You are on page 1of 22

5SSMN932 Introduction to Econometrics

2020/21
Coursework Questions
Instructions
The deadline for this Coursework is Friday 20th of November 2020, 10am.
Upload your answers in a single file via Turnitin using the link on KEATS.
Include at the end of the file all the Stata commands you used to answer the questions.
Keep your answers short, approx. 30-70 words and no more than 100 words for any
question. Maximum 700 words overall (excluding Stata tables and commands).

The data
In this coursework exercise you will investigate the relationship between a person’s labour
earnings and age, education, gender and a few other characteristics. The file cwdata.dta
contains data for full-time workers in the US. The data are from the US National Health
Interview Survey. The sample is chosen randomly from the whole population of the US.
The variables included in cwdata.dta are :
earnings Annual labour earnings (expressed in USD per year)
age Age in years
educ Years of schooling
woman Gender (0 man, 1 woman)
married Marital status ( 0 unmarried, 1 married)
region Region (1 Northeast, 2 Midwest, 3 South, 4 West)
manager Executive / managerial position (0 no, 1 yes)
height Height without shoes (in inches)

Questions:
1. Run a regression of annual earnings (earnings) on age (age). How do you interpret
the intercept and slope coefficients? (10 marks)

2. Run a regression of earnings on age, gender (woman), and education (educ). What is
the effect of age on yearly earnings according to this regression? (10 marks)

1. Are the results from the regression in (1) substantively different from the results in
(2) regarding the effects of age on annual earnings? Does the regression in (1) seem
to suffer from omitted variable bias? (20 marks)
[Hint: explain the direction of this bias.]

2. Include now some of the other control variables in the regression: regress earnings
on age, age2 , woman, educ, married, and manager. If age increases from 35 to 36
years, how are annual earnings expected to change? (20 marks)
[Hint: create a new variable using: gen age2=age^2. Alternatively, you can do this
automatically in Stata using the operator c.age##c.age.]

1
3. Run the same regression as in (4) but using the logarithm of earnings: regress log of
earnings on age, age2 , woman, educ, married, and manager. What is the effect of
an additional year of schooling on earnings according to this regression? (15 marks)
[Hint: create the new variable log earnings using: gen lnearn=ln(earnings)]

4. Run the same regression as in (5) but include also the interaction term woman x educ.
What does the coefficient of this interaction term measure? (15 marks)
[Hint: you can do this automatically in Stata using the operator woman##c.educ.]

5. Include now height in the regression from question (5). Does the height of a person
appear to be a determinant of earnings? Is the effect of height on earnings different
for women than for men? Specify and estimate a regression (or two) that you can
use to answer this question. (10 marks)

2
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

2020/21
COURSEWORK
Question 1
Run a regression of annual earnings (earnings) on age (age). How do you interpret
the intercept and slope coefficients? (10 marks)

The slope coefficient is 274.90, which means that with each additional year of age, a full-
time worker in the US earns an extra $274.90 yearly on average. This coefficient is
significant to 99% confidence level since the p-value is smaller than 0.01.

The intercept coefficient is 35,687.98, and it has no intrinsic meaning other than being the
y-intercept when earnings is plotted against age; it’s the expected earnings in USD for a
full-time worker who is 0 years old, which does not make sense in reality. It is significant to
99% confidence level.

1
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 2
Run a regression of earnings on age, gender (woman), and education (educ). What is
the effect of age on yearly earnings according to this regression? (10 marks)

The coefficient on age is 333.52, which means that with each additional year of age, a full-
time worker in the US earns an extra $333.52 yearly on average. This coefficient is
significant to 99% confidence level since the p-value is smaller than 0.01.

2
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 3
Are the results from the regression in (1) substantively different from the results in (2)
regarding the effects of age on annual earnings? Does the regression in (1) seem to
suffer from omitted variable bias? (20 marks)
By including variables such as gender and education, the coefficient on age changed from
274.90 to 333.52, a 21.3% increase.

For OVB, 2 conditions must hold: correlation between age and the new variables must be
non-zero, and the ommited variables are a determinant of yearly earnings.

