Table
1
.
Regressions for per Capita Growth
1
Answer the following questions based on the regression table below.
(
a
)
The table shows the results of running a regression where the dependent
(
or left hand side
)
variable is a country
’
s growth rate and the independent
(
or right hand side variables
)
are various factors that could affect the growth rate. The data used are cross
-
sectional
(
no time variation
)
.
I want to focus on the variable PRIM
6
0
,
which is a measure of education
(
primary school enrollment in
1
9
6
0
)
.
According to regression
(
1
)
,
what is the average effect of a one
-
unit increase in PRIM
6
0
on a country
’
s growth rate?
(
b
)
The numbers in parentheses are standard errors. Using this knowledge, write down which variables in regression
(
1
)
are statistically signifcant according to our
‘
rule of thumb
’
we learned in class.
(
c
)
Assume this is simply a cross
-
section regression with no use of difference
-
in differences
(
obviously
)
,
IV
,
or any other of the techniques we discussed in class. Tell me a story about why we should be careful about interpreting the coefficient on PRIM
6
0
as casual. I
’
m looking for a plausible story that would undermine a naive interpretation of the coefficient.
(a) The average effect of a one-unit increase in PRIM60 on a country's growth rate is 0.03.
(b) The variables in regression (1) that are statistically significant according to our 'rule of thumb' are PRIM60, GOVSHARE, and INF.
(c) We should be careful about interpreting the coefficient on PRIM60 as causal because there may be omitted variable bias. For example, it is possible that countries with higher levels of primary school enrollment in 1960 also have other characteristics that are positively correlated with economic growth. In this case, the coefficient on PRIM60 may be capturing the effects of these other unobserved variables rather than the true causal effect of primary school enrollment on growth. Additionally, it is possible that there are reverse causality issues, where countries with higher economic growth rates are more likely to invest in education. Therefore, a naive interpretation of the coefficient on PRIM60 as causal may lead to incorrect policy conclusions.