Computers were provided to randomly selected districts participating in a household survey in Vietnam to assess the impact on data quality of entering data within a day or two of completing the interview rather than several weeks later in the provincial capital. Provision of computers had no significant effect on the observed distribution of household expenditures and thus no effect on measured poverty. Provision of computers reduced the mean number of errors per household by 5 -23 percent,

more »

... nding on the type of error. Given the already low rate of errors in the survey, however, the goal of increasing the precision of the estimated mean of a typical variable can be achieved at a much lower cost by slightly increasing the sample size. Provision of additional computers did substantially reduce the time interviewers spent adding up and checking the data in the field, with the value of the time saved close to the cost of purchasing desktop computers. JEL Classification: C81, C93, C42, I32, O15 Household survey data are used for many policy and research purposes in developed and developing countries. Yet anyone who works with household survey data soon realizes that they can have many errors and inconsistencies. Statisticians call such errors nonsampling errors. Economists call them measurement errors. Reducing such errors in household surveys should lead to better research and better policies. Personal computers are a useful tool for this purpose. Data from survey questionnaires can be entered into personal computers using