Stata Code
Budget: $30 – $250 USD
In WFH.zip, you will find several Stata data files and a do file called ’Tables2014.do’, which will help you replicate the findings of “Does Working from Home Work? Evidence from a Chinese Experiment”
(Quarterly Journal of Economics, 2015) by Nicholas A. Bloom, James Liang, John Roberts,
Zhichun Jenny Ying. The paper can be replicated by running this file, but you will be asked to make adjustments to the analysis. Provide separate completed do file and document that answers the questions in complete sentences.
1. Table 2 Replication
(a) Replicate Columns 1 and 2 from Table 2 (titled The Performance Impact of WFH). In
1-2 sentences, how are these estimation strategies different?
(b) Create a variable called pre_experiment_mean that is equal to the mean of perform1
before the experiment starts for each person. This value will be constant for each
person. This is a bit tricky, be careful.
Replicate Column 2 from Table 2 again, but this time, control for the pre-treatment
mean of each person by adding it into the regression. Report the point estimate,
rounded 3 digits.
(c) Briefly comment on the difference between these three approaches, how is this
related to Footnote 17?
2. Table 2 Extenstion
(a) Estimate the diff-in-diff estimator (Column 1) across one measures of heterogeneity.
For example, estimate the treatment effect for workers who are above or below the
median productivity level as in Figure X. Report the point estimate and standard
error, rounded 3 digits. Explain the form of heterogeneity you are exploring and
why it may be interesting.
(b) Interpret the estimate. How does the treatment effect vary across your chosen
measure of heterogeneity? Do not forget to address both the estimate and its associ-
ated standard error.
3. Figure 6 Replication
(a) Replicate Figure VI (titled Performance of Treatment and Control Employees: Phone
Calls) It is okay to not provide the on-graph text labels, but be sure to label the axis and
give the figure a title and format the figure so it can be easily understood.
(b) Instead of graphing the treatment and control values, graph the difference be-
tween treatment and control over time (the average treatment effect). Does this
effect change over time? Why might it?. Be sure to format the figure so that it can
be easily undestood.
(Quarterly Journal of Economics, 2015) by Nicholas A. Bloom, James Liang, John Roberts,
Zhichun Jenny Ying. The paper can be replicated by running this file, but you will be asked to make adjustments to the analysis. Provide separate completed do file and document that answers the questions in complete sentences.
1. Table 2 Replication
(a) Replicate Columns 1 and 2 from Table 2 (titled The Performance Impact of WFH). In
1-2 sentences, how are these estimation strategies different?
(b) Create a variable called pre_experiment_mean that is equal to the mean of perform1
before the experiment starts for each person. This value will be constant for each
person. This is a bit tricky, be careful.
Replicate Column 2 from Table 2 again, but this time, control for the pre-treatment
mean of each person by adding it into the regression. Report the point estimate,
rounded 3 digits.
(c) Briefly comment on the difference between these three approaches, how is this
related to Footnote 17?
2. Table 2 Extenstion
(a) Estimate the diff-in-diff estimator (Column 1) across one measures of heterogeneity.
For example, estimate the treatment effect for workers who are above or below the
median productivity level as in Figure X. Report the point estimate and standard
error, rounded 3 digits. Explain the form of heterogeneity you are exploring and
why it may be interesting.
(b) Interpret the estimate. How does the treatment effect vary across your chosen
measure of heterogeneity? Do not forget to address both the estimate and its associ-
ated standard error.
3. Figure 6 Replication
(a) Replicate Figure VI (titled Performance of Treatment and Control Employees: Phone
Calls) It is okay to not provide the on-graph text labels, but be sure to label the axis and
give the figure a title and format the figure so it can be easily understood.
(b) Instead of graphing the treatment and control values, graph the difference be-
tween treatment and control over time (the average treatment effect). Does this
effect change over time? Why might it?. Be sure to format the figure so that it can
be easily undestood.