290 likes | 428 Views
Individual-Disaggregated Questionnaire Design & Data in LSMS-ISA: Too Many Shades of Grey?. Talip Kilic Living Standards Measurement Study Team Poverty & Inequality Group Development Research Group The World Bank.
E N D
Individual-Disaggregated Questionnaire Design & Data in LSMS-ISA: Too Many Shades of Grey? Talip Kilic Living Standards Measurement Study Team Poverty & Inequality Group Development Research Group The World Bank Presented at the 3rd annual Living Standards Measurement Study – Integrated Surveys on Agriculture (LSMS-ISA) project workshop at Zanzibar Beach Resort Hotel, Zanzibar, Tanzania, February 13-14, 2012
“While a great deal has been learned about what works and what does not when it comes to promoting greater gender equality, the truth remains that progress is often held back by the lack of data…” World Bank World Development Report 2012 Gender Equality and Development
Why Collect Individual-Disaggregated Data? • Analysis of welfare or productivity at the individual-level can signal whether there is a need for interventions targeted to specific groups or individuals within households • Analysis of intra-household resource allocation through individual-disaggregated data important for setting strategic priorities & understanding how policy interventions affect household members differently • Certain phenomena, such as gender equality in a given economic or social dimension, can be appropriately analyzed only if individual-level data are available
Motivational Insights from the WDR 2012 • Effective domestic public action to bring about gender equality “sometimes hinges on the production of a public good, such as the generation of new (global) information or knowledge” • Better & widely available gender-disaggregated data & comprehensive gender diagnostics crucial for evidence-based public action • Need for country-specific diagnostic studies also driven by cross-country variations in the nature of markets, institutions & social norms
How can LSMS-ISA help? • Nationally-representative • Multi-country, high-degree of heterogeneity in the nature of markets, institutions & social norms across project countries • Panel survey design • Multi-topic framework with a strong focus on agriculture, where the availability & quality of individual disaggregated data are particularly poor • LSMS traditionally focused on individual/gender-disaggregated data collection, as also reflected in the current set of LSMS-ISA instruments Yet… • Review of LSMS-ISA questionnaires & available data highlight several areas that could be improved & standardized across countries at a relatively low cost to enable richer analyses of intra-household resource allocation & gender-disaggregated outcomes
Paving the Path for Collecting More Comprehensive & Standardized Individual-Disaggregated Data • Agriculture • Ownership, Management vs. Control, Extension • Livestock • Wage Employment • Non-Farm Self Employment • Social Assistance • Remittances • Other Income • Durable Asset Ownership
Qx Review: Agriculture –Ownership * Information solicited only if parcel has a certificate.
Qx Review: Agriculture –Management • ** Determine joint vs. sole decision making, then solicit information accordingly. • *** Who decided what to plant on [PLOT]?
Qx Review: Agriculture - Control • ** Determine joint vs. sole decision making, then solicit information accordingly. • **** Information collected at the parcel-plot-crop-level.