1. Corr(Age, Woman)=0.0094 and Corr(Education, Age)=—0.054, non-zero.

2. Both variables affect yearly earnings since the Corr(Earnings, Woman)=—0.051 and
Corr(Earnings, Education)=0.388, which is why regression (1) suffers from OVB.

The direction of the bias will be negative as Corr(Education, Age) is negative, and since the
positive Corr(Age, Woman) is almost insignificant.

3
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

y S = β0S + β1S X1 + u S

y L = β0L + β1L X1 + β2L X2 + β3L X3 + u L

given X2 = δ0 + δ1X1 + v1 an d X3 = δ2 + δ3 X1 + v2

OVB = β1S − β1L = δ1 β2L + δ2 β3L

y S = 35,687.98 + 274.90(Age) + u S

y L = − 19,668.76 + 333.53(Age) − 2672.33(Wom an) + 4020.861(Educ) + u L

given Wom an = 0.538 + 0.00046(Age) + v1 an d Educ = 14.13 − 0.0143(Age) + v2

OVB = 274.90 − 333.53 = 0.00046(−2672.33) − 0.0143(4020.86) = −58.63

4
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 4
Include now some of the other control variables in the regression: regress earnings
on age, age2, woman, educ, married, and manager. If age increases from 35 to 36
years, how are annual earnings expected to change? (20 marks)

E(Ear nings | Age = 36) − E(Ear nings | Age = 35) =

2583.29(36) − 26.86(362) + wom an0 + educ0 + m ar r ied0 + m an ager0

−2583.29(35) + 26.86(352) − wom an0 − educ0 − m ar r ied0 − m an ager0

2583.29(36 − 35) − 26.86(362 − 352) = $676.41

Annual earnings are expected to increase by $676.41 if age increases from 35 to 36 years
holding gender, education, marital status and exec/manager constant.

5
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 5
Run the same regression as in (4) but using the logarithm of earnings: regress log of
earnings on age, age2, woman, educ, married, and manager. What is the effect of an
additional year of schooling on earnings according to this regression? (15 marks)

ln(△ear n) = βeduc △ educat ion → △ ear nings = 0.09667 * 100 % = 9.667 %


An additional year of schooling has a 9.67% increase in annual earnings.

6
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 6
Run the same regression as in (5) but include also the interaction term woman x
educ. What does the coefficient of this interaction term measure? (15 marks)

The coefficient on the interaction term woman x educ measures the increment on the
effect of education on annual earnings if woman=1 (if the subject is female), which would
be an additional 1.11%. Therefore, when the subject is male, the effect of an additional
year of schooling on annual earnings is an increase of 9.11%. However, when the subject
is female, the effect of an additional year of schooling on annual earnings is an increase of
9.105+1.108=10.213%

7
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Question 7
Include now height in the regression from question (5). Does the height of a person
appear to be a determinant of earnings? Is the effect of height on earnings different
for women than for men? Specify and estimate a regression (or two) that you can use
to answer this question. (10 marks)

The height of a person appears to be a determinant of earnings: an increase in height of 1


inch results in an increase in earnings by 1.32%, and the coefficient is significant to 99%
confidence level.

8
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

The effect on height for women and for men is different since:

1. When running the regression with the additional regressor woman x height, the
coefficient on the interaction term was -0.00923 (non-zero). The results were:

• For a male subject, when woman=0, an increase in height by an inch results in a


1.80% increase in annual earnings

• For a female subject, when woman=1, an increase in height by an inch results in a


1.796-0.923=0.873% increase in annual earnings, almost half than for men.

2. By looking at the OVB in the independent variable ‘height’ when the variable ‘woman’ is
not included:

y S = β0S + β1S X1 + u S

y L = β0L + β1L X1 + β2L X2 + u L

given X2 = δ0 + δ1X1 + v

OVB = β1S − β1L = δ1 β2L

9
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

ln(ear nings S ) = 6.7156 + 0.01229(Height) + ( . . . ) + u S

ln(ear nings L ) = 6.6494 + 0.01322(Height) + +0.0104(Wom an) + ( . . . ) + u L

given Wom an = 6.421 + 0.08755(Height) + v

OVB = 0.01229 − 0.01322 = 0.08755(0.0104) = −0.000928 or − 0.093 % (3s f )

