Analysis of attrition of business establishments in Uganda panel surveys
MetadataShow full item record
This study set-out to analyze attrition in economic panel surveys with specific emphasis on the factors associated with attrition in the Producer Price Index(PPI) Survey in Uganda. This was done using the PPI and Uganda Business Inquiry (2002, and 2008) datasets with key variables being time to attrition and status of attrition as failure variable. The dataset comprised 262 records of characteristics of business establishments, their management and time to attrition. Survival analysis techniques for interval censored data in R Software were adopted. The data were treated as right censored with the final time of observation as June 2015. The analysis was made for these ‘mixed case’ interval-censored survival data using nonparametric estimation of the survival function or Turnbull’s algorithm and finally a Cox proportional hazards model was fitted. It was established that the median survival time for a business establishment in the survey was 22 quarters (range:21-22 quarters)from the time of initiation. Attrition varied significantly by industry sector, location of the establishment and time of entry into the PPI survey (p<0.05). Particularly the business establishments in Greater Kampala had a 44% lower risk of attrition than those located up-country (HR= 0.56, p-value= 0.03), those that joined the PPI survey after 2004 had a 10%lower risk of attrition than those that were initiated in 2004 (HR= 0.90, p-value= 0.006), and those in the manufacturing sector have a 65% lower risk of attrition than those in the construction sector (HR=0.35, p-value= 0.006). However, the total number of staff employed and citizenship of managers of the business establishment were found insignificant (p>0.05). The main findings of this research effort suggest that the reasons for attrition of business establishments in the PPI survey lie within their own control. However, it is recommended that the Uganda Bureau of Statistics should rebase the producer price indices at least every five years, strengthen data collection in up-country business establishments and undertake in-depth qualitative investigation of the determinants of attrition.