Using Stata for Survey Data Analysis

Minot
 
 
 
 
 
Using Stata for Survey Analysis 
Page 50 
 

The results in Example 23 indicate that food expenditure rises with per capita expenditure, that is 

high-income people spend more on food than low-income people. Because both food expenditure and 
per capita total expenditure are expressed in logs, the coefficient is the income elasticity of food. In 
other words, a 1% increase in income (per capita expenditure) is associated with a 0.47 percent 
increase in food expenditure. The fact that the elasticity is less than one means that food expenditure 
rises more slowly than total expenditure, so the share of expenditure on food falls as income rises.
In addition, rural households spend somewhat more on food than urban households, after controlling 
for other factors. The positive coefficients on reg_1 and reg_2 indicate that households in the west 
and center spend more on food than households in the east. The t statistic in the sexhead is just 0.14, 
smaller than 2, so we can say that the effect of sex of head of household on food expenditure is 
statistically insignificant. Finally, larger households spend less on food, holding other factors 
constant.
Example 23. Using “regress” to examine determinants of meat expenditure 
 
 
probit  
This command carries out a probit regression analysis of the specified variables. The syntax is:
probit depvar indepvars [if exp] [in range] [, options]
Probit analysis is used when the dependent variable is a binary variable (with only two values), such 
as whether or not a household is poor. An alternative is the dprobit command which reports the 
derivative of the probability with respect to each independent variable instead of the coefficient.
Examples include:

---