Jump to ContentJump to Main Navigation
Q-SquaredCombining Qualitative and Quantitative Approaches  in Poverty Analysis$

Paul Shaffer

Print publication date: 2013

Print ISBN-13: 9780199676903

Published to Oxford Scholarship Online: September 2013

DOI: 10.1093/acprof:oso/9780199676903.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2017. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 25 February 2017

(p.119) Appendix 1: Q2 Research Designs and Methods

(p.119) Appendix 1: Q2 Research Designs and Methods

Source:
Q-Squared
Publisher:
Oxford University Press

The purpose of this appendix is to provide a quick reference guide to the main approaches to poverty analysis reviewed in Chapters 4, 6, and 7 of this book. The order of the studies follows the format of these chapters, beginning with the identification stage of poverty analysis before proceeding to the causal stage. Summary information on the methods, data sources, and value-added are presented.

We also include in the summary table information on the purpose of the Q2 design using Jennifer Greene and colleague’s (1989) typology discussed in Sections 1.3 and 8.2. Four of their five listed purposes of mixed-method research are relevant, namely development, triangulation, complementarity, and expansion. To recall, development refers to the use of methods from one approach to assist in the methodological development of another through, say, using focus groups to better structure the wording of fixed response surveys. Triangulation uses different methods to investigate the same phenomenon to assess, and/or bolster, the validity of research results. Complementarity relies on different methodological approaches to clarify, elaborate upon or better interpret the results of one method with those of another. Finally, expansion refers to the use of different methods to address related, but distinct, components of an overall research question such as the combined analysis of outcomes and processes.

A.1 Identification: Who are the Poor and What are their Characteristics?

A.1.1 Operationalizing Dimensions of Poverty

The studies in this section attempt to integrate locally meaningful definitions of poverty in a way which facilitates interpersonal comparisons and, in many cases, external validity of results. A common design is to combine information from dialogical inquiry with household survey data or to transform data numerically, or statistically, to facilitate interpersonal comparability.

(p.120)

Authors

Data sources

Purpose (Greene et al.. 1989)

Description and Q2 value-added

Barahona and Levy (2007)

PRAs Household surveys

Development

Locally meaningful definitions of poverty related to food insecurity, drawn from Participatory Rural Appraisals (PRAs), were included into a probabilistically sampled household survey. The value added of Q2 was to (i) include locally meaningful definitions of deprivation; (ii) enhance comparability of results over different populations by using the same definition of poverty; (iii) achieve external validity of results through probabilistic sampling.

Sharp and Devereux (2004)

Group discussion Household surveys

Development

Group discussions among teams of ‘qualitative’ and ‘quantitative’ researchers operationalized the concept of destitution, which was subsequently included into a probabilistically sampled household survey. The value added of Q2 was to (i) include locally meaningful definitions of destitution; (ii) enhance comparability of results over different populations by using the same definition of destitution; (iii) achieve external validity of results through probabilistic sampling.

Howe and McKay (2007)

PPAs household surveys

Development

Characteristics of chronic poverty uncovered in a Participatory Poverty Assessment (PPA) with national scope were ‘mapped’ onto similar indicators included in a nationally representative household survey. The value added of Q2 was to (i) rely on locally meaningful characteristics of chronic poverty; (ii) enhance comparability of results over different populations by using common correlates of chronic poverty; (iii) achieve external validity of results through probabilistic sampling.

Hargreaves et al. (2007)

PPAs

Development

Data on the characteristics of the poor from wealth rankings exercises were transformed numerically on the basis of the frequency with which they were mentioned, allowing for the calculation of a household wealth index. The value added of Q2 was to (i) rely on locally meaningful information on characteristics of poverty; (ii) enhance comparability of results over different populations by numerically transforming this information.

Campenhout (2006)

PPAs

Development

Well-being ranking results from PPAs were adjusted statistically by controlling for village and sub-village effects which could bias interpersonal comparisons. The value added of Q2 was to (i) rely on locally meaningful definitions of poverty reflected in well-being rankings; (ii) enhance comparability of results over different populations by statistically adjusting this information.

