From Home
Work? Evidence from A Chinese Experiment. [online] Nbloom.people.stanford.edu.
Available at: https://nbloom.people.stanford.edu/sites/g/files/sbiybj4746/f/wfh.pdf
[Accessed 2 Dec. 2019].
Appendix

Table 4. Regression Table for Treatment Group

Model (1) Model (2)
Age (year) 0.285*** 0.358***
(0.044) (0.048)
Tenure at Ctrip (month) -0.004 -0.015*
(0.007) (0.007)
Log Grosswage 0.211 0.469
(0.578) (0.584)
Daily commute in minutes -0.011*** -0.005*
(0.002) (0.002)
Whether have children 2.172***
(0.571)
Whether have independen~m -0.393
(0.600)
Whether is male 0.249
(0.271)
Whether married -3.646***
(0.388)
Whether volunteer to wo~m -4.549***
(0.378)
Observations 2886 2886
R-squared 0.030 0.120
F 22.570 43.458
df m 4.000 9.000
df r 2881.000 2876.000
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001
 Table 5. Regression Table for Control Group
Model (1) Model (2)
Age (year) -0.259*** -0.228**
(0.068) (0.076)
Tenure at Ctrip (month) 0.006 0.037*
(0.013) (0.016)
Log Grosswage -1.414 -1.923
(1.008) (1.007)
Daily commute in minutes -0.012*** -0.022***
(0.003) (0.003)
Whether have children 1.825*
(0.896)
Whether have independen~m 10.895***
(1.520)
Whether is male -1.467***
(0.421)
Whether married -3.921***
(0.918)
Whether volunteer to wo~m -0.618
(0.594
Observations 2223 2223
R-squared 0.022 0.057
F 12.383 14.966
df m 4.000 9.000
df r 2218.000 2213.000
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001