IPUMS PMA Survey Notes

Survey Types

IPUMS PMA comprises several different type of surveys, collected by the same organization and using similar data collection techniques. A basic understanding of the structure and characteristics of each survey is essential for leveraging these data well. Below are descriptions of each file found on the IPUMS PMA online data dissemination system.

Household and Female surveys

The core of IPUMS PMA are the household and female surveys conducted by Performance Monitoring for Action (PMA) in 11 countries since 2013. These are cross-sectional surveys of sampled households and women of childbearing age (15 to 49) who reside in the household. The household respondent provides information about the wealth, water sources, sanitation facilities, and other housing characteristics. The female interview includes questions about family planning use, fertility preferences, fertility history, and other topics.

Data file structure: Each record is a person within a household. Women of childbearing age have household characteristics attached to their record. These files include nonrespondents to both the household questionnaire and female questionnaire. By default, the online dissemination system selects only female respondents, but users may choose to include household members and/or nonrespondents in their data extracts.

How to get these data: On the online data dissemination system, choose the Family Planning - Person unit of analysis.

Service Delivery Point (SDP) surveys

In parallel with the Household and Female surveys, enumerators interviewed a knowledgeable staff person at up to 3 public and up to 3 private health facilities in the same enumeration areas (EAID) as the household and female surveys in the same years. This allows researchers to use microdata to describe the service delivery environment in which the women live. See our user note on how to use SDP data with household and female data.

Data file structure: Each record is a health facility. In order to use these data with the household and female files, researchers need to calculate summary statistics at the EAID level and merge to a household and female file of the same country(ies) and round(s).

How to get these data: On the online data dissemination system, choose the Family Planning - Service Delivery Point unit of analysis.

Maternal and Newborn Health

In Ethiopia, PMA has collected panel data on maternal and newborn health in separate cohorts, and the survey design has changed over time.

2016-2017 Cohort

Enumerators conducted household roster surveys of all households in sampled enumeration areas to find households with pregnant women in the Southern Nations, Nationalities, and People's region of Ethiopia. Then, a female screening survey was conducted for each pregnant woman. Enumerators followed up in person at 7 days and 6 weeks after the woman's delivery, and then either by phone or in person at 6 months after delivery.

Data file structure: Each record is an infant (both live and stillbirth). There are several cases of twins, in which case the attributes of the mother and the household are attached on each infant's record. Therefore, in each file, mothers of twins will be represented twice. There are also a small number of cases of households with more than one pregnant women. These households will also be represented more than once.

2019-2021 Cohort

The longitudinal panel on women and infants was selected simultaneously with the 2019 Ethiopia Household and Female cross-sectional data sample on family planning on reproductive health. Women who were pregnant were asked whether they consented to participate in the panel and were asked baseline interview questions about the pregnancy, antenatal care, family planning, and other topics. Women who were up to 9 weeks postpartum at baseline were asked baseline and 6-week follow-up questions in the same interview. Enumerators then returned to interview women who had been pregnant at baseline for a 6-week follow-up. All women were then interviewed at approximately six months and one year post-partum.

The 6-week and 6-month follow-up questionnaires each have a pre- and post-COVID version. The variables SURVEYVERSION_6W and SURVEYVERSION_6M indicate whether the woman received the original or revised questionnaire version. The revised version included questions about whether COVID restrictions impacted their ability to receive post natal health care or vaccinations, changed the planned location of delivery, and other questions about how COVID affected them.

Data file structure: The 2019 cohort data include records for household members and the women who are part of the longitudinal panel. IPUMS has adjusted the data format such that each non-household member record is an infant (both live and stillbirth), or a panel member woman if the pregnancy ended in miscarriage or abortion. There are multiple cases of twins, in which case the attributes of the mother and the household are attached on each infant's record. Therefore, in each file, mothers of twins will be represented twice. There are also a small number of cases of households with more than one pregnant women. These households will also be represented more than once.

How to get these data: On the online data dissemination system, choose the Infant unit of analysis.

Nutrition

In 2017 and 2018, PMA conducted Nutrition surveys in Burkina Faso and Kenya, at both the person level and at the service delivery point level.

Household and Female

Data file structure: Each record is a person within a household. Enumerators collected household roster information from randomly selected households from sampled Enumeration Areas (EAs).

Round 1 (2017)

Households were eligible for the household survey if there was at least one female aged 10 to 49 and at least one child under age 2 living in the household. Females aged 10 to 49 were eligible for the female survey if they lived in a household with a child under the age of 2. Children under the age of 5 living in a household with a child under the age of 2 and a female aged 10 to 49 were eligible for the female-child nutrition survey. The female aged 10 to 49 answered the survey for the child under 5.

