COVID-19 High Frequency Phone Survey of Households, Indonesia, 2020-2021 (ICPSR 38463)

Time Period(s) View help for Time Period(s)

Date of Collection View help for Date of Collection

Data Collection Notes View help for Data Collection Notes

Study Purpose View help for Study Purpose

The purpose of this study is to provide insights into the socio-economic impact of COVID-19 on households in Indonesia.

Sample View help for Sample

Three stages of sampling strategies were applied. For the first stage, districts (as primary sampling unit (PSU)) were selected based on probability proportional to size (PPS) systematic sampling in each stratum, with the probability of selection being proportional to the estimated number of households based on the National Household Survey of Socio-economic (SUSENAS) 2019 data. Prior to the selection, districts were sorted by provincial code. In the second stage, villages (as secondary sampling unit (SSU)) were selected systematically in each district, with probability of selection being proportional to the estimated number of households based on the Village Potential Census (PODES) 2018 data. Prior to the selection, villages were sorted by sub-district code. In the third stage, the number of households was selected systematically in each selected village. Prior to the selection, all households were sorted by implicit stratification, that is gender and education level of the head of households. If the primary selected households could not be contacted or refused to participate in the survey, these households were replaced by households from the same area where the non-response households were located and with the same gender and level of education of households' head, in order to maintain the same distribution and representativeness of sampled households as in the initial design.

Time Method View help for Time Method

Unit(s) of Observation View help for Unit(s) of Observation

Data Type(s) View help for Data Type(s)

Mode of Data Collection View help for Mode of Data Collection

Description of Variables View help for Description of Variables

This study covers the following topics:

• Knowledge and behavior (Rounds 1, 3)
• Employment and income loss (Rounds 1, 3, 5, 6, 7)
• Food security (Rounds 1, 2, 3, 4, 5, 6, 7)
• Access to health services (Rounds 2, 3, 4, 5, 6, 7)
• Digital transactions (Rounds 2, 4)
• Education (Rounds 2, 4, 7)
• Coping mechanisms (Rounds 2, 4, 6, 7)
• Concerns/Subjective Welfare (Rounds 2, 4, 5)
• Social safety-nets (Rounds 1, 3, 5, 6)
• Household roster (Rounds 1, 2, 3, 4, 5, 6, 7; Full updates only in R1 and R4)

Response Rates View help for Response Rates

The actual sample of households in the first round was 4,338 households or 85 percent of the 5,100 target households. However, the response rates in the following rounds are higher than expected, making the sampled households successfully interviewed in Round 2 4,119 (95% of Round 1 samples), and in Rounds 3, 4, 5, and 6, 4,067 (94%), 3,953 (91%), 3,686 (85%), and 3,471 (80%) respectively. The number of balanced panel households up to Rounds 3, 4, 5, and 6 are 3,981 (92%), 3,794 (87%), 3,601 (83%), 3,320 (77%) respectively.

Original Release Date View help for Original Release Date

2022-10-24

Weight View help for Weight

Household weights were first calculated independently by each initial survey and then combined altogether afterward. For this approach to be properly applied without potential bias, there should not be overlapped survey areas across different surveys. The household weights were calculated for both cross-section for each round and panel for all rounds of the survey. In each round of the survey, the initial sampling weight was calculated following the original sampling method of the survey from which the sampled households were drawn. A sampling weight trimming using the mean and standard deviation of the weights was then conducted to reduce weight variability. In particular, the weight trimming was applied to some outlier weights (only a small proportion of the samples), while keeping the total of the weights remaining the same. Afterwards, the weights were calibrated using a raking method to ensure the total estimates of the households with respect to designated variables were comparable with the population estimates of those variables from the National Household Survey of Socio-economic (SUSENAS) 2019. The designated variables included region (DKI Jakarta, Java Non-DKI Jakarta Urban/Rural, Outside Java Urban/Rural), gender of household's head, and level of education of household's head (junior secondary and lower, senior secondary, and tertiary). The attrition occurred when respondents were not able to be interviewed, which was mostly because their phones were unreachable or unanswered. A test for whether attrition was random showed that the dropped households were not associated with key households' characteristics, such as household head age, gender, education, region (DKI Jakarta, Java-non DKI Jakarta, and Outside Java), and wealth status. However, there was a weak association between households' participation and the area where they reside, as households in urban area were less likely to participate in the follow-up surveys than those in rural areas. The difference in participation rates between urban and rural samples was taken into account in the survey weight calculation. Therefore, for analysis requiring panel households, attrition bias is not a concern when interpreting changes between rounds.