Change in household and member weights of monthly series
by Mahesh Vyas
Household weights associated with all sample households in the monthly income and monthly expenses series since 2014 when the Consumer Pyramids Household Survey series begin will be changed on September 29, 2020. Monthly member weights used in member incomes data will also change from 2014. All old data files containing monthly income and monthly expenses data will be updated with the new weights. From October 1, all data of sample households of monthly income and monthly expenses and monthly member incomes data will reflect the new weights.
On average, the new weights of households in the monthly income and expenses series are 0.85 per cent lower than the old household weights. And, on average, the new weights for members in the monthly member income series are 0.55 per cent lower than the old member weights in monthly income data.
Links to the time series of old and new weights for the various strata are available at the end of this note. These are for your reference.
We apologise for the inconvenience the change in weights may have caused you. The least that we can do is that if you had downloaded the monthly income or monthly expenses data during your current subscription period, then these will be reversed. This will be done immediately.
The change has been necessitated because of the following reason. Hitherto, household weights associated with monthly income and expenses data reflected the estimated population of household as of the end of the survey execution period. The survey execution period to collect income and expenses data for all households and their corresponding members in the sample is four months. The existing algorithm for assigning weights to the sample observations of a month used the estimated population at the end of such a survey execution period.
However, it is more appropriate to use the estimated population of households at the end of the month for which data is collected rather than at the end of the month when the survey execution is completed. Although the survey execution is relatively short (four months), there is no justification for the weights to be set with respect to the estimated population at the end of the execution period.
To explain the above by way of an example, data for the month of say, January is collected over the months of the following four months, February, March, April and May. Note that the data on income during a month can be collected after the end of the month. Hitherto, the estimated population of households as of the end of May was used to derive the weights for the sample households and their members in January. Now, the estimated population of households and members in January would be used to derive the weights for sample households with the January income and expenses data.
Given that the population of households four months ahead is generally larger than in any month, the weights used so far were inflated correspondingly. The new weights are therefore lower than the earlier weights. The exception to this is weights for Sikkim rural region. The population of this region is projected to decline. Weights for this region is therefore revised upwards while those of other strata are systematically revised downwards.
The change in weights does not change any other data in the CPHS delivery. Also note that the weights of the Waves data do not change. These were and continue to be associated with end-of-Wave population estimates. The new weights, which are as of end-of-month would now be aligned with the approach used in creating Waves weights.
Strata-wise full time-series of old and new monthly weights: txt file
Strata-wise full time-series of old and new monthly weights: csv file