10
Module Code: 5SSMN932
Candidate Number: AB00913
Module Name: Introduction to Econometrics

Do-File
reg earnings age

reg earnings age woman educ

corr age woman educ

corr earnings woman educ

reg woman age

reg educ age

gen age2=age^2

reg earnings age age2 woman educ married manager

gen lnearn=ln(earnings)

reg lnearn age age2 woman educ married manager

reg lnearn age age2 married manager woman##c.educ

reg lnearn age age2 woman educ married manager height

reg lnearn age age2 educ married manager woman##c.height

reg lnearn age age2 educ married manager height

reg woman height

11
King’s Business School, King’s College London
Cover sheet for Coursework Assessment

Candidate ID: AB14123

Module Code: 5SSMN932

Module Name: Introduction to Econometrics

Word Count: 700

Please complete the above candidate and module information and


provide your answers below.

Where applicable students should clearly state the question(s) they are
answering (e.g. Question 1, Part A) so it can be clearly identified for
markers.

DECLARATION BY STUDENT

This assignment is entirely my own work.

I understand what is meant by plagiarism/collusion and have signed at


enrolment the declaration concerning the avoidance of
plagiarism/collusion.

I understand that plagiarism/collusion is a serious offence that may


result in disciplinary action being taken.

I understand that I must submit work BEFORE the deadline, and that
failure to do so may result in late submission penalties.

I understand that where a word limit is indicated, I should adhere to


the word limit and failure to do so may result in penalties.

By completing this assessment, you acknowledge that you have read


and understand the above. Please do not sign your name.
Question 1: Run a regression of annual earnings (earnings) on age (age). How do
you interpret the intercept and slope coefficients? (10 marks)

. reg earnings age

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(1, 17332) = 183.26
Model | 1.3165e+11 1 1.3165e+11 Prob > F = 0.0000
Residual | 1.2451e+13 17,332 718378743 R-squared = 0.0105
-------------+---------------------------------- Adj R-squared = 0.0104
Total | 1.2583e+13 17,333 725932604 Root MSE = 26803

------------------------------------------------------------------------------
earnings | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | 274.9016 20.30694 13.54 0.000 235.098 314.7053
_cons | 35687.98 854.9941 41.74 0.000 34012.11 37363.86
------------------------------------------------------------------------------

The OLS coefficient estimate is 274.9016 which tells us that if we increase the age of a
person by 1 year, we expect their average annual earnings to increase by $274.90.

We can reject the null hypothesis as the t-stat is 13.54(>2.567) and p value is
0.000(<0.01). The impact of age on earnings is significant at 1%.

The OLS intercept estimate of 35687.98 means that a person with an age of 0 will earn an
average of $35,678.98 per year - this is not useful in our regression analysis as it doesn’t
have a meaningful interpretation.

The R squared shows us that age only explains 1.05% of the variability in earnings.
Question 2: Run a regression of earnings on age, gender (woman), and education
(educ). What is the effect of age on yearly earnings according to this regression? (10
marks)

. reg earnings age woman educ

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(3, 17330) = 1167.78
Model | 2.1159e+12 3 7.0530e+11 Prob > F = 0.0000
Residual | 1.0467e+13 17,330 603964460 R-squared = 0.1682
-------------+---------------------------------- Adj R-squared = 0.1680
Total | 1.2583e+13 17,333 725932604 Root MSE = 24576

------------------------------------------------------------------------------
earnings | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | 333.5273 18.64785 17.89 0.000 296.9756 370.0789
woman | -2672.331 375.8246 -7.11 0.000 -3408.985 -1935.677
educ | 4020.861 70.7593 56.82 0.000 3882.166 4159.557
_cons | -19668.76 1287.359 -15.28 0.000 -22192.11 -17145.41
------------------------------------------------------------------------------

If we had full randomisation, with no conditional randomisation, we would expect the


coefficient estimate to be similar to (1) with a smaller standard error.

We observe that an additional year of age results in a $333.53 increase in average annual
earnings in this regression.

This value is statistically significant at the 1% level as the t-value is 17.89(>2.567) and p
value is 0.000(<0.01).

The standard error on age has decreased from 20.31 in (1) to 18.65 in (2).

