Analysis of unmatched data using propensity scores: cross-section analysis

Analysis of unmatched data using propensity scores: cross-section analysis

Using stratification of propensity for information technology and income

How can we derive the relationship between “ownership of information and communication technology” (ICT) on the one hand, and income defined as a multiple of the poverty line on the other hand? This study provides evidence of the impact of ICTs on poverty for a deliberately selected sample of sites from the poorest areas in four African countries (Kenya, Rwanda, Tanzania and Uganda).

The conducted survey for that purpose was initially designed to interview randomly selected individual family members in each one of the four countries. However, the paper underscores that the interviewed individuals were the one who happen to “be at home” at the time of the interview (i.e. biased data). Accordingly, it was important to correct for such a departure using available statistical means.

The authors claim that the method based on stratification of propensity scores was successful in reducing bias in the baseline covariates, allowing for a fair assessment of the partial effect of ICT ownership on the income. Then again, the document provides a simple intuitive method to deal with missing data typical in this type of studies.

The study demonstrates two conclusions:

  • it is clear that the relationship is definitely positive but its magnitude is subject to change depending on the controls that were used
  • the generalisation of the results can only be made relative to the (poor) areas included in the study or relative to areas of similar characteristics.