Transcript Data, Empirical methods, and Statistics

Fact and value
Policy issues (cont.)
Value neutrality
Data in sociology.
What is the relation of fact to value?
Social science theories can never determine what our
response to social problems should be;
– but they are relevant;
– they have policy implications;
– and this raises special problems of objectivity.
There are often conservative, liberal and radical
sociological analyses
Conservative: there has been too much change in the
direction of equality; we need to hold the line or go back.
Liberal: minor reforms can produce equality of
opportunity.
Radical: Inherited privilege is built into the social
structure; a progressive and free society requires
structural change.
e.g. 1: The Cherokee Reservation
The Indians are very poor; there is no wealth, no
jobs, no land; high anomie; high suicide; high
homicide; high drop-out.
1. Conservatives: (Murray) believe we have been
too generous and need to get tough.
2. Liberals: We need more support and
understanding; Vista built church row.
3. Radicals: (Feagin) We need to give back some of
what we have stolen from them.
 Compare issues in The Rabbit-Proof Fence: there
are policy dilemmas; even the goal of genetic and
cultural extermination was pursued for good
motives.
e.g. 2: Organic v.
solidarity

1.
2.
3.
Forced division
of labor
Murray’s main arguments are that social policies
(welfare, food stamps; headstart; affirmative action;
minimum wage; child labor laws) are ineffective.
Conservatives (Murray) believe that there is equal
opportunity today; the inequality that results is fair.
Liberals (Reskin) believe that modest reforms of
schools, mentoring, affirmative action, etc. would
produce equal opportunity; most inequality is fair.
Radicals (Feagin, Massey) believe that immense
inherited property and wealth produces unequal
education, unequal treatment by the law, unequal
political influence, segregation, structural strain, etc.
e.g. 3: Myrdal and Race relations
today
1. Conservatives (Murray, Thernstrom)
believe that equal opportunity is violated
by affirmative action and welfare.
2. Liberals believe that social supports for
the poor, especially poor children, (e.g.
ed., health, food) have to be improved.
3. Radicals (Feagin) believe that so long as
there are immense inherited group
inequality, it is a racist society.
Data, Empirical methods, and
Statistics
A basic element of any science is its
empirical access to the world.
In sociology, that often means statistics.
“You can prove anything you want with
statistics.” *63
but “You can prove anything you want
without statistics, too! At least with data
what is proved is more than just your
opinion – or mine.” *63
What you need from ch. 3
The growth of empirical data and methods of
statistical analysis has been one of the central
progressive developments in soc.
Therefore most later chapters and weeks will
require that you can look at a multivariate
crosstabulation
– i.e. % difference as a measure of association
– Between the independent and the dependent variables
– At fixed levels of the control variables.
Learn to look through the interpretation at the data
The empirical data of the early
sociologists
To the extent that the world is complex, one’s
methods of collecting and analyzing data must
be sophisticated.
Durkheim and the Chicago sociologists
concentrated on the data already collected by
government agencies: things like suicide rates
or rates of juvenile delinquency.
Some sociologists use such data, but it is limited
because it was not collected to help understand
social causality.
The main sources of data in
sociology today
1. Experimentation.
2. Participant Observation.
3. Survey Research.
Each has some strengths and some
weaknesses for the investigation of
causality in social structure where there
are multiple causes and reciprocal effects.
Objectivity, Reliability, Validity,
Generalizability
Virtues of good measures, theories and
analyses in the social sciences include:
Objectivity – that the measure reflects the real
qualities of the thing, outside of the mind of the
person observing it.
Reliability – that the measurement yields the
consistent results over observers and over time.
Validity – that the measurement procedure
measures what it is supposed to measure.
Generalizability – that the degree to which the
conclusion of the study can be applied outside of
the study participants.
Experiments
For establishing causality, a controlled
experiment has great advantages.
Specifically, it allows one to randomly assign cases
to the control group
And to manipulate the independent variable.
However, generalization of experimental
findings to the “real world” is problematic.
