Indeterminacy

1
Indeterminacy

• Brette Harrison
• February 2, 2003

2
Indeterminate Research Designs

• Causal relationships only
• Research Design a plan that examines how
  researchers will attempt to derive inferences
  from data using a prescribed model and data set
• Quantitative vs. Qualitative
• Indeterminate Research Design reveals a limited
  amount of information (if anything at all) about
  the causal hypotheses the RD fails to provide
  sufficient leverage to differentiate between
  various possible outcomes relevant to the
  hypotheses.

3
Research Designs continued

• Generally less obvious in qualitative experiments
• Greater volume of data, generally available in
  qualitative designs, may provide means to make
  the experiment determinant
• 2 possible reasons why research design may be
  indeterminate (King 119)
• More inferences to make than implications
  observed
• Multicollinearity

4
Inferences gt Observations

• Inference using facts we know to draw
  conclusions about facts we do not know
• Rule one fact ( or observable implication)
  cannot give independent information about more
  than one other factIn practice, we usually need
  many more than one observation to make a
  reasonably certain causal inference (King 119)
• Comparison is not necessarily a valid methodology
  for determining causal relationships

5
Example

• Assume we have three case studies, each of which
  describes a pair of countries attempts to create
  a free-trade area. All three case studies are
  extensive in their examination of economies of
  scale, labor mobilization, exchange rates, and
  the final result. Throughout the experimental
  process we cite ten factors important to the
  creation a successful free trade zone.
  Regardless of the plausibility of these ten
  explanatory variables, the research design
  renders it impossible for us to determine which,
  if any, of the hypotheses are correct. There are
  not enough observations to evidence the multiple
  inferences.

6
Formal Analysis

• E(Y)3X15X2
• E(Y) is the causal model
• Y is the dependent variable
• X1 and X2 are the parameters of the model
• Even if values are assigned to the parameters (X1
  and X2), and the model produces a final result,
  there is no unique solution to an equation with
  two unknownsall possible solutions are equally
  consistent with the single observation.

7
Multicollinearity

• Any situation where we can perfectly predict one
  explanatory variable from one or more of the
  remaining explanatory variables (King 122)
• An instance where the explanatory variables are
  so closely intertwined that it becomes impossible
  to determine the causal influence of one variable
  over the other

8
Example

• Suppose in our study of successful free trade
  areas we find (1) cooperation between countries
  that share a border are more likely to be
  successful than those that are isolated from one
  another and (2) countries that are able to
  benefit from comparative advantages are more
  likely to succeed than countries without
  comparative advantage. Both explanatory
  variables examine the negative effects of
  domestic unemployment on the creation of free
  trade areas. However, if for some circumstance
  of time or resources, we only observed (1)
  neighboring countries that benefited from
  comparative advantages and

9
Continued

• (2) non-neighboring countries without comparative
  advantage. In this instance, the two parameters
  are perfectly correlated and neither variable can
  sufficiently support the hypothesis
  independently.

10
Maximize leverage

• Build confidence in your estimates
• Observe as many implications as possible (time
  and money are not factors in determining validity
  and reliability)
• Diversify sample set (sensitivity to situation
  1)
• Defining unit of analysis
• avoid nearly unique circumstances as it limits
  examples
• manageable and numerous units
• A successful project is one that explains a lot
  with a little. At best, the goal is to use a
  single explanatory variable to explain numerous
  observations on dependent variables (King 123).

11
Works Cited

• King, Gary, Robert O. Keohane, and Sidney Verba.
  1994. Designing Social Inquiry. New Jersey
  Princeton University Press.
• Etheridge, Marcus E.,ed. 2002. The Political
  Research Experience Third Edition. New York
  M.E. Sharpe.