Mixing Qualitative and Quantitative Methods of Analyzing Poverty Dynamics

1
Mixing Qualitative and Quantitative Methods of
Analyzing Poverty Dynamics

• Chris Barrett
• Cornell University
• March 11, 2003
• SAGA Workshop
• Nairobi, Kenya

2
Imperative of Poverty Research

• Governments committed to
• (i) poverty reduction and
• (ii) increased participation of the
• poor in priority setting exercises.
• Need to know
• - who are the poor?
• - whose poverty is chronic, whose transitory?
• - why are some chronically poor?
• - what one-off interventions can help put people
  on an accumulation path out of poverty?

3
Advances in Poverty Research

• Significant recent progress in both qualitative
  (QUAL) and quantitative (QUANT) methods of
  poverty analysis
• - rapid rise of participatory poverty assessment
  (PPA) methods
• - emergence of widespread, nationally
  representative household survey data, notably
  longitudinal panels

4
Ongoing Challenge of Poverty Research
Poverty as a complex, multidimensional concept
- Static/dynamic - chronic/transitory - assets
vs. income/expenditures - econ/non-economic
outcomes - experience/prospect of poverty -
outcomes/processes

Powerlessness, vulnerability and resource
insufficiency central to most conceptualizations
but hard to pin down
5
So are QUAL and QUANT complements or
substitutes?Considerable conflict among
practitioners of each legacy of 80s/90s
disciplinary segregation. But do the methods
necessarily conflict too???So complex a concept
requires iteration between (sequential mixing) or
integration of (simultaneous mixing) methods for
accurate triangulation. We must learn the
lesson of the blind men and the elephant.
6
Be clear about focus of question(1) Data
collection methods(2) Data types (3) Data
analysis methods(4) Audience
Dimensions of QUAL-QUANT Difference
7
Data collection methods
Dimensions of QUAL-QUANT Difference
Analytical Coverage General
Specific
Census
Random Sample Surveys
PRA
Autobiography
Passive Active Population Involvement in
Research
8
Dimensions of QUAL-QUANT Difference

• Data types
• Qualitative Quantitative
• Categorical Ordinal Cardinal
• Each data collection method can yield both
  non-numerical and numerical data types

9
Dimensions of QUAL-QUANT Difference

• Data analysis methods
• Qualitative Quantitative
• Inductive Deductive
• Related to the specific-general data collection
  methods distinction, theres often (not always) a
  difference in analysis methods.

10
Dimensions of QUAL-QUANT Difference

• Audience
• Local community
• Or global/national policymakers
• Local empowerment
• or the big picture and speaking truth to power

11
Myths about QUAL-QUANT differences

• (1) One more/less extractive than the other
  (ethical superiority)
• (2) One more/less contextual than the other
  (historical superiority)
• (3) One inherently numerical/non-numerical
• (statistical superiority)
• (4) One more rigorous than the other
• (scientific superiority)
• Bad practice is bad practice, whatever the
  method...
• Key question When and how is good practice
  within one strand still wanting? How can the
  other fill the blanks?

12
Mixing Methods

• Improve analysis by mixing the two take the
  con out of econometrics and generalize beyond
  the part of participatory methods
• Harness statistical power of quantitative methods
  for description/aggregation along with narrative
  power of qualitative methods for nuanced and
  textured analysis of complex (unmeasurable?)
  concepts
• Triangulation to uncover mechanisms behind poverty

13
Mixing Methods

• Sequential mixing or classical integration
• Practitioners of each method do their best with
  their own tools on same problem, sometimes taking
  outputs from one as intermediate inputs to
  another. Then triangulate to get an integrated
  result.

14
Sequential mixing

• Example Understanding welfare transitions
  (BASIS CRSP)
• Step 1 Panel survey data collection to construct
  transition matrices and change measures.
  Poort Nonpoort
• Step 2 Draw several households Poort1
• from each of 6 cells in matrix
• and do detailed oral histories. Nonpoort1
• Why? Capture omitted variables, check
  transitions, 2nd method of inference, problem of
  identifying thresholds econometrically, value of
  stories for policy audiences.

15
Mixing Methods

• Simultaneous mixing or Bayesian integration
• Iterative approach to using one method to inform
  another, then back to the first, etc., keeping
  multiple methods interactive throughout the
  research process to update researchers priors
  continuously.
• Feedback loop yields a homeostatic research
  mechanism
• ethnography precedes participatory which in
  turn precedes survey in dictionary ought to
  be the case in the field, too!
• Ongoing, creative tension between methods helps
  ensure originality, robustness and relevance of
  results

16
Simultaneous mixing

• Example improving pastoralists risk management
  (PARIMA)
• (1) Participatory risk mapping to identify
  relevant threats open-ended, spatially-explicit,
  pseudo-cardinal
• (2) Quarterly repeated surveys with open-ended
  sections and mixed modules
• (i) complex property rights climate
  forecasting, resource conflict land use history
  livelihoods strategies, etc.
• (ii) complementarity at multiples levels of
  analysis and different methods (e.g., livestock
  marketing with data from households, markets and
  traders)
• Effective means to ensure inference consistent
• (i) across methods (a test of robustness) and
• (i) with local understandings of the problem(s)
  (a test of relevance)

17
The Way Forward Walking On Two Legs

• Development scholars and practitioners
  increasingly recognize the complementarity
  between qualitative and quantitative methods.
• Best done in multidisciplinary teams with a
  range of joint and parallel efforts.
• Situate participatory poverty appraisals within
  sampling frame
• Introduce more open-ended and subjective
  questions of causality and interpretation into
  survey instruments
• Explicitly foster self-critique, cross-checking
  within teams and feedback briefings with
  different interested parties
• Nonetheless, much remains to be done (vocabulary,
  data cross-referencing, respectful dialogue, etc.
  

18
Thanks very much for your comments and questions!