The syntax below computes a general vocabulary factor based on the 10 word vocabulary test in the General Social Survey, a survey which can be downloaded here. The item loadings were as follows:

This factor showed little incremental validity over WORDSUM, which is computed simply by summing correct responses. Below, REGR factor 1 is the first factor extracted with principle axis factoring.

The answer, then, is: No.

**Synatax**

**Syntax**

If WORDA=1 NWORDA=1.

If WORDA=0 or WORDA=9 NWORDA=0.

If WORDB=1 NWORDB=1.

If WORDB=0 or WORDB=9 NWORDB=0.

If WORDC=1 NWORDC=1.

If WORDC=0 or WORDC=9 NWORDC=0.

If WORDD=1 NWORDD=1.

If WORDD=0 or WORDD=9 NWORDD=0.

If WORDE=1 NWORDE=1.

If WORDE=0 or WORDE=9 NWORDE=0.

If WORDF=1 NWORDF=1.

If WORDF=0 or WORDF=9 NWORDF=0.

If WORDG=1 NWORDG=1.

If WORDG=0 or WORDG=9 NWORDG=0.

If WORDH=1 NWORDH=1.

If WORDH=0 or WORDH=9 NWORDH=0.

If WORDI=1 NWORDI=1.

If WORDI=0 or WORDI=9 NWORDI=0.

If WORDJ=1 NWORDJ=1.

If WORDJ=0 or WORDJ=9 NWORDJ=0.

FACTOR

/VARIABLES NWORDA NWORDB NWORDC NWORDD NWORDE NWORDF NWORDG NWORDH NWORDI NWORDJ

/MISSING PAIRWISE

/ANALYSIS NWORDA NWORDB NWORDC NWORDD NWORDE NWORDF NWORDG NWORDH NWORDI NWORDJ

/PRINT INITIAL EXTRACTION

/CRITERIA MINEIGEN(1) ITERATE(25)

/EXTRACTION PAF

/ROTATION NOROTATE

/SAVE REG(ALL)

/METHOD=CORRELATION.

But there is still a very slight improvement for educational attainment, no ? Of course, no incremental validity occurs for the correlation with parents’ income and education probably because a person’s IQ is more correlated with his attained SES than his parents’ SES. Also, have you used the weights for your correlation analysis ? Using the syntax :

WEIGHT BY wtssnr.

Regarding the skin color variable, the fact that it correlates higher with wordsum than with income or education makes me think about what Jensen used to say in Educability & Group differences. SES has less genetic component, and so is more malleable than IQ (and so we expect a lower BW/color difference concerning SES-related variables). From a colorism point of view we would also expect a larger BW/color difference with variables having larger environmental component.