Microsoft PowerPoint - the NCRM EPrints Repository

advertisement
What is Analytic Induction?
Martyn Hammersley
The Open University
NCRM Research Methods Festival,
St Catherine’s College, Oxford, July 2012
Why should we be interested
in analytic induction?
• It has historical significance: it was
closely associated with the Chicago
School of Sociology.
• Appeals are still made to it as a
distinctive approach to qualitative or
case study research.
• Above all, it highlights some important
issues that all social science faces,
qualitative work especially.
A historical sketch
• Florian Znaniecki’s The Method of Sociology.
• In this there are echoes of the accounts of
scientific method to be found in J. S. Mill and
Aristotle.
• The classic applications of AI are the work of
Lindesmith on opiate addiction, Cressey on
financial trust violation, and Becker on
marihuana use
More recent examples that appeal to this
approach include: Bloor 1978; Katz 2001a;
Gilgun 1995, and Rettig et al 1996.
What is AI?
• It is a distinctive conception of the proper
form of social scientific method.
• It claims to specify what is required if we
are to achieve reliably sound conclusions
about the causes of some type of outcome.
• It is aimed at identifying the necessary and
jointly sufficient conditions that must be
met for the type of outcome concerned to
occur.
The steps of AI
1. Initially specify the outcome to be explained.
2. Collect data on a small number of cases in
which the outcome is present.
3. Identify what these cases share and formulate
a hypothesis on that basis.
4. Collect data on further cases.
5. If one is found that does not fit the hypothesis,
then either reformulate that hypothesis or redefine the outcome.
6. Continue until data from new cases no longer
force any revisions.
An example
Cressey (1953) started by studying people who
had been convicted of embezzlement and
equivalent offences, looking for factors they
shared in common. From this he developed a
hypothesis, but deviant cases were found so
that he had to revise it. At one point he
reformulated the type of case he was trying to
explain, excluding those people who had
taken on positions of financial trust specifically
to steal money, focusing solely on those
already in positions of financial trust who had
subsequently decided to ‘borrow’ money.
With what does AI contrast?
• Statistical inference from trends in a
sample to conclusions about general
patterns of causal relationship.
• Theoretical speculation that does not
engage with the details of particular
cases.
• Descriptive accounts of particular cases
that infer causal relationships without
testing these through comparative
analysis.
How does AI differ from
Grounded Theorising?
• It is aimed at generating an explanation for
a quite specific type of outcome.
• It is explicitly designed to test as well as to
develop explanations/theories.
• It includes the possibility of redefining what
is to be explained, i.e. the outcome.
The curious co-existence of AI and GT in the
development of Chicago Sociology (see
Hammersley 2010a).
How does AI compare with
Qualitative Comparative Analysis?
• Both are concerned with developing and
testing explanations for a quite specific type of
outcome.
• Both tend to produce explanations involving
configurations of causal factors.
• AI is developmental in character in a way that
QCA generally is not.
• AI does not involve systematic checking of all
combinations of relevant causal factors.
Variation in interpretations of AI
• Whether it starts from pre-conceived
hypotheses (but, if so, how is it inductive?).
• Whether it involves searching for negative
cases. Or, more precisely, whether it
examines all the types of case that must be
investigated if a reliable conclusion is to be
reached about the necessary and jointly
sufficient factors causing a type of outcome.
• Whether the goal is to produce universal laws
or more restricted forms of generalisation.
AI as starting with a hypothesis
‘Analytic induction is unlike other
qualitative approaches since it begins
with a pre-existing theoretical viewpoint
or premise that guides the investigator’s
approach to the cases that are
examined (Gilgun, 1995; Miller,1982)’
(Rettig et al 1997:208).
Does AI look for cases where the
candidate causal factor is present
but the outcome is absent?
If not, it can only discover necessary
conditions, at best.
Different products?
• Becker (1998:194-212) has distinguished
between ‘classical’ and ‘less rigorous’ forms.
• Various recent commentators have referred to
‘modified analytic induction’ (Bogdan & Biklen
1992:69-72), which is aimed at developing
‘descriptive hypotheses that identify patterns
of behaviors, interactions, and perceptions’
(Gilgun 1995:269), or ‘working hypotheses
and concepts that illuminate other similar
situations’ (Rettig et al 1997:209), rather than
‘universal’ generalisations.
The issues that AI raises
• What is the form of an adequate scientific
explanation of any social phenomenon?
• What is required if we are to develop such an
explanation?
• What kind of causal relations operate in the
social world, and is it possible for the
strategies that social scientists currently
employ to discover them?
• The key issue of discovering the conditions
under which causal relations operate: revising
the explanandum.
The nature of causation
AI relies upon a very strong
interpretation of causation, in which a
cause = a set of necessary and jointly
sufficient conditions that must be met
for some outcome to occur.
We need to consider whether this is
appropriate.
Necessary and jointly
sufficient conditions
• Necessary = what must be present if the
outcome is to occur
• Sufficient = anything that, if present, ensures
that the outcome will occur, though it need
not be present for the outcome to occur.
• For example, the presence of flammable
materials can be regarded as a necessary
condition for a fire to occur, while a spark can
be treated as a sufficient condition in these
circumstances.
