I examine here the magnitude of the BW gap at different levels of parental education, using the NLSY97. You need to download a collection of variables in the NLS Investigator. The relevant ones for the present analysis are R1302400, R1302500, R1482600, R0538700, R0538600, and R9829600. You can check these variables using the “Variable Search” option in the NLS Investigator. And you may need a SPSS or PSPP program. In the SPSS (or PSPP) program, click on File, Open, and then Syntax, to upload the collection of variables you have downloaded from the NLS I (you should have transferred  your files into a new file located in your “Documents and Settings” folder). Recall that your handle file should look like this, or otherwise the recoding will not work.

Now, I want to create a new variable, collapsing these two variables :

R1302400 “CV_HGC_BIO_DAD 1997”
R1302500 “CV_HGC_BIO_MOM 1997”

Father’s grade and mother’s grade, respectively. For the information, in the NLSY97, these two variables have a value going from 0 to 20 (where 0 = None, 1 = 1st Grade, 2 = 2nd Grade, and so on). What I need to do, is to create a 3-categories ‘parental education’ variable.

Here’s my method. First, go to “transform”, and “create variable” in your SPSS program. Select the R1302400 variable in the list and give it a new name (e.g., FGRADE). Then, click on “old and new value”. Click on “Range”, and enter 1 in the first box, 11 in the second box, click on “value” in the panel “new value” and enter 1. Hit the box “Add”. Click on “Value” in the panel “old value” and enter 12. Click on “value” in the panel “new value” and enter 2. Hit the box “Add”. Click on “Range”, and enter 13 in the first box, 20 in the second box, click on “value” in the panel “new value” and enter 3. Hit the box “Add”. The box “old –> new” should look like this if you did this right :

1 thru 11 –> 1
12 –> 2
13 thru 20 –> 3

Repeat the process for R1302500. Then, in your syntax editor, enter the following code :


Black-White IQ gap by Parental Education in the NLSY97Click on “Run”, “All”. Then, click “Comparison of means”, and “Means”.

Your list of dependent variables should be :
And your list of independent variables should be :
Layer 1 : WHITE (i.e., the name of my recoded variable for non-hispanic whites, see here)
Layer 2 : KEY!RACE_ETHNICITY R1482600
Layer 3 : PARENTSGRADE (i.e., the name of my ‘parental education’ variable)

The ASVAB scores displayed in the image opposite are unweighted. Note that I have restricted the sample to US born people (using the recoded variable RUSBORN, detailed here). Under the column KEY!RACE_ETHNICITY, 1 corresponds to the black sample, 2 to the hispanic sample, 3 to mixed race, and 4 to the non-black/non-hispanic sample. Under the column WHITE, 1 corresponds to the non-hispanic whites.

The d gap is about 0.78, 1.01, and 0.98 for PARENTSGRADE =1, PARENTSGRADE =2, and PARENTSGRADE =3, respectively. And below is a table generated in Excel which shows the weighted d gaps.

NLSY97 Black-White IQ d gap by Parental Education (Excel function)

As we can see, and consistent with previous findings (Jensen, 1973, pp. 241-242; Herrnstein & Murray, 1994, pp. 287-288; Jensen, 1998, p. 358; Gottfredson, 2003, Table 2; Chuck, Jan. 19, 2012; Hu, Jan. 1, 2013), the BW IQ gap increases with the level of parental education, which is clearly at odds with the environmental hypothesis, whatever the cause of such increasing gap might be.