HU Journal of Public Economics Discussion


Need to write a Referee Report about a Journal “The impact of COVID-19 on student experiences and expectations: Evidence from a survey”. I have a pdf version and an example about the referee report. It’s about causal inference. The guideline and the article are attached

1 attachmentsSlide 1 of 1attachment_1attachment_1.slider-slide > img { width: 100%; display: block; }
.slider-slide > img:focus { margin: auto; }

Unformatted Attachment Preview

Referee Report: American Economic Journal: Applied Economics
Manuscript #: 2020-0635
Title: The Ties That Bind Us: Social Networks and Productivity in the Factory
This paper seeks to examine the effects of caste-based social networks on productivity in a
garment assembly setting where production is team-based. To answer this question, the
authors exploit daily variation in own caste composition within the production line driven by
worker absences and reassignment of workers across production lines. The authors report
two main findings: First, an increase in own caste workers is associated with an increase in
individual productivity. Second, an assembly line’s productivity (i.e. average and minimum
productivity of workers in the line) is positively correlated with the homogeneity of its caste
This paper tackles an important question in the personnel economics literature. The authors
have carefully conducted the empirical analysis to address the question of interest. I do have
several major comments that I hope will improve the paper’s interpretation and contribution
to the literature.
Major Comments:
Background and Context
o While the authors provide a detailed explanation of the production technology
and compensation scheme I am not completely convinced about the following
statement (last line of page 9):
Given the production externalities in the assembly line, the performance of coworkers in an assembly line can impact the earnings of a worker.
This statement is crucial to the theoretical framework and empirical analysis
of this paper. I believe there is no controversy about the production externality.
Yet, the authors indicate that workers are compensated by a fixed salary plus
a promotion chance (by grade) and overtime work. The latter two are
determined likely by the line supervisor who has good knowledge of the
capability and productivity of each worker in his/her production line. Thus, I
do not understand why a slow coworker in the production line will directly
impact a worker’s earnings if salary is fixed pay. Alternatively, the authors can
provide direct evidence to support this statement using payroll data but I
believe they mention that this is not possible.
I think a more convincing argument can be made if the authors assume that
there is a non-pecuniary component to worker’s utility. For instance, a slow
coworker may induce extra effort by workers in the same line to make up for
the efficiency loss caused by their slow coworker.
o The study takes data from two factories – one export-oriented and one
domestic. Given that there might be drastic differences in terms of
management and production technologies between the two factories (export
factory should use more advanced technologies than the domestic factory) it
looks more reasonable to look at them separately. If one factory has few
observations for an independent analysis it might be worth dropping them and
only look at one factory.
o The authors state that prior to this study the factory only recorded line level
productivity and not individual worker productivity. This is a concern especially
for external validity as the period of analysis coincides with the period of
productivity recording. Knowing that their individual productivity is recorded
workers may respond to caste networks differently than usual. The authors
should address this issue when interpreting the findings.
o Given the large number of migrant workers, it seems worthwhile to explore
same hometown networks. If the productivity effect is driven by social
networks relevant for job referrals, then hometown networks should also be a
relevant proxy.
o The main measure of productivity is efficiency. Yet, I wonder if there is data on
the number of operations or tasks performed by each individual worker. For
instance, skilled workers may perform multiple tasks and the number of tasks
may also depend on the productivity of coworkers. If there is a slow coworker
in the production line attaching zippers to a garment then the supervisor may
ask a fast worker to also attach zippers after finishing her own task.
Theoretical Framework
o Based on the second paragraph of this section (page 16), supervisor incentives
play an important role in incentivizing workers. The authors may want to
include a brief discussion about connections between supervisors and workers
and how it may influence productivity. If the authors can find evidence of
exogenous variation between assignment of workers to line supervisors, I
believe a potentially interesting direction is to examine how social networks
between supervisors and workers affect worker productivity (i.e. Bandiera,
Barankay, Rasul (2009)).
o While the authors argue that networks can help increase productivity without
having to increase wages I wonder if that is possible in a competitive labor
market when firms are competing with others to hire employees. That is, if
networks induce workers to exert more effort and, accordingly, higher cost of
working then in the long run if that cost is not compensated for the workers
would eventually switch their jobs.
Methodology and Results
o The authors show that the worker’s caste and the production line is exogenous.
Given the results of this paper it seems that supervisors may want to
strategically group workers of the same caste to increase productivity. Is there
a reason why they cannot do this other than because the management does
not have information about the workers’ caste?
o Personally, I think the magnitude of the estimate seems overly large to be
attributed to pressure from social networks: a 1 percentage point increase in
own caste composition leads to a 30 percent increase in productivity. What
might be more convincing is if the authors can conduct the analysis with
subsamples of certain tasks or include task-specific fixed effects (I am not
questioning the SAM approach to efficiency but tasks do vary to a great degree
in terms of complexity and skill levels which might be responsible for the large
variation across days).
o Instead of month fixed effect the authors can also use week fixed effects to
better address time variation.
o I cannot find the results claimed by the author in the first paragraph of page
Indeed our results show that the effect of… is larger for least performing
workers as compared to the average productivity worker.
o The authors use the finding from Table 9 to support the mechanism. I think a
more relevant interaction term for arguing job referral network effect is
Proportion own caste X referee in same production line. The reason is if
referee is in different production line than there is no directly associated
pecuniary or non-pecuniary benefits/costs.
o An alternative interpretation of the findings is that same caste production lines
are more likely to coordinate and specialize in certain tasks leading to an
increase in the overall efficiency. For example, suppose task A requires high
skill and high effort and task B requires low skill and low effort. In a nonhomogeneous group, task assignment may be random but in a homogenous
group the high skill worker conducts task A despite the high cost and low skill
worker conducts task B.
o In the last paragraph the authors argue that if the mechanism is altruism then
productivity dispersion within the line is declining with caste homogeneity. I
do not understand why this is true. Maybe more explanation is needed for this
Bandiera, O., Barankay, I. and Rasul, I. (2009) “Social Connections and Incentives in the
Workplace: Evidence from Personnel Data,” Econometrica, 77(4), 1047-1094.

Purchase answer to see full

Explanation & Answer:
3 pages

Journal of Public Economics

User generated content is uploaded by users for the purposes of learning and should be used following Studypool’s honor code & terms of service.