Myths structure our perceptions of history and of
reality by providing guiding narratives that help
to create, define and bond communities, most
typically by giving them an origin and a destiny.
Patricia Fara.1
On his New York Times blog in 2008, Steven Levitt dubbed Robert Jensen “the
Indiana Jones of economics” and, with that persona, Jensen recounted how he and
his colleague Nolan Miller had succeeded in finding “an elusive fabled prize shrouded
in mystery [and] travel to far-off lands”. They had been searching for a “verified
example” of a “violation of one of the most sacred and holy laws in economics”, the
possibility of which had “excited and intrigued economists for over a century”. Other
economics blogs were also enthusiastic, if perhaps not quite so effusive, about the
quest.2
The occasion for the excitement was a paper by Jensen and Miller, published
in 2008 by the American Economic Review, reporting the results of a field experiment
which they characterised as providing “the first rigorous, empirical evidence of real -
world Giffen behaviour”. The experiment, which involved subsidising the purchases
of rice and wheat by poor households in China, removed the “minor embarrassment“
that economists had been unable to provide convincing evidence for their theoretical
argument that the “Law of Demand, while descriptively valid in many situations, may
not apply to the very poor facing subsistence concerns.” Although Alfred Marshall
had “first publicized this idea in the 1895 edition of his Principles of Economics” using
the example of bread, and a discussion of Giffen behaviour could be found in
“virtually every basic economics course”, empirical confirmation had been lacking for
Marshall’s “conjecture”. If subsequent work had indicated that the “demand for
neither wheat nor bread was upward sloping in Britain during Marshall’s time”, the
“standard textbook example of a Giffen good, potatoes during the Irish potato famine
of 1845 – 1849”, first published in Paul Samuelson’s Economics, was now
“discredited”. More recent laboratory experiments on the topic were of little use as
their conditions were “far removed from reality”, while other econometric work
suffered from specification problems. Jensen and Miller concluded, however, that
“the absence of previously documented cases most likely results from inadequate
1
Epigraph: Fara 1999, p.168. Any emphases in material quoted here appear in the original
texts.
2 Levitt 2008; Jensen 2008. For the blog commentary, see Economist 2007; Moroney 2007;
Rodrik 2007; Thoma 2007.
2
data or empirical strategies.” More generally, as neoclassical economic theory was
the only analysis that could predict the behaviour of the Giffen phenomenon, their
results also provided a vindication of the theory which “in recent years … has come
increasingly under attack” [Jensen and Miller 2008, pp.1553, 1554, 1558, 1576].3
Jensen and Miller thus presented their paper as the denouement of a long search for
empirics that would support Marshall’s conjecture. It was precisely because the
search had been marred by ‘inadequate data or empirical strategies’ that “Giffen
behavior has long played an important, though controversial role in economic
pedagogy, as well as in the history of economic thought” [Jensen and Miller 2008,
p.1554]. Their analysis, however, actually undermined the picture of a single
research project driven by the attempt to establish whether an initial conjecture had
empirical validity. The Jensen and Miller results did not provide evidence for a “Giffen
good” where, at least over a range, the market demand curve slopes upward.
Following the previous secondary literature, Jensen and Miller note that the effects of
the behaviour of the very poor could be swamped in a market, so their results were
evidence of “Giffen behaviour” and not of a Giffen good. Indeed, because Jensen
and Miller rejected the possibility of identifying such a market demand curve, their
paper, in effect, sounded the Last Post for that project. In an analysis that is
symptomatic of the more recent experimental turn in economics, the relevance of
Giffen behaviour was now depicted in terms of devising welfare programs for the
poor [Jensen and Miller 2008, pp.1558, 1575 – 76]. If Jensen and Miller were not
clear about the significance of that point, some of the blog commentary claimed that
they had identified a demand curve or, at least, a Giffen good.4
More careful excavations at the sites of Marshall and Samuelson have unearthed a
series of anomalies that further problematise the history of a single Giffen research
project in economics. First, Marshall did not discuss the possibility of a market
demand curve with a positive slope in the Principles. Moreover, the subsequent
3
Jensen and Miller list the following requirements for Giffen behaviour. Households must be
“poor enough that they face subsistence nutrition concerns”; their diet consists of “a basic
(staple) and a fancy good”; the basic good provides the “cheapest source of calories”,
accounts for “a large part of the diet/budget, and has no ready substitute” [Jensen and Miller
2008, p.1556]. This is formally consistent with the usual textbook account where the basic
good must be inferior and the income effect dominates the substitution effect.
4
That commentary may have been influenced by the way that Jensen referred to the quest
for a ‘Giffen good’ that violated ‘the Law of Demand’ [Jensen 2008].
3
discussion in Marshall’s Memorandum on Fiscal Policy of International Trade (1908)
contradicted the account in the Principles (Section 1 below). Marshall’s references to
the statistician Sir Robert Giffen (1837–1910) as the source for his different accounts
are also problematic. It is an odd feature of the Jensen and Miller paper that, while
they often refer to Giffen behaviour (or goods), there is no explicit reference to
Giffen.5
Hence they make no mention of the second anomaly, which is that there is
no evidence that Giffen ever made the claim(s) that Marshall attributed to him.6
Indeed, it will be shown here that Giffen rejected a key assumption in Marshall’s
Memorandum (Section 2). The third anomaly is that, by the mid-1920s, discussion of
an upward-sloping demand curve attached no particular significance to an illustration
referenced by Marshall’s Principles because other and quite different explanations
were regarded as just as, if not more, important. That focus was replicated in the
early editions of Samuelson’s Economics, so that it was not until the early 1960s that
Giffen and the Irish famine were introduced in that text as the only possible
illustration for a demand curve with a positive slope (Section 3). Examination of these
anomalies shows that there have been different histories of Giffen behaviour,
conceptualised in different ways and constructed for different purposes since 1895.
If the case of Samuelson’s Economics, for example, provides a striking example of
shifting textbook rhetoric, it also illustrates how the 1930s ‘transformation’ of
neoclassical economics as “a general logic of choice” [Hicks 1934, p.54] restricted its
explanatory domain.