Suppose a national survey of women in the United States was conducted in the years 1972, 1974, 1976, 1978, 1980, 1982, and 1984. Suppose the survey data from each year is pooled to create a pooled cross-sectional data set consisting of 13,000 observations. You want to use these data to estimate a multiple linear regression model using OLS to explain the number of children born to a woman during this time period. To do this, you include variables to control for education, age, and ethnicity, as well as dummy variables to control for location within the United States. You also include dummy variables for the years 1974, 1976, 1978, 1980, 1982, and 1984. For example, if an observation comes from the year 1974, y74 would equal 1, while the remaining dummy variables for other years would equal O.
The coefficient estimates and standard errors are summarized in the following table:
Determinants of Women's Fertility Dependent Variable: kids Base Year: 1972 Independent Variable coefficient Standard Error educ -0.135 0.20 age 0.277 0.111 age2 -0.0055 0.001 black 0.88 0.177 east 0.217 0.133 north 0.363 0.121 west 0.198 0.167 farm -0.053 0.147 -0.053 0.147 farm othrurual -0.163 0.175 town 0.084 0.124 0.212 0.160 0.190 0.171 -0.09 0.171 smcity y74 y76 y78 y80 y82 y84 -0.060 0.182 -0.080 0.188 -0.420 0.175 -0.650 0.173 Constant -7.731 3.101 n 13,000 R2 0.127 0.115
The coefficient of y84 is predicted to have . This implies, holding education, age, race, and location of residence constant, 100 women in 1984 are predicted to have children compared to 100 women in 1972.