The U.S. Poverty Measure and the Politics of Statistics Rebecca M. Blank U.S. Department of Commerce
Much of today is about “How to measure poverty” I want to talk about “How to implement and institutionalize a poverty measure”
Why haven’t we changed the U.S. poverty measure in 45 years? Larger question: What lessons can we learn about effective statistical policy from the statistical politics that have swirled around the poverty measure and prevented changes?
What isn’t unique about poverty measurement: The question of how to define and measure poverty is not obviously any more complex or politically sensitive than many other statistical questions that we regularly answer
Political economy lessons from the endless U.S. discussion about the poverty measure: A. What we learned about implementation by doing it badly B. What we learned about advocating for a new measure by doing it badly
What we learned about implementation by doing it badly 1. Do not define a statistic through specific regulations that require political action to change
What we learned about implementation by doing it badly 2. Do not set regulations inside an office that is part of the Executive Office of the President
What we learned about implementation by doing it badly 3. The less a statistical definition is used in program eligibility and benefit determination, the easier it is to improve and change that statistic
What we learned about advocating for a new measure by doing it badly 1. Do not assume that change is a rational process requiring appropriate statistical arguments and a reasonable alternative measure
What we learned about advocating for a new measure by doing it badly 2. Do not ignore how the definition is utilized in program eligibility, thus creating disincentives toward change
What we learned about advocating for a new measure by doing it badly 3. Don’t make the perfect the enemy of the good (and don’t want too much from a single statistic)
What we learned about advocating for a new measure by doing it badly 4. Don’t think there’s only one way to accomplish what you want
What do we need for a good long- term statistical measure? 1. Agreement that the problem/issue is important enough to justify the effort of producing a statistic
What do we need for a good long- term statistical measure? 1. Agreement that the problem/issue is important enough 2. A quasi-independent statistical agency charged with developing, defining and producing the statistic
What do we need for a good long- term statistical measure? 1. Agreement that the problem/issue is important enough 2. A quasi-independent statistical agency 3. Data and definitions that are ‘good enough’
What do we need for a good long- term statistical measure? 1. Agreement that the problem/issue is important enough 2. A quasi-independent statistical agency 3. Data and definitions that are ‘good enough’ 4. A statistical agency that can, over time, balance the need for updating against the costs of changes