The R squared has increased significantly as there are more variables in this regression; it
now explains 16.82% of the variation in earnings. Therefore, the effect of age on yearly
earnings is more accurate in this model.
Question 3: Are the results from the regression in (1) substantively different from the
results in (2) regarding the effects of age on annual earnings? Does the regression in
(1) seem to suffer from omitted variable bias? (20 marks)

The coefficient on age is 274.90 in (1) compared to 333.53 in (2) so the effect of age on
earnings has increased by 21.33%.

The standard error on age has decreased from 20.31 in (1) to 18.65 in (2). The R-squared
shows that (2) explains 15.77% more of the variability in earnings. The Root MSE
decreases from 26803 to 24576. Therefore, the coefficient estimate in (2) is more
accurate.

• OVB = age coefficient in short regression - age coefficient in long regression

• OVB = 274.9016 - 333.5273 = -58.6257

Age and earnings are positively correlated and the value for OVB is negative so the overall
OVB is negative. This means that ignoring the effect of education and gender understates
the effect of age on earnings in (1).

Gender is an omitted variable but it does not contribute to OVB on age because age has
no statistically significant effect on gender. The p-value 0.217>0.1 and the t-stat
1.23<1.282 so we cannot reject the null hypothesis at the 10% level.

. reg age woman

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(1, 17332) = 1.52
Model | 153.078682 1 153.078682 Prob > F = 0.2172
Residual | 1741912.12 17,332 100.502661 R-squared = 0.0001
-------------+---------------------------------- Adj R-squared = 0.0000
Total | 1742065.2 17,333 100.505694 Root MSE = 10.025

------------------------------------------------------------------------------
age | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
woman | .189194 .1532988 1.23 0.217 -.1112871 .4896751
_cons | 40.7872 .11444 356.41 0.000 40.56289 41.01152
------------------------------------------------------------------------------
Question 4: Include now some of the other control variables in the regression:
regress earnings on age, age2, woman, educ, married, and manager. If age
increases from 35 to 36 years, how are annual earnings expected to change? (20
marks)

. gen age2 = age*age


. reg earnings age age2 woman educ married manager

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(6, 17327) = 1402.86
Model | 4.1139e+12 6 6.8565e+11 Prob > F = 0.0000
Residual | 8.4687e+12 17,327 488755126 R-squared = 0.3270
-------------+---------------------------------- Adj R-squared = 0.3267
Total | 1.2583e+13 17,333 725932604 Root MSE = 22108

------------------------------------------------------------------------------
earnings | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | 2583.294 137.1386 18.84 0.000 2314.489 2852.099
age2 | -26.85759 1.590982 -16.88 0.000 -29.97608 -23.73911
woman | -2392.674 338.6715 -7.06 0.000 -3056.504 -1728.844
educ | 3849.855 64.60724 59.59 0.000 3723.218 3976.492
married | 20600.91 354.0357 58.19 0.000 19906.96 21294.85
manager | 8705.34 550.7602 15.81 0.000 7625.795 9784.886
_cons | -76243.18 2913.888 -26.17 0.000 -81954.69 -70531.67
------------------------------------------------------------------------------

If age increases from 35 to 36 years, annual earnings will increase by:


• 35 age earnings = 35(2583.294) +35^2(-26.85759) = 57514.74225
• 36 age earnings = 36(2583.294) + 36^2(-26.85759) = 58191.14736
• Diff in earnings = 58191.14736 - 57514.74225 = 676.40511

Annual earnings are expected to increase by $676.41 with an increase in age from 35 to
36.

Having age2 in the regression shows us that the marginal effect of age on yearly earnings
is non linear, it is quadratic. Since the coefficient on age2 is negative, annual earnings will
increase with age until a given age (at the top of the parabola) and decrease hereafter.

The adjusted R-squared in this model is 15.87% higher than (2) and the Root MSE
decreases from 24576 to 22108 showing that this regression is more accurate.

All coefficients are statistically significant at the 1% level.

We can construct an F-statistic to test a joint hypothesis that the coefficients of both age=0
and age2=0. The F-stat (287.16) is significantly higher than 3 so the null hypothesis can be
rejected at the 5% level.