For example, replication and applicability
of the Zimbardo experiment have been
debated.
Participant observation
The method analogous to anthropology is direct
observation
– Anthropology mainly uses interpretation of observation
– Urban ethnographies are similar.
There have been many ethnographies analogous
to the movie, 187,
– And we shall examine some of them.
– But in direct observation of a concrete situation there
are problems both of interpretation and of
generalizability.
Survey research
Therefore, the most commonly used method in
sociology is survey research.
– Asking questions of a large sample.
– Virtually every chapter of Sociology, Micro, Macro and
Mega will illustrate points with survey findings.
– Usually these are taken from the General Social
Survey,
– Often analyzed in cross-tabulations.
The raise problems of measurement,
interpretation and spurious association
(indefinitely many) controls.
An example: WLTH POV
The question:
: “In a free society it is all
right if a few people accumulate a lot of wealth
and property while many others live in poverty.
Validity: What does it measure?
What fraction of the population agrees”? Why?
Association: What makes one more likely to
agree or to disagree?
AGREE or DISAGREE
– The most usual measures of association will be
percent differences in Crosstabulations
– e.g. How do you think income, race and gender affect
one’s attitude on WLTH POV?
Crosstabulation showing the effect of
income on agreement w. WLTH POV
DEPENDENT VAR:
WLTH POV
INDEPENDENT VAR:
INCOME
HIGH
MIDDLE
LOW
AGREE
NEITHER
DISAGREE
T
189
82
127
398
(48%)
(21%)
(32%)
100%
168
85
199
452
(37%)
(19%)
(44%)
100%
103
54
184
341
(30%)
(16%)
(54%)
100%
Interpretation:
The main point of this table:
There is a moderately strong positive association
between income and agreement that “In a free society it
is alright if a few people accumulate a lot of wealth and
property while many others live in poverty.”
Upper income respondents are 18% more likely to
agree.
Note: look at percent differences calculated on the
independent variable.
But this answer to the question, “How does social
position affect this attitude?” raises further questions
about Why? When? How is it changing:?
Spurious association: why
empirical association does not
prove causality
Finding that there is a general empirical
association between two variables is
usually a necessary but not a sufficient
demonstration of the effect of the
independent variable (cause) on the
dependent variable (effect.)
Whenever multiple causes are operating,
one must separate them analytically.
In social structure, multiple causes operate
e.g. falling bodies
Newton’s laws imply that everything will tend to
fall toward the center of the earth at 32’ per
second per second.
However Newton’s laws are not contradicted by
the fact that virtually nothing falls at that rate,
and many things do not fall at all.
Because there are other forces, particularly
frictions, at work.
However, Newton’s laws do require that when
those frictions are less, then falling bodies do
approach the predicted rate of descent.
A bad test: observe the association
of fire engines and damage
Much
Damage
Little
Damage
Many
Engines
45
5
(90%)
(10%)
Few
Engines
5
45
(10%)
(90%)
e.g. fire engines and fire damage
If fire engines reduce fire damage, then
one would expect a negative association:
– when there are more engines there should be
less damage;
– and when there are fewer engines, there
should be more damage.
However, one would almost certainly
observe a positive association:
– More engines, more damage
because fire
no. of engines
As well as no. of engines
fire damage
A better test: the controlled association
of fire engines and damage
Big Fires
Small Fires
Much
damage
Little
damage
Many
engines
40
1
(98%)
(2%)
Few
engines
2
3
(40%)
(60%)
Much
damage
Little
damage
Many
engines
5
4
(56%)
(44%)
Few
engines
3
42
(7%)
(93%)
A possible “libertarian” theory of
government inefficiency
When we had little government fire companies,
individuals took more responsibility for
prevention and extinction,
and fire damage was less.
Therefore, we would be better off without fire
departments.
Comment: though the empirical facts that the
theory appeals to are correct, the theory is
almost certainly wrong, as shown by the Great
San Francisco fire.