Question

You need to use two data sets for this exercise, JTRAIN2 and JTRAIN3. The former is the outcome of a job training experiment. The file JTRAIN3 contains observational data, where individuals themselves largely determine whether they participate in job training. The data sets cover the same time period. (i) In the data set JTRAIN2, what fraction of the men received job training? What is the fraction in JTRAIN3? Why do you think there is such a big difference? (ii) Using JTRAIN2, run a simple regression of $r e 78$ on train. What is the estimated effect of participating in job training on real earnings? (iii) Now add as controls to the regression in part (ii) the variables $r e 74,$ re75, educ, age, black, and hisp. Does the estimated effect of job training on $r e 78$ change much? How come? (Hint: Remember that these are experimental data.) (iv) Do the regressions in parts (ii) and (iii) using the data in JTRAIN3, reporting only the estimated coefficients on train, along with their $t$ statistics. What is the effect now of controlling for the $\quad$ extra factors, and why? (v) Define avgre $=(r e 74+r e 75) / 2 .$ Find the sample averages, standard deviations, and minimum and maximum values in the two data sets. Are these data sets representative of the same populations in 1978$?$ (vi) Almost 96$\%$ of men in the data set JTRAIN2 have avgre less than $\$ 10,000 .$ Using only these men, run the regression and report the training estimate and its $t$ statistic. Run the same regression for JTRAIN $3,$ using only men with avgre $\leq 10 .$ For the subsample of low-income men, how do the estimated training effects compare across the experimental and nonexperimental data sets? (vii) Now use each data set to run the simple regression $r e 78$ on train, but only for men who were unemployed in 1974 and $1975 .$ How do the training estimates compare now? (viii) Using your findings from the previous regressions, discuss the potential importance of having comparable populations underlying comparisons of experimental and nonexperimental estimates.

Name: Problem 10, Chapter 9: Introductory Econometrics
Uploaded: 2020-08-20T03:49:06-08:00
Duration: 16 min 23 s
Channel: Krish Desai

You need to use two data sets for this exercise, JTRAIN2 and JTRAIN3. The former is the outcome of
a job training experiment. The file JTRAIN3 contains observational data, where individuals themselves
largely determine whether they participate in job training. The data sets cover the same time period.
(i) In the data set JTRAIN2, what fraction of the men received job training? What is the fraction in
JTRAIN3? Why do you think there is such a big difference?
(ii) Using JTRAIN2, run a simple regression of $r e 78$ on train. What is the estimated effect of
participating in job training on real earnings?
(iii) Now add as controls to the regression in part (ii) the variables $r e 74,$ re75, educ, age, black,
and hisp. Does the estimated effect of job training on $r e 78$ change much? How come? (Hint:
Remember that these are experimental data.)
(iv) Do the regressions in parts (ii) and (iii) using the data in JTRAIN3, reporting only the estimated
coefficients on train, along with their $t$ statistics. What is the effect now of controlling for the
$\quad$ extra factors, and why?
(v) Define avgre $=(r e 74+r e 75) / 2 .$ Find the sample averages, standard deviations, and minimum and maximum values in the two data sets. Are these data sets representative of the same
populations in 1978$?$
(vi) Almost 96$\%$ of men in the data set JTRAIN2 have avgre less than $\$ 10,000 .$ Using only these men, run the regression and report the training estimate and its $t$ statistic. Run the same regression for JTRAIN $3,$ using only men with avgre $\leq 10 .$ For the subsample of low-income men, how do the estimated training effects compare across the experimental and nonexperimental data sets?
(vii) Now use each data set to run the simple regression $r e 78$ on train, but only for men who were unemployed in 1974 and $1975 .$ How do the training estimates compare now?
(viii) Using your findings from the previous regressions, discuss the potential importance of having
comparable populations underlying comparisons of experimental and nonexperimental estimates.

Introductory Econometrics

Jeffrey M.… 6th Edition

Chapter 9, Problem 10

Instant Answer

Solved by Expert Krish Desai

08/20/2020

Step 1

Ensure that the data is clean and ready for analysis. Show more…

Show all steps

Thanks for your feedback!