. test age age2


( 1) age = 0
( 2) age2 = 0
F( 2, 17327) = 287.16
Prob > F = 0.0000
Question 5: Run the same regression as in (4) but using the logarithm of earnings:
regress log of earnings on age, age2, woman, educ, married, and manager. What is
the effect of an additional year of schooling on earnings according to this
regression? (15 marks)

. gen lnearn=ln(earnings)
. reg lnearn c.age##c.age woman educ married manager

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(6, 17327) = 1515.51
Model | 2613.46181 6 435.576969 Prob > F = 0.0000
Residual | 4980.00844 17,327 .287413196 R-squared = 0.3442
-------------+---------------------------------- Adj R-squared = 0.3439
Total | 7593.47025 17,333 .438093247 Root MSE = .53611

------------------------------------------------------------------------------
lnearn | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | .0609042 .0033256 18.31 0.000 .0543857 .0674227
|
c.age#c.age | -.0006405 .0000386 -16.60 0.000 -.0007161 -.0005648
|
woman | -.0634451 .0082127 -7.73 0.000 -.0795428 -.0473473
educ | .0966711 .0015667 61.70 0.000 .0936002 .099742
married | .533651 .0085853 62.16 0.000 .5168229 .550479
manager | .2053434 .0133558 15.37 0.000 .1791647 .2315222
_cons | 7.563876 .0706611 107.04 0.000 7.425373 7.702379
------------------------------------------------------------------------------

The OLS coefficient estimate for educ is 0.09667. This means that an additional year of
schooling results in an increase in lnearn of 0.09667 which is equal to a 9.7% increase in
average annual earnings.

The large t-value (61.70>2.567) and small p-value (0.000<0.01) for education show that
the coefficient estimate is statistically significant at the 1% level.

The Root MSE is not comparable as the units are different for earnings. Similarly, we
cannot compare the R-squared because the dependent variable is different in (4).

This is a linear log regression. The log scale for earnings is an approximate for the
percentage scale. Homoskedasticity is more likely to hold so a normal distribution is more
plausible and it helps reduce the influence of outliers.
Question 6: Run the same regression as in (5) but include also the interaction term
woman x educ. What does the coefficient of this interaction term measure? (15
marks)

. reg lnearn age age2 woman educ married manager woman##c.educ


note: 1.woman omitted because of collinearity
note: educ omitted because of collinearity

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(7, 17326) = 1301.74
Model | 2617.16716 7 373.881023 Prob > F = 0.0000
Residual | 4976.3031 17,326 .287215924 R-squared = 0.3447
-------------+---------------------------------- Adj R-squared = 0.3444
Total | 7593.47025 17,333 .438093247 Root MSE = .53593

------------------------------------------------------------------------------
lnearn | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | .0608511 .0033245 18.30 0.000 .0543348 .0673674
age2 | -.0006392 .0000386 -16.57 0.000 -.0007148 -.0005636
woman | -.2135431 .0425881 -5.01 0.000 -.29702 -.1300661
educ | .0910505 .002214 41.13 0.000 .086711 .0953901
married | .5338715 .0085826 62.20 0.000 .5170489 .5506942
manager | .2064662 .0133549 15.46 0.000 .1802893 .2326432
|
woman |
1:woman | 0 (omitted)
educ | 0 (omitted)
|
woman#c.educ |
1:woman | .0110802 .0030849 3.59 0.000 .0050336 .0171269
|
_cons | 7.639869 .0737374 103.61 0.000 7.495337 7.784402
------------------------------------------------------------------------------

Keeping all other variables in the regression constant, the predicted difference in lnearn
between males and females at any level of education = -0.2135431 + 0.0110802 * educ

For a man, an extra year of schooling is associated with a 9.1% increase in average
annual earnings, whereas for women it is a 10.2% increase.

The coefficient on the interaction term, 0.0110802, is the difference in the effect of
education for women versus men. This value is statistically significant at the 1% level as
the t-value (3.59>2.567) and p-value (0.000<0.01) indicate.

Including the interaction term only slightly increases the R-squared from (5) showing that
this regression explains 0.05% more of the variation in earnings. The Root MSE actually
decreases from 0.53611 to 0.53593. The standard error on educ increases from
0.0015667 to 0.002214.

