The effect of poor vision on economic farm performance: Evidence from rural Cambodia
The data shows 9 variables. 0. Household ID 1. Age (Age in years) 2. Education (Years in school) 3. Gend (Male =0, Female =1) 4. Area cult (Total area of cultivation in hectares for all plots that belong to the farm) 5. Householdsize (Number of people living in the household) 6. Net_profit (Gross margins of all produce valued at average product prices minus cost for seeds, fertilizer, insecticides, fungicides, herbicides, machine hours, land and costs for hired labor for all cropping activities (transplanting, weeding, application of agrochemicals, harvesting and irrigation. Relates to the growing season 2017-2018. All values are transformed to USD/year) 7. Eyesight Calculates the results from Landolt C-Test, classifying respondents into “poor vision” and “good vision”. The threshold is an average visus on both eyes≥0.7 8. Eyesight: Upper bound comparison We shift the threshold of assignment to the “good vision” group to a visus≥0.75 9. Eyesight: Lower bound comparison We shift the threshold of assignment to the “good vision” group down to a visus≥0.45 Hypothesis: Farmers with poor vision have lower gross margins. In the paper we conduct various types of propensity score matching and mahalanobis distance matching techniques to test the hyptothesis. First, we develop a model that predicts vision status of farmers. This is a propbit model that includes variables 1-5 as predictors. This model gives us the propensity score which is used in a second step to match farmers with similar propensity score and different vision status. With this approach, we can decrease the bias in our sample. As robustness checks we code the variable Eyesight in 3 different ways (variables 7-9). Our findings show that farmers with good vision have gross margins that are around 630 USD/year higher than those with poor vision.