Beegle et al. (2009)

Household surveys

Development

Household survey respondents were asked to situated themselves on a six-step well-being ladder and respond to a series of ‘vignettes’ in which they placed four hypothetical families on the same ladder. Data from the vignettes were then used, inter alia, to rescale responses to the first question to facilitate interpersonal comparability. The value added of Q2 was to (i) rely on people’s perceptions about the level of their own well-being; (ii) enhance comparability of results over different populations by statistically adjusting this information; (iii) achieve external validity of results through probabilistic sampling.

(p.121) A.1.2 Weighting Dimensions of Poverty

The studies in this section aim to rely on locally meaningful weighting scheme to determine the relative importance of various dimensions of poverty. A core distinction is between indirect approaches, which rely on econometric analysis of correlates of well-being rankings, and direct approaches in which respondents are asked to provide the weights directly.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Kebede (2009)

PPAs Household Surveys

Development

Well-being ranking results from PPAs and observable household characteristics from household surveys were combined with econometric analysis of the correlates of different ranking groups. The authors interpret the regression coefficients as an indication of the social value, or weight, afforded characteristics such as income, assets, land, number of adults and housing. The value added of Q2 was to used statistical analysis of household survey data to infer, weights used in well-being rankings.

de Kruijk and Rutten (2007)

Household surveys

Development

A nationally representative household survey was administered which asked respondents to rank twelve dimensions of well-being in terms of their perceived priority. Rankings were averaged separately for men and women and relative weights calculated for use in the Human Vulnerability Index. The value added of Q2 was to (i) rely on people’s perceptions about the relative importance of dimensions of well-being; (ii) achieve external validity of results through probabilistic sampling.

Woodcock et al. (2009)

Household survey

Development

The development of the Wellbeing Research in Developing Countries Quality of Life Questionnaire (WEDQoL) in Thailand began with an exploratory phase where respondents were posed open-ended questions about the sources of happiness. Results were subsequently codified into 51 items in the fixed response WEDQoL, and respondents asked to rate them in terms of their perceived necessity for well-being. The core value added of Q2 was to rely on people’s ratings of the relative importance of dimensions of well-being to elicit weights.

Esposito et al. (2012)

Focus group discussions Household surveys

Development

Focus group discussions were held to elicit a short-list of highly valued literary practices. Such practices were then included in a household survey which respondents were asked to simultaneously rank by allocating fifty beans amongst them. Unlike sequential ranking in the two aforementioned studies, simultaneous ranking forces respondents to jointly valuate the importance of all five practices. The core value added of Q2 was to rely on people’s simultaneous ratings of the relative importance of literary practices.

(p.122) A.1.3 Setting Poverty Thresholds

Q2 analyses have attempted to set the poverty line at a point which reflects locally meaningful thresholds. The studies in this section present three different approaches based on data discontinuities, conceptual thresholds and the ‘consumption adequacy’ question.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Hargreaves et al. (2007)

PPA

Development

Drawing on PPA data, analysis was conducted to determine if particular statements were overwhelmingly made about the poor, very poor and other well-being groups. Such ‘data discontinuities’ were detected by visual inspection and used to distinguish thresholds between well-being groups. The core value-added of Q2 was to base poverty thresholds on local perceptions of the characteristics of poverty.

Barahona and Levy (2007)

PPA Household survey

Development

There is a built-in ‘conceptual threshold’ associated with the idea of ‘not having enough food’ used as the poverty measure in the household survey, drawing on results of a prior PPA. The core value added of Q2 was to base food security thresholds on local perceptions of adequacy.

Pradhan and Ravallion (2000)

Household survey

Development

A consumption adequacy question (CAQ) was posed in households surveys, whereby respondents we asked if their level of consumption (food, housing, clothing, etc.) is more than, less than, or just adequate to meet family needs. By regressing responses to the CAQ on consumption expenditure, subjective consumption poverty lines were calculated. The core value added of Q2 was to base the consumption poverty line on people’s perceptions of consumption adequacy.

(p.123) A.2 Causal Analysis of Poverty Status and Dynamics

A.2.1 Determinants of Poverty Status—I: Combining Outcomes and Processes

The studies in this section illustrate one of the major contributions of Q2 research to causal analysis, namely to combine analyses of outcomes and processes. Such analyses have improved, or broadened, aspects of the causal framework, including causal variables, weights, mechanisms and the causal ‘tree’, while also directing attention to issues of external validity, as discussed in Section 6.1.