Necessary (flammable material)
and sufficient (spark) conditions
for fire
Presence of
flammable
materials
Fire
Spark
Alternative interpretations of
causation
• Identifying a significant sufficient
condition
• Identifying a significant necessary
condition
• Identifying a condition that typically
increases the chances of some
outcome of interest occurring
Conclusion
• Analytic induction is an instructive model
for thinking about causal analysis.
• It illustrates that qualitative researchers
have been and can be engaged in this
enterprise.
• While there may be questions about
whether it is a viable model for many
areas of social inquiry, it raises very
important methodological issues.
References
Becker, H. S. (1998) Tricks of the Trade, Chicago, University of Chicago Press.
Bloor, M. (1978) ‘On the analysis of observational data: A discussion of the worth and uses of inductive
techniques and respondent validation’, Sociology, 12, 3, pp545-52.
Bloor, M. and Wood, F. (eds.) Keywords in Qualitative Methods, London, Sage.
Bogdan, R. and Biklen, S. (1992) Qualitative Research for Education, 2nd ed. Boston MS, Allyn and Bacon.
Cooper, B. et al (2012) Challenging the Qualitative-Quantitative Divide, London, Continuum, Ch.5.
Cressey, D. (1953) Other People’s Money, Glencoe ILL, Free Press.
Denzin, N. (2006) ‘Analytic induction’, in Ritzer, G. (ed.) Blackwell Encyclopedia of Sociology, Oxford Blackwell.
Gilgun, J. (1995) ‘We shared something special’, Journal of Marriage and Family, 57, pp265-81.
Gilgun, J. (1992) Qualitative Methods in Family Research, Thousand Oaks, Sage. (see pp22-40)
Goldenberg, S. (1993) ‘Analytic induction revisited’, Canadian Journal of Sociology, 18, 2, pp161-76
Hague, B. (2008) ‘Précis of “An abductive theory of scientific method”’, Journal of Clinical Psychology, 64, 9,
pp1019-22.
Hammersley, M. (1989) The Dilemma of Qualitative Method, London, Routledge.
Hammersley, M. (2008) Questioning Qualitative Inquiry, London, Sage.
Hammersley, M. (2010a) ‘A historical and comparative note on the relationship between analytic induction and
grounded theorising’, Forum: Qualitative Social Research, 11(2), 2010, Art. 4, Available at: http://nbnresolving.de/urn:nbn:de:0114-fqs100243
Hammersley, M. (2010b) ‘Aristotelian or Galileian? On a Puzzle about the Philosophical Sources of Analytic
Induction’, Journal for the Theory of Social Behaviour, 40, 4, pp393-409.
Hammersley, M. (2011) ‘On Becker’s studies of marihuana use as an example of analytic induction’, Philosophy of
the Social Sciences, 41, 4, pp535 – 566.
References Contd.
Johnson, P. (1998) ‘Analytic induction’, in Symon, G. and Cassell, C. (eds.) Qualitative Methods and Analysis in
Organizational Research, London, Sage.
Katz, J. (2001a) ‘Analytic induction revisited’, in Emerson, R. (ed.) Contemporary Field Research, Second edition,
Prospects Heights ILL, Waveland Press.
Katz, J. (2001b) ‘Analytic induction’ in N. J. Smelser and P. B. Baltes (eds.) International Encyclopedia of the
Social and Behavioral Sciences, Oxford, Elsevier
Lindesmith, A. (1968) Addiction and Opiates, Chicago, Aldine.
Lindesmith, A. ‘Symbolic interactionism and causality’, Symbolic Interaction, 4, 1, pp87-96.
Manning, P. (1982) ‘Analytic induction’, in Smith, R., Manning, P. (eds.) Handbook of Social Science Methods:
Qualitative Methods, Cambridge MS, Ballinger.
Miller, S. (1982) ‘Quality and quantity: Another view of analytic induction’, Quality and Quantity, 16, 4, pp281-95.
Pascale, C-M (2010) Cartographies of Knowledge, Thousand Oaks CA, Sage.
Patton, M. Q. (2002) Qualitative Research and Evaluation Methods, 3rd edition, Thousand Oaks CA, Sage.
Ragin, C. (1994) Constructing Social Research, Pine Forge Press.
Ratcliff, D. ‘Analytic induction as a qualitative research method of analysis’. Available at (accessed 1.5.12):
http://www.mieonline.org/home/articles/23/1/Analytic-Induction-as-a-Qualitative-Research-Method-ofAnalysis--Donald-E-Ratcliff/Page1.html
Rettig, K., Tam, V. and Magistad, B. (1997) ‘Using pattern matching and modified analytic induction in examining
justice principles in child support guidelines’, Marriage and Family Review, 24, 1/2, pp193-22:
Robinson, W. S. (1951) ‘The logical structure of analytic induction’, American Sociological Review, 16, 6, pp812818.
Turner, R. (1953) ‘The quest for universals’, American Sociological Review, 18, 6, pp604-11.
Znaniecki, F. (1934) The Method of Sociology, New York, Farrar and Rinehart
Download