Recession Raises Poverty Rate to a 15-Year High

New York Times article "Recession Raises Poverty Rate to a 15-Year High" (September 16th, 2010) fails to acknowledge the poverty rate's intertemporally unstable nature. It compares poverty rates over time, an improper comparison due to the way poverty rates are calculated.

The share of residents in poverty climbed to 14.3 percent in 2009, the highest level recorded since 1994. The rise was steepest for children, with one in five affected, the bureau said.

One might think that individuals in the U.S. were only as well off as they were in 2004, when chained GDP/capita was approximately the same as it is this quarter. However, Corrections contends that we are actually even better than this. Below, find graphically depicted U.S. GDP over time in chained 2005 dollars (click to enlarge), and U.S. GDP per capita over time in chained 2005 dollars (click to enlarge). Note that chaining dollars is an attempt to introduce new products for comparison (otherwise, comparing cell phone prices from 1960 and today would not be well-defined).

The poverty rate from 1994 is measuring something completely different from the poverty rate in 2009. Poverty thresholds have changed multiple times, under a "sliding scale" approach. For example, in 1964, ~2.6% of U.S. households owned a color television. Circa 1994, 97% of U.S. households owned a color television. Beyond this, these televisions were not only cheaper, but were of better quality, programming, and durability.

Why is this relevant? Because increasing product quality is not properly measured, even with chained GDP. Below, we take minivans as an example.

In 2003, using Barry, Levinson and Pakes's instrumental method for demand estimation (also discussed in a previous post), Amil Petrin, in his phenomenal paper, "Quantifying the Benefits of New Products: The Case of the Minivan" (JPE 2002) estimates the value in dollars to consumers from the advent of the minivan by Chrysler in its first five years (1984-1988) as $2.8 billion in consumer surplus, and $2.9 billion in total surplus.

This surplus comes solely from an improvement in product quality that is largely unreflected in price due to monopolistic competition by competitors (GM and Ford introduced their own minivans in 1985). This sort of change will not be reflected by even chained GDP numbers. Because of increasing product differentiation (Petrin's "new goods" problem) and monopolistic competition (or competition), life is getting better than we're measuring with our best measures of product-chained GDP. Quality of life is higher.

Poverty indicators are not appropriate for "long" time spans because of innovation. Locally, we might think 2008 and 2009 are comparable. But in 1994, the internet had yet to be invented. Since then, as Austan Goolsbee and Peter Klenow estimate in "Valuing Consumer Products by the Time Spent Using Them: An Application to the Internet" (AER 2006) (gated) (ungated), the median individual gained $3000/year because of the advent (and widespread use) of the internet. The reason for this large gain is largely due to increased price competition and increased value of time. Since 1994, almost everyone in the United States is vastly better off than they were. (Another example might be how much individuals would have paid for a smart phone in 1994, given the millions that have them now and how "little" they paid for them relative for 1994 willingness-to-pay). This massive increase in consumer surplus generated from an increased value of time is unmeasured by GDP (underestimated) and poverty measures (overestimated). Use of these to compare long-run trends is ill-advised, especially when one has an ideological/Malthusian axe to grind.


Unfortunately, the notion that GDP growth generally underestimates utility gains is almost universally ignored in long-run intertemporal comparisons of utility.