Authors

Data Sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Place et al. (2007)

Ethnographies Household surveys

Complementarity Expansion

Examination of the effect of agricultural technologies on the poor which combined an analysis of outcomes, drawing on panel data from household surveys, with processes relying on semi-structured interviews and participant observation. The value added of Q2 was to (i) distinguish between outlier and tendency cases; (ii) help interpret the meaning of variables in the household survey; (iii) probe the reasons for the testing and adoption of new technology; (iv) explain counter-intuitive results from the household survey. The Q2 design enhanced or facilitated the understanding of causal variables, mechanisms and the causal tree and allowed for an assessment of the external validity of the ethnographic results

Woldehanna et al. (2005)

Household surveys Semi-structured interviews

Complementarity Expansion

Econometric analysis of household survey data on correlates of child schooling and labour was followed-up by semi-structured interviews to provide a richer understanding of the statistical results. The core value added of Q2 was to explain the reasons for counterintuitive econometric results and, in the process, provide a fuller account of causal mechanisms and the causal tree.

(p.124) A.2.2 Determinants of poverty status—II: the rural livelihoods approach

The series of studies in this section rely on the sustainable livelihoods framework whereby ‘forms of capital’ or assets are transformed into livelihoods via mediating processes related to social relations, institutions, and so on. Q2 research has been used to highlight different elements of this overall analytical framework.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 Value-added

Ellis and Freeman (2004)

PRAs Household surveys

Complementarity Expansion

Household survey data on assets, incomes, shocks and livelihood activities were combined with information from the PRAs on mediating processes, related primarily to institutions, to enrich the overall analysis. The core value added of Q2 was to explain the reasons for some of the descriptive statistical findings from the household surveys and, in the process, provide a fuller account of causal mechanisms and the causal tree.

A.2.3 Determinants of poverty dynamics—I: interviewing the transition matrix

Poverty dynamics, or the flows of households into and out of poverty, can be represented in terms of a poverty transition matrix, which distinguishes between households who remain poor, escape from poverty, enter into poverty, and remain non-poor. All of the Q2 studies in this section have contributed to a fuller explanation, and better understanding, of this transition matrix.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Barrett et al. (2006)

Household Surveys Oral histories Semi-structured interviews

Complementarity Expansion

Econometric analysis was conducted to test for the existence of poverty traps and detailed case studies subsequently undertaken of households within the transition matrix to uncover the reasons behind well-being trajectories. The core value added of Q2 was to integrate causal weights and mechanisms with a more detailed account of the causal tree.

Baulch and Davis (2008)

Household surveys Life histories

Complementarity Expansion

Descriptive statistical and econometric analysis of panel data were combined with life histories of households in different quadrants of the transition matrix to provide a richer depiction of trajectories of change. The core value added of Q2 was to integrate causal weights and mechanisms with a more detailed account of the causal tree.

Adato et al. (2006, 2007)

Household surveys Oral histories Semi-structured interviews

Complementarity Expansion

Similar to Barrett et al. (2006), econometric analysis was conducted to test for the existence of poverty traps and detailed case studies subsequently undertaken of households to uncover the reasons behind well-being trajectories. The core value added of Q2 was to integrate causal weights and mechanisms with a more detailed account of causal variables and the causal tree.

Authors

Data Sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Krishna et al. (2006a) Krishna et al. (2006b)

Focus group discussions Semi-structured interviews

Development Expansion Triangulation

Following identification of the main reasons for well-being trajectories, logistic (logit) regression models were estimated of the likelihood of falling into, or escaping, poverty using the previously uncovered variables. The core value added of Q2 was to integrate causal weights and mechanisms with a more detailed account of causal variables.

Krishna and Lecy (2008)

Focus group discussions Semi-structured interviews

Development Expansion Triangulation

A ‘net events’ variable, which is the difference between positive and negative events or ‘reasons’, was calculated and included in econometric analysis of the determinants of well-being transitions. The core value added of Q2 was to integrate causal weights and mechanisms with a more detailed account of causal variables and the causal tree.

(p.125) A.2.4 Determinants of poverty dynamics—II: the ‘Stages of Progress’ approach

Causal analysis using the Stages of Progress (SoP) methodology entails, first, situating households within the poverty transition matrix using recall, second, eliciting reasons for escape from, and descents into, poverty, and, third, modelling ‘reasons’ econometrically (Krishna 2010b). The Q2 contribution in the studies in this section has been to identify causal variables and to assign causal weights.

