This question is based on David Card’s (1990) study of the effect of immigration on the employment status of natives. If

[answerhappygod]

This question is based on David Card’s (1990) study of the
effect of immigration on the employment status of natives. If you
recall from class, the Mareil Boatlift was an unexpected shock to
Miami’s immigrant population that occurred when Castro unexpectedly
allowed a large number of Cuban’s to immigrate to the US in 1980.
The episode saw an increase in Miami’s labour force of about 7
percent – a huge year-over-year increase. The idea was that, if
immigrants affect the labour market out comes of natives, it should
be detectable here.
(a)
In order to study the impact of immigration on the labour market
outcomes of locals, economists often use a regression set up such
as this:
Yit=α+δImmigit+ηitYit=α+δImmigit+ηit
Where YitYit is the average employment outcome (either
wages or unemployment) for native workers in city ii in
time tt. The variable ImmigitImmigit is the fraction
of immigrants in the city. Explain why an OLS regression
of YitYit on ImmigitImmigit may not yield the
causal effect of immigrants on native outcomes. Your explanation
should be specific to this context.
(b)
The boatlift episode that David Card examines can potentially
overcome these problems. Suppose we are interested in the following
equation:
lnwageit=α+δDit+ηit(1)(1)lnwageit=α+δDit+ηit
Where lnwage is the log hourly wage of a worker. The
variable DitDit is a treatment indicator. In this case,
the treatment is belonging to Miami after 1980, when the
immigration shock occurred. What are the assumptions we need to
make about ηitηit in order for a difference in difference
strategy to estimate a causal effect of the Boatlift?
(c)
In order to carry out a difference-in-difference estimator in
this case, we have to perform some data manipulation. The data
contains information of 5 cities, Miami and four other comparison
cities. Follow the instructions
The data for this question: Get data
(d)
Estimate a difference-in-difference specification for the
relationship in (1). Include a specification without controls and
one with controls for yearsch, exper, and female.
Use the data frame df2 (ie, include only the years
1977-1983)
(e)
The difference-in-differences assumptions cannot ever be
verified. However, one can assess plausibility by examining
“pre-trends”. To do this, one plots the mean outcome variable over
time for the treatment group and the controls and compares them.
Using the data frame df ie, using all
years, plot the trend in lnwageitlnwageit over
time. There is a sketch of the code below. Interpret the output
from the figure: Does the difference-in-differences common trend
assumption seem plausible in this case? Overall, what is your
assessment of the impact of the boatlift?