In Figure 2, we see a striking difference between the B-W and CPS data.
The combined effect of the various discrepancies between the B-W estimated figures and the CPS data, is apparent in Figure 3.(13) Here, the B-W and CPS estimates of the married female hourly wage are plotted together with the fertility rate, for women aged 20-24 and 25-34.
Similarly, there is a continued upward trend in the B-W series in the period 1967-78, which results largely from a spurious decline in estimated hours worked, as explained above.
In [Y.sub.m] -0.4019 -0.0791 -0.4166 (-2.416) (-0.949) (-1.374) In [Y.sub.m] -0.8281 -0.9372 -1.9482 (-1.393) (-3.497) (-5.740) Adjusted [R.sub.2] 0.1358 0.5883 0.5813 Durbin-Watson 0.671 0.631 1.032 p 0.600 0.609 0.451 Degrees of freedom 23 23 23 Results are set out in the top panel of Table 2 for two of the age groups used by B-W in their original analysis, 20-24 and 25-34, and also for the age groups 25-29 and 30-34, since these five year groupings were available using CPS data.
In the bottom panel of Table 2, an attempt has been made to improve the model by replacing total fertility rates with marital fertility rates, since the B-W model specifically addresses marital fertility.(16) It is apparent that this attempt serves only to make the model worse: in this specification, the adjusted [R.sup.2] all drop to .59 or lower (including a remarkably low [R.sup.2] of only .14 for the 20-24 age group), with unacceptably low Durbin-Watson scores, and the signs on all variables are uniformly "perverse" (although not always significant).