(p.126) A2.5 Model specification

The studies in this section provide examples of how various types of narrative information have proved useful for purposes of econometric modelling, by facilitating selection of causal variables and specification of the causal tree.

Authors

Data Sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Rao et al. (2003)

Focus group discussions Household surveys

Development

Results of focus group discussions facilitated the selection of an ‘instrumental variable’ for inclusion in an econometric model of the effects of condom use on revenue among sex workers. The core value added of Q2 was to aid in the specification of causal variables.

de Weerdt (2010)

Focus group discussions Life histories Household surveys

Development Complementarity Expansion

Econometric analysis, conducted to predict 2004 asset values on the basis of 1993 household characteristics, was combined with dialogical methods to explain why certain households had ‘defied their economic destiny’. The dialogical methods facilitated identification of an interaction variable between remoteness and initial conditions which was subsequently used in econometric modelling. The value added of Q2 was to aid in the specification of causal variables and their interrelationships, or the causal tree, along with an understanding of the underlying causal mechanisms at work.

Quisumbing (2011)

Focus group discussions Household surveys

Development Expansion

Focus group discussions facilitated the specification of variables which were subsequently included in a household survey, along with the construction of interaction terms which were included in the econometric modelling. The core value added of Q2 was to aid in the specification of causal variables and understanding of the causal tree.

(p.127) A.3 Causal analysis in impact assessment

A.3.1 Combining results and mechanisms

The studies in this section illustrate how Q2 research designs have allowed for the combined examination of the magnitude of impact, or ‘results’, and the underlying reasons, or ‘mechanisms’.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Adato (2008)

Semi-structured interviews Participant observation Household Surveys

Complementarity Expansion

Experimental and quasi-experimental (regression discontinuity techniques) analyses of household survey data were combined with the results of ethnographic research to present a combined study of the magnitude of, and reasons for, impact.

Broegaard et al. (2011)

Focus group discussions Semi-structured interviews Household surveys

Complementarity Expansion

Quasi-experimental analysis (propensity score matching) of household survey data was combined with narrative information from semi-structured interviews and focus group discussions to present a combined study of the magnitude of, and reasons for, impact.

Authors

Data Sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Ravallion (2001)

Narrative information Household surveys

Development

A chance encounter between Ms Analyst and Ms Sensible Sociologist reveals information on the determinants of programme participation, specifically budgetary allocation to school boards. A budget allocation variable could be subsequently used to estimate a model of programme participation and as an ‘instrument’ in a model of schooling.

Rao and Ibáñez’s (2005)

Focus group discussions Semi-structured interviews Household survey

Development

In the first stage matching, narrative information was used to select five similar communities to those included in the study incorporating ‘unmeasured’ variables such the number of churches and youth groups, and ‘unobservables’ such as political culture and social structure. The authors maintain that use of these additional sources of information contributed to minimising the problem of selection bias.

Authors

Data Sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

Shaffer (2012)

Household surveys

Triangulation Development

Propensity score matching drawing on outcome data from household surveys was combined with perceptual information on how respondents ‘would have acted in the absence of the program’. The two analytical techniques came to similar results with respect to health and school fee exemptions, thereby bolstering the validity of research results through triangulation.

(p.128) A.3.2 Identifying comparison groups

The studies in this section illustrate how Q2 approaches, specifically various forms of narrative information, are used to assist in the formation of comparison groups.

A.3.3 Conducting counterfactual thought experiments

In this study, a counterfactual scenario was constructed by way of a though experiment relying on perceptual data from household surveys and compared with results of a propensity score matching exercise.

Authors

Data sources

Purpose (Greene et al. 1989)

Description and Q2 value-added

De Silva and Gunetilleke (2008)

Focus group discussions Household survey

Complementarity Expansion

Impact results drawing on variables in household survey data, such as access to toilets, water and energy, were compared with impact indicators identified by focus group participants. Satisfaction with changes in the former were off-set by dissatisfaction with such factors as the loss of a quiet rural environment and land devoted to paddy cultivation.

(p.129) A3.4 Defining benefits

This study compares changes in standard indicators of project impact with those identified by participants in focus group discussion. Accordingly, Q2 analysis is used to address the metric which should be used to gauge programme success or failure. (p.130)