How Survey-to-Survey Imputation Can Fail

This paper proposes diagnostics to assess the accuracy of survey-to-survey imputation methods and applies them to examine why imputing from the Household Income and Expenditure Survey into the Labor Force Survey fails to accurately project poverty trends in Sri Lanka between 2006 and 2009. Survey-to-survey imputation methods rely on two key assumptions: (i) that the questions in the two surveys are asked in a consistent way and (ii) that common variables of the two surveys explain a large share of the intertemporal change in household expenditure and poverty. In addition, differences in sampling design can lead validation tests to underestimate the accuracy of survey-to-survey predictions. In Sri Lanka, the causes of failure differ across sectors. In the urban sector, the primary culprit is differences between the two surveys in the design of the questionnaire. In the rural and estate sectors, the set of common variables in the prediction model does not adequately capture changes in poverty. The paper concludes that in Sri Lanka, survey-to-survey imputation between the Household Income and Expenditure Survey and the Labor Force Survey cannot produce accurate poverty estimates unless the Labor Force Survey adds additional questions on assets and is redesigned to use a questionnaire that is compatible with the Household Income and Expenditure Survey. Alternatively, a new welfare-tracking survey that satisfies these conditions could be established.

Bibliographic Details

Main Authors:	Newhouse, D., Shivakumaran, S., Takamatsu, S., Yoshida, N.
Language:	English en_US
Published:	World Bank Group, Washington, DC 2014-07
Subjects:	AVERAGE WAGES, BIASES, CALCULATION, CHANGES IN POVERTY, CONFIDENCE INTERVALS, CONSUMPTION DATA, CONSUMPTION EXPENDITURE, CONSUMPTION EXPENDITURES, DECLINE IN POVERTY, DRINKING WATER, DUMMY VARIABLES, EMPLOYMENT INCOME, EMPLOYMENT STATUS, ESTIMATES OF POVERTY, FOOD CONSUMPTION, HOUSEHOLD BUDGET, HOUSEHOLD CONSUMPTION, HOUSEHOLD DEMOGRAPHICS, HOUSEHOLD EXPENDITURE SURVEYS, HOUSEHOLD HEAD, HOUSEHOLD HEADS, HOUSEHOLD INCOME, HOUSEHOLD SIZE, HOUSEHOLD SURVEY, HOUSEHOLD SURVEYS, HOUSEHOLD WELFARE, HOUSING, INCOME GROWTH, INEQUALITY, LIVING STANDARDS, NATIONAL POVERTY, NATIONAL POVERTY RATE, PER CAPITA CONSUMPTION, POOR, POOR PROVINCES, POVERTY ANALYSIS, POVERTY ASSESSMENT, POVERTY DATA, POVERTY ESTIMATES, POVERTY INDICATOR, POVERTY LINES, POVERTY MAPPING, POVERTY MAPS, POVERTY MEASUREMENT, POVERTY MEASURES, POVERTY RATE, POVERTY RATES, POVERTY REDUCTION, PRECISION, PREDICTION, PREDICTIONS, PROBABILITIES, PROBABILITY, REDUCTION IN POVERTY, REDUCTION OF POVERTY, REGIONAL DIFFERENCES, REGIONAL LEVEL, REGIONAL LEVELS, REGIONAL PERSPECTIVE, RELIABILITY, RURAL, RURAL AREAS, RURAL POPULATION, RURAL POVERTY, RURAL PUBLIC, RURAL SECTOR, RURAL SECTORS, SAMPLE DESIGN, SAMPLING ERRORS, SELF-EMPLOYMENT, STANDARD DEVIATION, STANDARD ERRORS, UNEMPLOYMENT, VILLAGE LEVEL, WAGE INCOME, WAGE RATES, WAR, WELFARE INDICATORS, WELFARE MONITORING,
Online Access:	http://documents.worldbank.org/curated/en/2014/07/19754254/survey-to-survey-imputation-can-fail https://hdl.handle.net/10986/19364
Tags:	Add Tag No Tags, Be the first to tag this record!