Distributional Differences in Upper Tail Observation of Per capital Household Expenditure in Tanzania in 2001 and 2007

Recent economic growth in Tanzania has not been matched by a corresponding reduction in poverty when compared with similar episode of growth in countries such as Uganda, Ghana, etc. This has raised heated debate, whereby some analysts argue that problem lies on the GDP is compiled, and other on the validity of Household Budget Survey (HBS) data. This paper contributes to that debate by  analyzing the distribution of upper tail observation using he Generalized Pareto Distributions show that HBS 2007 is characterized by extreme values of per capita expenditure when compares to 2001. However, this feature is not captured by analyses confined to inequality indexes such as Gini coefficient and General Entropy measures. Given these findings and the documented increase in sample replacement rate in 2007 household budget survey, the paper discusses the policy implications on how to manage the sampling and replacement of rich households in Household Budget Surveys