Depending on the child's age and other factors, each child was eligible for a different set of nutrition questions:

  • The first group was the youngest biological child in the care of the female under the age of 2 (these variables have a prefix INF and can be found in the variable groups Infant Nutrition and Infant Food Consumption).
  • The second group was children between age 2 and 5 who were the youngest child in the care of the female aged 10 to 49 (these variables have a prefix YK and can be found in the variable groups Young Child Nutrition).
  • The last group was children aged 2 to 5 who were not the youngest biological child in the care of the female (these variables have a prefix OTK and can be found in the variable groups Other Young Child Nutrition).

    • See the variable ELIGTYPE for more details. Users should note that children under 2 who were not the youngest child in the care of the female aged 10-49 were not eligible for the nutrition questionnaire.

    The original data was disseminated in three files - the household roster, female survey, and children nutrition survey. IPUMS restructured these data into a single data file using the following steps:

    1. Attached data from the female survey onto each of the women's records on the household roster using a individual female ID number (FCQINSTID) provided in the original data.
    2. Characteristics from the female's record was merged onto the child's record also using the female ID number, which represented the mother or caretaker of the child.
    3. Using children's birth month, birth year, and household identifier, the data from the child survey was attached to the corresponding record of the child in the household roster. This step is important because the child's gender was on the household roster, but not the Child file.
    4. Attached household level characteristics (water sources, wealth, etc) to children from the Child file that were not matched to records in the household roster.
    5. PMA had also constructed records for children who had died, according to the responses from the female-child survey. These records contained the birth year and death year of the child. IPUMS attached this data to the woman's record and omitted the records of the deceased children from the file.
    Round 2 (2018)

    All households were eligible for the household survey. All children under age 5 were eligible for the child survey. The primary caretaker of the child answered the questions for the child. Twenty-five percent of the households surveyed (see variable FQSELECTED) for the household roster were randomly selected for the female survey, even if a female aged 10 to 49 was not a part of the household. Females age 10 to 49 in these selected households were eligible to respond to the female survey. The data in this round were disseminated in one file per country. Unlike Round 1, the mother or caretaker's identification number is not present on the child record, so it is not possible to attach mother's characteristics onto the child's record.

    How to get these data: On the data dissemination system, choose the Person - Nutrition unit of analysis.

    Service Delivery Point

    Data file structure: Each record is a health facility. In order to use these data with the household and female files, researchers need to calculate summary statistics at the EAID level and merge to the Nutrition household and female file of the same country and round by EAID. See our memo about using household and service delivery point (SDP) data together.

    How to get these data: On the data dissemination system, choose the Service Delivery Point - Nutrition unit of analysis.

    COVID-19 surveys

    The PMA COVID-19 survey is a follow-up telephone survey administered to women who participated in an in-person baseline survey for a broader panel study. This baseline survey was collected between November 2019 and Februrary 2020 - prior to the appearance of COVID-19 in most countries.

    When the outbreak of COVID-19 grew into a global pandemic in the spring of 2020, PMA representatives partnered with the Ministries of Health in DRC, Kenya, Burkina Faso, and Nigeria to design a shorter - approximately 30 minute - survey responding directly to the effect of COVID-19 on women and their households.

    Women were selected for the baseline survey if they were age 15-49 and resided in a household screened at random from a sample cluster represented by EAID. All women surveyed at baseline where eligible to participate in the COVID-19 follow-up, provided that they 1) agreed to the interview, and 2) owned or had access to a telephone.

    You'll find survey weights adjusted for the probability that a given woman had access to a telephone recorded in the new variable CVQWEIGHT. This weight is normalized for the target population of each sample (note that two of the samples are not nationally representative):

    You'll find more detail about the construction of PMA COVID-19 survey weights here. For information about response rates for each sample, check out sample-specific Dataset Notes.

    To link female records between the baseline surveys and the COVID-19 surveys, drop household members who are not eligible women (use CONSENTFQ) and merge by FQINSTID. If you are using Stata, use the following code to trim trailing zeros and change the format of FQINSTID in order to merge:

    
recast str50 fqinstid
replace fqinstid = trim(fqinstid)

    Client Exit Interviews

    The client exit interview surveys were introduced in 2019 in select countries. Interviewers approached women leaving family planning clinics to ask about the services they received during their visit. Only women aged 15-49 were eligible to be interviewed, and are only interviewed once in a sample. Most questions on the survey were only given to women who had received either information on family planning or a method. Questions include whether the woman received her preferred method, whether advantages or disadvantages of certain methods were discussed, and how the visit experience was (length wait time, polite staff, etc).

