This paper adopts a ``revealed preference" approach to the question of what can be inferred about bias in a political system. We examine a dynamic environment in which individuals differ by income each period.
Long run preference profiles are unobserved to an outside observer but are known to belong to a well behaved class in which individual preferences are ordered by income in each state. Policy data is summarized by a Markov policy rule. The observer makes inferences about the underlying distribution of political power as if political power were derived from a wealth-weighted voting system with weights that can vary across states. The weights determine the nature and magnitude of the wealth bias. Positive weights on relative income in any period indicate an ``elitist" bias in the political system whereas negative weights indicate a ``populist" one.
We ask: what class of weighted systems can rationalize a given policy rule as a weighted-majority outcome each period? Our first result shows that without further knowledge, all forms of bias are possible: any Markov policy rule can be shown to be rationalized by any system of wealth-weighted voting. An additional single crossing restriction on preferences can, however, rule out certain weighting systems. We then augment policy data with polling data and show that the set of rationalizing wealth-weights are bounded above and below, thus ruling out extreme biases. In some cases, polls can provide information about the change in political inequality across time.
- elitist bias,
- populist bias,
- weighted majority winner,
- rationalizing weights,
- ``Anything Goes Theorems".
Available at: http://works.bepress.com/roger_lagunoff/13/