You need to use two data sets for this exercise, JTRAIN2 and JTRAIN3. The former is the outcome of a job training experiment. The file JTRAIN3 contains observational data, where individuals themselves largely determine whether they participate in job training. The data sets cover the same time period. (i) In the data set JTRAIN2, what fraction of the men received job training? What is the fraction in JTRAIN3? Why do you think there is such a big difference? (ii) Using JTRAIN2, run a simple regression of $r e 78$ on train. What is the estimated effect of participating in job training on real earnings? (iii) Now add as controls to the regression in part (ii) the variables $r e 74,$ re75, educ, age, black, and hisp. Does the estimated effect of job training on $r e 78$ change much? How come? (Hint: Remember that these are experimental data.) (iv) Do the regressions in parts (ii) and (iii) using the data in JTRAIN3, reporting only the estimated coefficients on train, along with their $t$ statistics. What is the effect now of controlling for the $\quad$ extra factors, and why? (v) Define avgre $=(r e 74+r e 75) / 2 .$ Find the sample averages, standard deviations, and minimum and maximum values in the two data sets. Are these data sets representative of the same populations in 1978$?$ (vi) Almost 96$\%$ of men in the data set JTRAIN2 have avgre less than $\$ 10,000 .$ Using only these men, run the regression and report the training estimate and its $t$ statistic. Run the same regression for JTRAIN $3,$ using only men with avgre $\leq 10 .$ For the subsample of low-income men, how do the estimated training effects compare across the experimental and nonexperimental data sets? (vii) Now use each data set to run the simple regression $r e 78$ on train, but only for men who were unemployed in 1974 and $1975 .$ How do the training estimates compare now? (viii) Using your findings from the previous regressions, discuss the potential importance of having comparable populations underlying comparisons of experimental and nonexperimental estimates.

Play audio

Feedback

Krish Desai and 60 other educators are ready to help you.

Ask a new question

Labs

Want to see this concept in action?

NEW

Explore this concept interactively to see how it behaves as you change inputs.

View Labs

Key Concepts

Subsample Analysis

Subsample analysis involves analyzing specific segments of the overall sample, such as low-income or previously unemployed individuals, to assess whether the treatment effect is consistent across different groups. This method helps in identifying heterogeneity in treatment effects and ensures comparability between groups tested across different datasets.

Population Comparability

Population comparability concerns the degree to which different datasets or samples represent the same underlying population. When comparing experimental data with observational data, ensuring population comparability is critical because differences in the characteristics of the samples can lead to variations in estimated treatment effects and affect the external validity of the findings.

Statistical Significance and t-Statistics

t-Statistics are used as a measure of statistical significance in regression analysis. They indicate whether an estimated effect is likely to be a true signal or if it might have arisen by chance, thereby providing insight into the reliability of the treatment effect estimates.

Causal Inference

Causal inference refers to the process of drawing conclusions about the causal effect of a treatment on an outcome. It relies on methodologies such as experimental design and controlled regression analysis to move beyond mere correlations and make stronger claims about causality.

Control Variables

Control variables are included in regression models to adjust for confounding factors that might skew the estimated effect of the treatment. Their inclusion helps in isolating the treatment effect by accounting for other influences on the outcome, thereby reducing omitted variable bias.

Observational Study and Selection Bias

Observational studies rely on individuals’ self-selection into treatment, which can introduce selection bias. This bias occurs because the characteristics that influence participation may also affect the outcome, making it challenging to disentangle the true effect of the treatment without additional adjustments.

Experimental Design

Experimental design involves the use of random assignment to create treatment and control groups that are comparable on both observed and unobserved characteristics. This helps to ensure that any observed differences in outcomes can be causally attributed to the treatment rather than other confounding factors.

Regression Analysis

Regression analysis is a statistical technique used to estimate the relationship between a dependent variable and one or more independent variables. In the context of causal inference, it provides a framework for estimating treatment effects and assessing how these effects change when additional control variables are included.

Recommended Videos

Transcript

00:01 All right, hello everybody.

00:02 This is a pretty long problem.

00:04 We're going to be comparing two different data sets that measure essentially the same thing with some slight differences.

00:10 And we're going to be basically discussing the importance of having comparable populations.

00:16 So again, there's a pretty long one.

00:17 So let's buckle in and get straight into it.

00:19 First thing we're going to need is we're going to need our woldridge package.

00:27 This contained their data sets.

00:30 I'm always, i'm always.

00:30 Also going to be using the dplyr package.

00:37 This package allows me to basically take account of all the data points and observations in a dataset.

00:45 It just makes my life a little bit easier.

00:48 So yeah, all right, we're gonna be using the woldridge library and the dplyr library.

