I am reviewing an article and am having difficulty understanding the methodology employed by the author, and am wondering if anyone has some guidance (or cites) for me.  Here are the details of the methodology (I am trying to protect the specific topic and submission):

Subject: Aggregate predictors of votes in statewide congressional election in the U.S.
Unit of Analysis: town/city

Independent Variable: change in unemployment rate over a one-year period prior to the election
Dependent Variable: total vote for a candidate for statewide office

Methods: OLS Regression, with three models (other independent measures not shown here)
[here is where I need help]

Model 1: uses change in unemployment rate
Model 2: uses change of unemployment rate squared
Model 3: uses change of unemployment rate cubed

My question: I believe I understand the reason for using unemployment rate change squared (since the relationship between it and vote total is not linear, but curvilinear), but what is the advantages of cubing the rate?  Is this something that is accepted among those of you who employ such methods.

Thank you for any guidance.  Feel free to respond off-list to [log in to unmask] if you prefer.

Joe Cammarano
Providence College
Providence, RI 02918-0001


**********************************************************
             Political Methodology E-Mail List
   Editors: Diana O'Brien        <[log in to unmask]>
            Jon C. Rogowski <[log in to unmask]>
**********************************************************
        Send messages to [log in to unmask]
  To join the list, cancel your subscription, or modify
           your subscription settings visit:

          http://polmeth.wustl.edu/polmeth.php

**********************************************************