‘Research Methods’ 91
Let me give an example. Suppose your project concerns food wastage due to the
limitations of distribution networks, such as, for example, the loss that occurs when
a shipping container of vegetables spends an extra day in a warehouse. This can
be caused by factors such as: the vegetables should not have been shipped there in
the first place; there is no truck or driver to distribute the vegetables after shipping;
there is no capacity to receive the vegetables at the retail outlets. You have pro-
posed a mathematical model that (in principle) a grocery chain could use to try and
reduce such loss, by, for example, integrating driver rosters with data extrapolated
from recent sales. As a measure, you use the tonnage of vegetables that are thrown
away due to their having reached a sell-by date; that is, you deem your model to be
a success if it can predict a schedule that reduces the weight of vegetables that are
discarded.
This measure sounds plausible, and is emotionally appealing because the idea
of throwing away tons of vegetables seems such a waste. But do you think that
it is sufficient to simply use ‘tonnage discarded’ without some rationale, or with-
out some consideration of the alternatives? I hope that you agree with me that the
answer is no! It certainly isn’t good enough to use a measure on the grounds that
it has some emotive impact. To see what I mean, consider some of the alternative
measures (not all of which you will agree with, but I think you will see that some
kind of argument could be made for each of them): the retail value of the discarded
vegetables; the energy saved by not delivering vegetables that won’t be sold; or
the value of the vegetables that can be discarded at point of production instead of
point of sale (allowing them to be used, say, to feed pigs or as compost, rather than
landfill). To take the first of these, an economist could argue that it is reasonable to
discard cheaply grown, cheaply harvested local vegetables in preference to expen-
sive vegetables that have been shipped over a distance, even if this increases the
total tonnage thrown away.
Even if it is ‘obvious’ (a loaded term! If it is obvious, you should be able to
briefly explain why) that your measure is correct, it then needs to be applied to a
concrete problem, such as a particular distribution scenario. In this scenario, you
might postulate a mix of vegetables sourced from local and remote farms with a
mix of lifetimes and of shipping and storage environments (chilled, humid, frozen,
and so on), and a mix of local distribution mechanisms and kinds of retail outlet.
The richer this scenario, the more realistic it might be—and such research would
seem to be of limited value if it only applies to idealized scenarios—but the more
assumptions it embodies. This is a version of the study-versus-case-study problem;
the more specific the concrete problem, the greater the leap of faith required to be-
lieve that the solution will generalize to other cases. To address this issue, you might
need to apply your model to hundreds of scenarios representing different mixes of
the various factors.
Something I’ve found intriguing is how easily, under prompting, some students
find problems with the measures they have been using. A particular example was a
Masters student I was working with, Vivienne, who was looking at the effectiveness
of different forms of the ‘traffic light’ labelling that is used to indicate the healthi-
ness of retail food. For each of the measures she had proposed, and even one she