Doing Economics



Yüklə 1,91 Mb.
Pdf görüntüsü
səhifə16/94
tarix25.06.2022
ölçüsü1,91 Mb.
#62274
1   ...   12   13   14   15   16   17   18   19   ...   94
Doing Economics What You Should Have Learned in Grad School But

2.5.6 Tables
Before closing out this section, I would like to discuss some miscellaneous
pieces of advice regarding tables of empirical results. In no particular order:
• The titles of your tables should be self-explanatory: “OLS Results
for the Effect of Participation in Contract Farming on Household
Income,” or “OLS Results for the Effect of Years of Education on
Wage by Gender.” The titles should thus tell us what is being
estimated (e.g., OLS), what the relationship of interest is (i.e., the
effect of participation in education on wage), and what subset of
your sample, if any, it applies to (i.e., male and female
respondents separately).
• Coefficient estimates and standard errors should be reported with
the same number of decimal places throughout your tables—
usually two or three.
• Some people like to omit control variables, preferring instead to
include a line that says “Controls? Yes” in the second (i.e.,
bottom) half of the table. Though this is fine to save space in a
published article, a working paper should show everything to the
readers (especially the reviewers and the editor). The obvious
exception is for individual, household, or community fixed
effects, of which there are usually too many to list. If you must
include a line at the bottom that says “Controls? Yes,” make sure
the notes to the table (i.e., right under the table) include a detailed
list of which controls are included—a careful reader will want to


know whether you condition on colliders or include as control a
variable that lies on the causal path between the treatment and
outcome variables.
11
• The last lines of the table should list the number of observations,
the R
2
(I prefer the usual R
2
to the adjusted one, because this tells
me how much of the variation in y is explained by the variables
on the right-hand side, without any arbitrary correction for the
number of observations and parameters), maybe the results of a
test of joint significance of the variables on the right-hand side,
and various lines indicating which controls are included (e.g.,
state fixed effects, a linear time trend, year fixed effects, state-
specific linear trends, state-specific quadratic trends, region–year
fixed effects, and so on).
• Finally, the notes to the table should present all symbols for
statistical significance (typically, * for statistical significance at
less than the 10 percent level, ** at less than the 5 percent level,
and *** at less than the 1 percent level; for completeness and
transparency, none should be omitted), and additional symbols if
necessary.
12
For instance, you may have adjusted your p-values
for multiple comparisons, bootstrapped your standard errors, or
done some randomization inference, all of which would lead to
different inferences and critical levels of statistical significance,
in which case you might use the symbols 


††
, and 
†††
to denote
significance at less than the 10, 5, and 1 percent level for this
additional version of the standard errors.
• Present estimation results for the same estimation sample. That is,
as the number of control variables increases, the sample size is
nonincreasing due to missing variables. If the sample size
decreases as you throw controls on the right-hand side, this
involves an apples-to-oranges comparison (different estimation
samples are representative of different populations). Instead, take
your smallest sample size (as dictated by missing observations)
and use that sample for all specifications.
For variable names, use plain English words like “Years of
education,” “Age squared,” and “Female” and not Stata or R
codenames like “Edu,” “AGE_2,” or “SEX.”


• Ultimately, it always helps to put yourself in your reader’s shoes,
and the right question to ask yourself (or a friend who owes you a
favor) is this: When given only the tables, can one write down the
exact regression that was estimated? Or is one left with more
questions than one has answered after looking at the tables?

Yüklə 1,91 Mb.

Dostları ilə paylaş:
1   ...   12   13   14   15   16   17   18   19   ...   94




Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©azkurs.org 2024
rəhbərliyinə müraciət

gir | qeydiyyatdan keç
    Ana səhifə


yükləyin