Woman and educ were omitted from the regression due to perfect multi collinearity - there
shouldn’t be any linear relationships between the independent variables and these
variables were included twice by mistake.

Question 7: Include now height in the regression from question (5). Does the height
of a person appear to be a determinant of earnings? Is the effect of height on
earnings different for women than for men? Specify and estimate a regression (or
two) that you can use to answer this question. (10 marks)

. reg lnearn age age2 woman educ married manager height

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(7, 17326) = 1316.99
Model | 2637.17613 7 376.739448 Prob > F = 0.0000
Residual | 4956.29412 17,326 .286061071 R-squared = 0.3473
-------------+---------------------------------- Adj R-squared = 0.3470
Total | 7593.47025 17,333 .438093247 Root MSE = .53485

------------------------------------------------------------------------------
lnearn | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | .061562 .0033185 18.55 0.000 .0550573 .0680666
age2 | -.0006462 .0000385 -16.79 0.000 -.0007217 -.0005708
woman | .0104731 .0115343 0.91 0.364 -.0121354 .0330815
educ | .0945106 .0015809 59.78 0.000 .0914118 .0976093
married | .5345593 .0085656 62.41 0.000 .5177698 .5513489
manager | .2035318 .0133258 15.27 0.000 .1774118 .2296518
height | .0132224 .0014522 9.10 0.000 .0103759 .0160689
_cons | 6.649389 .1227089 54.19 0.000 6.408867 6.88991
------------------------------------------------------------------------------

An increase in height by an inch is predicted to increase annual earnings by 1.32%. This


is statistically significant at the 1% level as the t-value is 9.10 and the p-value is 0.

The R-squared increased from 0.3442 in (5) to 0.3473 in (7) showing that adding height to
the regression explains 0.31% more of the variation in lnearn.

The p-value for women is 0.364(>0.1) which shows that this is the only coefficient that is
not statistically significant and the null hypothesis cannot be rejected at the 10% level.

Woman is a dummy variable. The regression below shows that when all other variables
are kept constant, the predicted difference in lnearn between males and females at any
height = 0.632785 - 0.0092387 * height

For men, an additional inch of height is associated with a 1.8% increase in yearly
earnings, whereas for women it is only a 0.9% increase.

All the coefficients are statistically significant at the 1% level.


. reg lnearn age age2 educ married manager woman##c.height

Source | SS df MS Number of obs = 17,334


-------------+---------------------------------- F(8, 17325) = 1154.29
Model | 2640.14346 8 330.017932 Prob > F = 0.0000
Residual | 4953.3268 17,325 .285906309 R-squared = 0.3477
-------------+---------------------------------- Adj R-squared = 0.3474
Total | 7593.47025 17,333 .438093247 Root MSE = .5347

--------------------------------------------------------------------------------
lnearn | Coef. Std. Err. t P>|t| [95% Conf. Interval]
---------------+----------------------------------------------------------------
age | .0614679 .0033178 18.53 0.000 .0549647 .067971
age2 | -.0006453 .0000385 -16.77 0.000 -.0007208 -.0005699
educ | .0943169 .0015816 59.63 0.000 .0912167 .0974171
married | .534104 .0085645 62.36 0.000 .5173167 .5508913
manager | .2039526 .0133229 15.31 0.000 .1778384 .2300668
|
woman |
1:woman | .632785 .193513 3.27 0.001 .2534801 1.01209
height | .0179631 .0020672 8.69 0.000 .0139112 .022015
|
woman#c.height |
1:woman | -.0092387 .0028678 -3.22 0.001 -.0148598 -.0036176
|
_cons | 6.32227 .159247 39.70 0.000 6.010129 6.63441
--------------------------------------------------------------------------------

Stata commands:

. reg earnings age


. reg earnings age woman educ
. reg age woman
. gen age2 = age*age
. reg earnings age age2 woman educ married manager
. test age age2
. gen lnearn=ln(earnings)
. reg lnearn c.age##c.age woman educ married manager
. reg lnearn age age2 woman educ married manager woman##c.educ
. reg lnearn age age2 woman educ married manager height
. reg lnearn age age2 educ married manager woman##c.height

You might also like