    The sampling frame for client exit interviews begins with the random sampling of primary sampling units called EAs, which are small geographic areas of about 200-300 households. Households are sampled from these EAs for the household and female family planning surveys, and up to three public and three private health facilities in these areas are also visited for the service delivery point surveys. A subset of these facilities that provide family planning services were chosen if they reached a certain threshold of clients per day, and interviewers approached women who left the facility after an appointment to complete a client exit interview.

    The data in the client exit interviews can be attached to the service delivery point (SDP) data in the same year and country using the variable FACILITYID. For some countries, there is a follow up survey for the respondents within a 6 month period. These data will be available in 2022.

    How to get these data: On the data dissemination system, choose the Person - Client Exit Interview unit of analysis. The default case selection is for women aged 15-49. When selecting All Cases, you include observations of women who were not eligible for interview.

    Longitudinal Family Planning Panel

    The initial survey design of Performance Monitoring for Action was a high-frequency, cross-sectional survey on family planning, water, and sanitation indicators. Beginning in 2019, PMA refocused on reproductive and sexual health indicators, and adjusted the survey design to add a contraceptive calendar and a longitudinal panel of women of childbearing age to observe contraceptive and fertility dynamics.

    Burkina Faso, Kenya, Nigeria, and the Democratic Republic of the Congo began fielding their baseline panel survey in 2019. Baseline surveys for Cote d'Ivoire, Uganda, and India were fielded in 2020. PMA refers to the panel in terms of phases. Phase 1 is the baseline sample, Phase 2 is the first follow-up, and Phase 3 is the last follow-up. Each phase is around one year apart.

    Data file structure: See our guide on longitudinal data file structures.

    How to get these data: Navigate to our online data dissemination system and choose the Family Planning topic under the Person unit of analysis. Then click on "Select Samples". You will now have the ability to view either the cross-sectional samples or the longitudinal samples using radio button selections.

    See our tutorial video on the longitudinal data for more details on how to access cross sectional or panel data from the online data dissemination system.



    In addition to following up with the panel women over time, PMA also adjusted the sampling so that a cross-sectional sample could still be produced, enabling the calculation of an uninterrupted time series of representative indicators. Additional households were sampled to maintain national or regional representativeness if more than 10 percent of households left the sample after the first round.

    Which weight to use: For analyses involving women from multiple phases of the longitudinal panel, see the weighting guide. For cross-sectional sub-samples, use FQWEIGHT.

    Sample Composition of Cross-Sectional and Longitudinal Samples

    Source: PMA Usernotes

    For countries and years in which a longitudinal panel was conducted, there are cross-sectional and longitudinal versions of that survey that contain mostly the same individuals. Only households identified as the cross sectional sample in the original data are included in the cross sectional version of the sample (CROSS_SECTION). The longitudinal data files include all households and females interviewed in each Phase, but some panel women and households have been removed from the cross-sectional version of the data files in order to achieve a representative sample.

    Household types (see HHTYPE), as identified in Phase 2 data:

    The Phase 2 data also include women recruited for the panel who were not interviewed in Phase 1 if they aged into eligibility (age 15), if they moved into a sample household, or if they were a resident in a replacement cross-section dwelling (see HHREPLACED).

    Female response types include:

    PMA created a flow chart for each country to demonstrate the attrition rate of women and households.

    Loss to Follow Up

    Women could leave the panel if they or their entire household moved out of the primary sampling unit area (EA), moved to a household that was not sampled within the EA, aged out (turned 50 years old), did not consent to the follow-up survey, died, or could not be found. Women could enter the panel at Phase 2 if they were in a sampled household and turned 15 between Phase 1 and Phase 2, they moved into a sampled household, or their household was selected to be added to the panel to replace households lost from attrition. The following bullet points describe the scenarios in which a woman could be lost from the panel:

    Household identifiers: Because households move between dwellings, disband, or change their composition, PMA constructed multiple household identifiers to retain as much information as possible.

    Linked Service Delivery Point Files

    PMA's sampling design for surveys conducted between 2013 and 2018 collect data from households and service delivery points from the same enumeration area (EAID, the primary sampling unit) for four rounds in a row before randomly selecting a new set of EAs. Often, the same facilities were sampled in multiple rounds. The variable FACILITYID identifies the facility within rounds that were conducted in the same set of EAs.

    Beginning with the redesign in 2019, PMA refocused on longitudinal data collection with purposeful repeated interviews of the same facilities over time.

    IPUMS has linked the SDP data using FACILITYID for rounds where the survey was conducted in the same set of EAs. Users should note that for a small number of matched cases (less than 4% of total cases), linked records of facilities had different EAID codes over time. Users should also expect a match rate of about 50% to 80% for most samples, especially those before the survey design explicitly returned to the same facilities for multiple survey rounds.