Objectivity and Evidence.
Is value-free social science
I. What is value-free social
I shall presuppose that there are positive facts: facts
that do not imply that something is good or bad,
better or worse, valuable or not. (Note: this is
I shall allow that there may or may not be facts
about values. (Assuming there are such facts is one
very simple way of securing moral objectivity.)
Examples of positive facts: (candidate facts)
1) The force exerted by the magnet is 2 dynes.
2) The grass in my garden now is greener than in
my friend’s garden in Northern Alberta.
3) The charge of an electron is…
4) “The import elasticity of demand for
automobiles in Britain in 1979 is 1.3.” [Blaug.
5) Nancy is 5’5” tall.
B. Positive (value-free) social science
Establishes (true or false) claims reporting positive facts.
Without presupposing the truth of any value claims (i.e.
“objectively”). An „objective claim‟ is a claim about
positive facts that is established objectively.
C. Why can‟t we have value-free social science?
Central social science claims inevitably fail to report
It is hard to eliminate value assumptions from social
II. Some distinctions
A. Amartya Sen-
“Accounts, Actions and Values: Objectivity of Social Science”
Sen: Nobel-prize winning social welfare theorist,
economist/philosopher famous for his work on famine.
B. Sen distinguishes
The objectivity of claims
The goodness of accounts
The goodness of actions
Economic claims and judgements about the goodness of
accounts can be objective. Judgement about the
goodness of actions are moral judgements.
C. Sen‟s example
BBC Panorama programme on brain death and
kidney donations, October 1980. The
programme cast doubt on the certainty of death
of allegedly dead patients when their kidneys
1) Were the claims true/ well supported?
2) Was it a good account of brain death and kidney
3) Was the action of broadcasting the report right?
“‟Should the BBC have given such an account?‟ is a
question about action judgement, not about account
This question requires a “moral judgement”. Neither
questions 1) nor 2) do.
“The same applies to choosing…; selecting
questions…, and picking the ways of
D. Sen on ACTIONS
“The problem here isn‟t fearing that scientific action might
be value-loaded, but fearing that it might not. Value-
loading here is not so much a right as a duty. An action
by a person that is contrary to his or her values…
remains pernicious in terms of his or her own values,
even if it happens to be related to science.
… Actions related to science are like all other actions,
calling for evaluation, assessment, and scrutiny.”
See also P. Kitcher on Truth, Democracy,…
III. Is value-free social science
Why might many central social science claims fail
to describe positive facts?
the social facts we want to study (e.g. human
development, unemployment) can in many
cases only be identified with (or measured by)
a set of positive facts given a value
The United Nations‟ „Human Development Index‟: The HDI
is an equal weighting of – life expectancy, literacy, level of
education with a specific choice of how to measure each
of these in turn. There is a sense in which a country‟s
having a certain value for the HDI is a positive fact. BUT,
there is a sense in which it is not. Why? The choice of
measurement procedure reflects a view about what
constitutes human development. E.g. no indicators of
political freedom are included (and historically, we know
this is for political reasons).
1. What lies behind the rules for how to measure it? E.g. Are
women who look after the family unemployed? Maybe;
maybe not. But not children.
2. We have a view about who should be working, e.g. single
Models and methods may presuppose/imply value
assumptions without note.
In standard „search and matching‟ models of
unemployment individual welfare is measured
only by income. They do not add in the values of
self-esteem or social inclusion, which would lead
to different results.
Often we assume that social welfare is some
aggregation… a representative agent. But consider: in a
discussion of employment models, A. Atkinson shows
that the not uncommon “efficiency” criterion used in
those models does not allow for distribution values:
the social value of income is indifferent in those models
to distributional effects. This is clearly opposite to the
criterion of John Rawls: social value has to do with
improving the position of the least well off without
harming anyone else. So… the identification of social
welfare with aggregated individual welfare
presupposes a value commitment…(and this value
commitment is seldom noted in macro texts.)
Use Sen‟s distinction: Various measure are objective- the
methods establishing the values they take do not require
value judgements. But using the measures in particular
ways is an action and hence subject to moral scrutiny.
Perhaps even just calling concepts by certain names is
subject to moral scrutiny.
Logically: if you show something is true of a concept
defined in a certain way, you mustn‟t illicitly carry the
results over to a different concept with the same name.
IV. Detailed Example
A. A.B. Atkinson-
“The Welfare Basis of Macroeconomics” (lecture, BAAS,
„Golden Rule‟: invest till the (steady state) rate of
return (r)= rate of growth (n)
Policy advice. How established? Via a theoretical model.
The standard models begin with an individual welfare
Where β is discount rate for utility of future times and U is
utility. (This is a simple version of a standardl formula,
used by ‘Chicago School’ economists like R. Lucas.)
2. Justification for using a positive β (by
Armartya Sen): We may reasonably think
the existence of future generations is
uncertain… nuclear war for example.
3. BUT (points out Atkinson) if you think that
it is 40% probable we don‟t survive into the
next century this gives β= .5%: “an order
of magnitude smaller” than Lucas‟s (1+β)=
4. Consider other models.
Overlapping generations: (1+r) =
(1+n)(1+β) – Modified Golden Rule
4. Consider other models.
Overlapping generations: (1+r) = (1+n)(1+β) –
Modified Golden Rule
Our model adds a “representative” utility for each time.
A classic utilitarian will be concerned with total utility:
∑(1+β)-t(1+n)tU(ct). This implies (1+r)=(1+β).
Bottom line: We can go from r=n to r=n+β to r=β
depending on discounting and how we deal with agents.
Often told: someone else picks the ends. Positive social
science shows the means to those ends. Our example
shows how hard this is: We pick a general end, say, a
savings rate that leads to steady state equilibrium. We
look to the models of economics to tell us what rate
achieves this end. BUT (as we have seen) there are
generally a lot more value choices to be made in the
internal structure of the model.
It is only when ALL these value choices „appear out
front‟ – e.g. if you want steady state equilibrium where
future generations are treated differently from current at
exactly a discount rate β and where you maximise not
total utility but the sum of representative utilities, etc.,
then set r=… -- that you get value-free accounts of
the means. And it is not at all clear that can be
- Provide an example of a standard social science concept
that you think is value free and an example of one you
think is not.
- Provide an example where the use of a concept may be
open to praise or blame though the concept itself is
relatively value free.
- Explain and illustrate Sen‟s distinction between the
objectivity of claims and
i. The goodness of accounts
ii. The goodness of actions