00:56 Okay, all right, so first thing we will, and of course, sorry, our two datasets are the j train 2 and the j train 3 datasets.

01:06 Okay, let's get into it.

01:08 So first we want to find the fraction of members of the first data set that actually received job training.

01:14 So the way we're going to do this is we're going to say count and then we're going to find the subset of j train 2 where the condition for our subset subset is train equals 1.

01:26 So basically, whoever in this subset did receive job training, then we're going to divide that simply by the number of members of j -train 2.

01:35 And we will get about 41 .6%.

01:39 Right? we can do this exact same thing for j -train 3, and i'm only actually just going back in same function, same line of code, just changing the variable.

01:49 And we will see that here only 6 .92 % of the members received job training.

01:58 And so this difference is really important because in the first one, it's an experiment, right? we've done an experiment to find the results of job training on real earnings.

02:09 And the second one is just observational and there are a lot more members and very few of them actually went through job training.

02:17 So that's an important difference.

02:19 In fact, if we look at this, we can see that in j train two, there are only 445 people using the experiment.

02:25 But if we take this observational look, there are 2 ,675 people that responded to the observational survey.

02:35 So that's a pretty important difference.

02:37 Okay.

02:38 Now we're going to run a regression.

02:41 Actually, let me keep going.

02:46 There we go.

02:47 Okay.

02:47 Now we're going to run a regression on j train 2 to calculate the effect of training on real earnings in 78.

02:56 So pretty simple.

02:58 We're going to do j train 2 regression is a linear model, right, of re78 on training.

03:08 And our data is j train 2.

03:10 Again, this is a very standard, basically linear model function.

03:15 And if we take our summary, we will find that the effect of training is that, first of all, it is significant at the 5 % confidence level, or significance level.

03:24 But we will find that in the experimental study, if you receive training, your real earnings would go up by $1 ,794 .3.

03:37 Okay? so that's our estimated effect of participating in job training.

03:41 Now we're going to add some controls.

03:45 So our j -train controlled is going to be the same formula, but we're also going to add in re70.

03:56 Plus re75 plus education plus age plus black plus hispanic and when we take the summary of this we'll see that the training estimate actually didn't change that much is still significant at the 5 % confidence level but yeah this didn't change too much right it's still pretty close here it was 1 .7943 and here it's 1 .6801 so this isn't changing too much simply because it's an experiment and it's already designed to control for these factors, right? all of these are designed to be controlled for in an experiment that's trying to study this specific relationship.

04:40 Otherwise, it wouldn't be a good experiment if, you know, all these other factors influenced it.

04:45 So, yeah, so that's an important difference between something experimental versus something non -experimental.

04:51 And we will actually check that out right now by trying this exact same thing with the j train 3 dataset.

04:59 So here, same function, but instead this is a j train 3 regression.

05:05 Keep in mind, this one was observational, not experimental.

05:11 And same thing here with our controlled.

05:14 So we have our j train 3 regression and our j train 3 controlled.

05:18 And if we take the summary of these two, and the summary of our controlled, we will see that, so in our regular regression, there's actually a negative impact from training onto real earnings if you don't control for other factors.

05:39 When we do control for other factors, we do find a positive impact, but it's very minimal.

05:45 And that's because only 6 .915%, if you recall, actually receive job training in the previous one, which means that, you know, if people choose to receive job training, it doesn't really have a big impact on real earnings.

06:05 Okay.

06:06 All right.

06:07 So, and so if we look at the t statistics, we'll also find that training isn't statistically significant at all.

06:15 It's very, it's not even close.

06:18 This is a 0 .803 compared to the 0 .05, you need to be significant at the 5 % significance level.

06:24 So, and the effect of controlling for other factors now shows that, yeah, these other factors really do determine the real earnings in 78 as opposed to training.

06:35 It doesn't actually have a huge effect.

06:39 Okay, cool.

06:41 So that's a good comparison to make.

06:44 Now we're going to add into both of these a new variable.

06:49 This is called average real earnings.

06:52 And we're going to call this, we're going to define this as the earnings, as the average of the earnings in 74 and 75.

07:04 Sorry, i forgot it to 2 here.

07:06 But our formula is just going to be the real earning 74 plus the real earning 75 divide by 2.

07:12 And we'll do the same thing for j -train 3.

07:21 Awesome.

07:22 Now we have our two things.

07:23 In fact, just to show you, if i view this, go all the way.