Qx Review: Remittances • ** Level of Obs: Remitting-individual in 2008/09; No Module in 2010/11 • ***Level of Obs: Recipient Other Income
Building a Case… • … for collecting more comprehensive information on intra-household resource allocation in a standardized fashion across LSMS-ISA instruments
Case Study # 1 • We revisit the link between woman’s control of income & child nutrition using data from the Malawi IHS3 • Estimate the impact of woman’s share of household income on height-for-age z-scores for boys 6-59 months of age relative to girls in the same cohort • The primary shortcomings of the existing studies are that they either • focus on a specific component of HH income OR • compute a total HH income aggregate but assign the control of income among individuals based on questionable assumptions • Contrast our results with those that would be obtained by assigning control of income among household members based on assumptions that are commonly found in the literature
Data • Malawi Third Integrated Household Survey (2010/11) • Sample: 2,522 Children (6-59 Months) • Biologically implausible/potentially outlier z-scores excluded • Of Malawian couples with multiple children (household fixed-effects) • Definitions of woman’s share of household total income • V1: Each income component, w/the exception of wage employment, assigned 100% to HH Head; wage income assigned to earners • V2: Each income component, w/the exception of wage employment, split 50/50 between head & spouse; wage income assigned to earners • V3: Replication of the approach of Haddad & Hoddinott (HH), 1994; 1995 • V4: Preferred definition that incorporates available ownership/control info • Current income definition excludes livestock & fishery net revenues • Farm self-employment income = Aggregated plot-level net revenues
Methodology • Linear regression with household fixed-effects, controlling for time-invariant unobservable household heterogeneity yih = c0 + µ Maleih +γ FSI*Maleih + β Childih+ ρParentalih + νh + ξih • i & hDenote child & household, respectively • yihZ-score of height for age • c0 Constant • MaleDummy variable identifying male child • FSI*Male Woman’s share of household income interacted with Male • Child Vector of child age & birth order categories • Parental Vector of parental characteristics • ν Household fixed-effects • ξ Error term
Case Study # 2 • We revisit the evidence on gender differences in agricultural productivity in sub-Saharan Africa (SSA) by using detailed plot-level agricultural data collected within a multi-topic framework as part of the Malawi IHS3 • In targeting sustainable poverty gains through smallholder-based agricultural growth, national development plans have emphasized the alleviation of gender differences in agricultural productivity • Increased productivity among female farmers argued to result in: (i) poverty alleviation through positive impact on overall smallholder productivity growth, & (ii) improvement of nutritional outcomes
Shortcomings of Current Literature on Gender Differences in Agricultural Productivity in SSA • Household-Level Analyses • Plot-Level Analyses • Data do not identify plot managers within households • Main explanatory variable: Gender of head of household • Concerns regarding the plausibility of assumptions in sub-Saharan African settings • Plot-level outcomes linked to managers/owners • Cohabitation of male & female managers/owners allows for estimation of within-household differences However… • Rich agricultural data often not solicited in multi-topic framework • External validity concerns: Small/non-representative samples • Can management & ownership be used interchangeably?
Methodology ln(Yih)= c0 + β Femaleih + µ Mih +γ Pih + ρln(Iih )+ δln(Lh )+ νh + ξih • i & h denote agricultural plot & household, respectively; • Y – KGs of maize production/ha; • Female • identifies a plot managed by a female in Specification 1. • is replaced in Specification 2 with two indicator variables identifying (i) exclusively female owned, and (ii) mixed male & female owned plots. The comparison category is exclusively male owned plots. • M – vector of variables on other characteristics of the plot manager; • P – vector of variables plot physical characteristics; • I – vector of variables capturing the extent of non-labor input use; • L – vector of variables capturing household & hired labor input; • ν – household fixed-effects; • Only plots from HH cultivating multiple maize plots inform the estimates • ε & c0 – the error and constant term, respectively. […]
Conclusions • Management is not synonymous with ownership • Extent of data collection on intra-household allocation of resources in LSMS-ISA instruments encouraging but not all-encompassing • Modules allowing for the computation of a total household income aggregate largely exist, however unevenness persists in terms of whether individuals that keep/decide on the use of earnings from different sources are identified • Within existing structures, benefits likely to considerably outweigh the costs of including an additional question identifying individuals that decide on the use of earnings from each possible source
Topics for Discussion • Reactions? • Lingering differences in the way that plot managers are identified (wording, single vs. joint-management, etc…): Easy fix? • To complement information on parcel owners & plot managers, Uganda is the only country that identifies (up to 2) individuals that • hold use rights to parcels that are rented in (parcel-level) • decide on the use of the crop output (parcel-plot-crop-level) • Unevenness across questionnaires in terms of the availability of module(s) on other income receipts • Convenient to assume wage income is in the control of earner? • Ethiopia/Nigeria: How has been the experience of identifying household members that own durable/farm assets?
Individual-Disaggregated Questionnaire Design & Data in LSMS-ISA: Too Many Shades of Grey? Talip Kilic Living Standards Measurement Study Team Poverty & Inequality Group Development Research Group The World Bank Presented at the 3rd annual Living Standards Measurement Study – Integrated Surveys on Agriculture (LSMS-ISA) project workshop at Zanzibar Beach Resort Hotel, Zanzibar, Tanzania, February 13-14, 2012