Ingo Rohlfing

People familiar with the development of qualitative methods know that
process tracing has developed rapidly over the last years. As the
discussion about a method becomes broader and deeper, it becomes more
important to systematize and sort the field in order to understand what
it can and cannot achieve. Beach and Pedersen (BP) have left a mark on
the field with their comprehensive [[[book on process
tracing]{style="color:#3366ff;"}](http://www.press.umich.edu/2556282/process_tracing_methods){target="_blank"}]{style="color:#3366ff;"}
and argue, among other things, that there are three varieties of process
tracing. In my view, this is incorrect and ignores an important variant
of process tracing that can be done and has been implemented a great
many times in the empirical literature.

BP distinguish between *case-oriented* and *theory-oriented* process
tracing, which is a useful distinction (and corresponds to case-centered
and theory-oriented research in [[[my
book]{style="color:#3366ff;"}](http://www.palgrave.com/page/detail/case-studies-and-causal-inference-ingo-rohlfing/?k=9780230240704){target="_blank"}]{style="color:#3366ff;"}).
Case-oriented research aims to develop a comprehensive explanation for
one substantively interesting case. The two theory-oriented variants are
theory-testing and theory-building. What theory-testing process tracing
is need not be addressed in more detail here. The problem lies in the
idea of theory-building process tracing which is taken as equivalent to
exploratory research. It is common to treat theory-building research and
exploratory research as being synonymous, but they are not because there
are two variants of exploratory research.

We need to distinguish between process tracing that builds theory or, on
a smaller scale, a hypothesis from scratch and process tracing that
modifies an existing hypothesis. The latter variant is commonly referred
to as puzzle solving and is one of the [[[classic means of motivating
empirical
research]{style="color:#3366ff;"}](http://www.amazon.com/Political-Interests-Identities-Institutions-Comparative/dp/0472087231/ref=sr_1_1?ie=UTF8&qid=1403695551&sr=8-1&keywords=grofman+puzzle){target="_blank"}]{style="color:#3366ff;"}.
By and large, a hypothesis works but there are some cases that it fails
to explain and that represent the puzzle one aims to resolve by
modifying it. In contrast, genuine hypothesis-building research
addresses a heretofore ignored or new phenomenon such as a country's
complete abandonment of nuclear plants or policy-making about
pre-implantation diagnostics. We do not solve a puzzle here because
there is no theory in light of which the new phenomenon could constitute
a puzzle.

One might wonder what the big deal is about the fourth variant because
these seem to be subtleties concerning ways in which to systematize
process tracing. However, there is more to this because process tracing
aiming to build and modify a hypothesis is built on different types of
cases that call for different case selection strategies (see chapter 3
of my book). This can be easily demonstrated by bringing classic types
of cases to the table. Two types of cases that are considered
particularly interesting for exploratory research are failed most-likely
and passed least-likely cases (related to theory-infirming and
theory-confirming cases in [[[Lijphart's seminal 1971
article]{style="color:#3366ff;"}](http://www.jstor.org/stable/1955513){target="_blank"}]{style="color:#3366ff;"}).
A failed most-likely case does not pass a test it should have passed; a
passed least-likely case masters a test we did not expect to be
successful.

It is apparent that these two types serve to modify a hypothesis (some
strategies of hypothesis modification are described in chapter 4 of my
book). It is equally apparent that genuine exploratory research
concerned with new empirical phenomena cannot be a failed most-likely
case or a passed most-likely case because there is no hypothesis we can
test in the first place (a similar argument holds for deviant cases).
The two types of cases we can use are typical cases and diverse cases,
as they only require information about the outcome, which is usually
available in exploratory research.

The distinction between case-oriented and theory-testing process
tracing, and confirmatory and exploratory process tracing thus is not
all that we need to derive all possible types. Doing justice to
exploratory process tracing and its proper implementation additionally
demands it to distinguish between the analysis of new phenomena and the
modification of an existing hypothesis that works well, but not
perfectly.


People familiar with the development of qualitative methods know that process tracing has developed rapidly over the last years. As the discussion about a method becomes broader and deeper, it becomes more important to systematize and sort the field in order to understand what it can and cannot achieve.

Rogue Scholar Beiträge

Three varieties of process tracing? It’s four