Table 5

Impact of Marriage Law on Crude Birth Rate

Log Births per 1,000 in Population
All Treatment Counties UsedMississippi Not Included
1234567
Post X treatment−0.0158
(0.0185)
−0.0424*
(0.0216)
−0.0424*
(0.0216)
−0.0584***
(0.0218)
−0.0558**
(0.0274)
−0.136***
(0.0286)
−0.113***
(0.0307)
Treatment (=1 if Mississippi or county bordering Mississippi)0.0781***
(0.0213)
0.0903***
(0.0253)
0.0487
(0.0397)
0.0769**
(0.0308)
0.0575*
(0.0324)
0.104***
(0.0338)
0.0952***
(0.0358)
Post (=1 if year ≥ 1957)−0.269***
(0.00609)
−0.242***
(0.0126)
−0.310***
(0.0144)
−0.320***
(0.0168)
−0.300***
(0.0185)
−0.322***
(0.0182)
−0.371***
(0.0221)
Manufacturing wage−0.00117**
(0.000479)
−0.00139***
(0.000497)
−0.00142***
(0.000463)
−0.00130***
(0.000461)
Farms per 1,000 in population−3.222***
(0.228)
−4.138***
(0.408)
−3.336***
(0.277)
−3.353***
(0.280)
Percent employed in manufacturing−0.000440
(0.000336)
−8.80e–06
(0.000362)
−0.000470
(0.000351)
−0.000599**
(0.000352)
Percent employed in agriculture0.000953***
(0.000359)
Fixed effects usedNoneNoneState, YearState, YearState, YearState, YearState X Yea:
Control groupAll counties in southern U.S.A.Counties in states neighboring Mississippi only
Observations4,6501,3721,3721,029686810810
R-squared0.1390.1820.3710.5010.5030.5140.5+1
  • Notes: Dependent variable is log births per 1,000 of the population. Data from city and county data book for years 1948, 1950, 1954, and 1960 are used. Population at the county level for 1950 and 1960 is used to compute the percentages (1954 numbers are divided by the average population between 1950 and 1960). Treatment is 1 if state is Mississippi and border counties. Agricultural employment data are only available for 1950 and 1960. Manufacturing employment and wages data are not available for 1948. Regressions reflect estimation of Equation 1 in the paper. Outliers in the dependent variable (top and bottom 5 percent of the counties with changes in birth rates) removed. Including these counties increases the magnitude of the effect to around 8–9 percent in Columns 1–5 and the effects remaining statistically significant. These estimates are presented in Appendix Table 13.

  • * significant at 10 percent;

  • ** significant at 5 percent;

  • *** significant at 1 percent, robust standard errors, clustered at the county level