How should we measure the poverty rate?

Perhaps we shouldn’t.

The idea behind a poverty rate is that we set an income line below which people’s resources are deemed insufficient for a minimally decent standard of living. The poverty rate is the share of people in households with income below that line.

Because it’s a binary measure, it’s a crude one. Suppose a lot of the poor at time 1 have incomes just below the poverty line. The economy then improves, or the benefit amount for a government transfer program is increased, so at time 2 a number of those people have moved above the line. It will appear that poverty has been sharply reduced, even though the amount of genuine progress is small. Similarly, suppose a number of people who formerly had very low incomes move into the work force and experience an income rise, but that rise doesn’t quite get them above the poverty line. This is a significant improvement, but it won’t show up at all in the poverty rate.

This problem is well known among social scientists. Some therefore also calculate the “poverty gap” — the distance between the poverty line and the average income of those below the line. To that we can add inequality among the poor. Measures exist to incorporate either or both of these. But they are complicated and thus difficult to communicate to a nontechnical audience. One common measure, for instance, is the poverty rate multiplied by the poverty gap. This is better than the poverty rate by itself, but the numbers yielded by the measure don’t have an intuitive feel.

Another problem with poverty rates is that much hinges on where the line is drawn, so we end up mired in interminable debates about exactly where that should be (here, here).

Is there a useful alternative? I think so.

Instead of a relative poverty rate, such as the official measure used by the European Union, I recommend the p50/p10 income ratio. Relative poverty is essentially a measure of inequality within the lower half of the distribution, so why not use a measure that more clearly conveys that? The 50/10 ratio is an inequality measure already familiar to social scientists, and it’s fairly simple to explain and understand. And as the first of the following two charts shows, the 50/10 ratio is very similar to the poverty rate multiplied by the poverty gap (the correlation is .96). The second chart shows that the poverty rate is a less effective proxy for the rate x gap.

Instead of an absolute poverty rate, such as the official poverty measure in the United States, we can use absolute household income at the tenth percentile (p10) of the distribution. Across countries and over time, this measure is very similar to the absolute poverty rate multiplied by the absolute poverty gap. But it’s much simpler and easier to comprehend. Also, it’s a low-end analogue to median (p50) household income, a common indicator of the living standards of the middle class.

Why the tenth percentile rather than the fifth or the fifteenth? Actually, I’d prefer the fifth, but there sometimes is reason to worry about data quality as we get close to the very bottom of the distribution. The tenth is reasonably close but not too close to the bottom, it’s a nice round number, and it already is commonly used in inequality measures such as the 50/10 ratio and the 90/10 ratio. But in truth, the choice of the tenth is arbitrary; it’s no more representative than the seventh or the twelfth or any other point at the low end of the distribution.

So we have good alternatives to the two most common poverty rate measures. But what about political impact? Isn’t the poverty rate a helpful tool in pressing policy makers to keep their eye on the least well-off? Maybe. Yet hardly any of Europe’s rich nations had an official poverty rate measure prior to the EU’s introduction of one a decade ago, while here in the U.S. we’ve had an official poverty rate for nearly half a century. The absence of an official poverty rate doesn’t seem to have impeded government commitment to the poor in Europe. And I’m not sure the presence of one has helped a whole lot here.

I don’t expect policy makers or social scientists to stop using poverty rates any time soon. And it won’t be disastrous if they don’t. But we could do better.

9 thoughts on “How should we measure the poverty rate?

  1. What would help is if the Census Bureau would publish household income for the 10th percentile for all areas (and quintiles as well), instead of publishing number of households within absolute dollar amounts of income. The latter can’t be compared over time.

  2. stick with the p10/p50, especially in international comparisons. Whatever happens below p10 (or above p90) the accounting is heavily dependent on local and varying over time, tax laws, business cycle, etc.)
    but the result is boring: p10 / p50 = 0.50, p90 / p50 = 2.0, over time US vs DE, 1913, 1930ties, today, if you net the results properly (health, taxes, subsidies like EITC, household size accounting like DIW SOEP (most appropriate thing I have seen)

  3. The trouble with “grading on a curve” comes up mostly in international comparisons. In dollar terms, even on a purchasing power parity rather than an exchange rate basis, the poor are much, much poorer than the poverty line. For example, more than 50% of Mexcans are below the U.S. poverty line and in plenty of countries in the Third World the overwhelming share are below the U.S. poverty line on a PPP basis.

    But, it is possible to survive, in poverty but to survive, with monetary incomes that would put you in the gutter and starving in NYC for a whole year in someplace like Ghana where one would eke out a living as a small scale subsistence farmer. For apples to apples comparisons one really needs non-monetary comparisons of housing standards, nutrition levels, etc. By those measures, places like Afghanistan, which don’t have tropical subsistance farming economies and are poor look truly abject, while some places that are at the bottom by purely economic measures look better.

    Indeed, part of the source of opposition to foreign aid comes from an inability of an American to have a meaningful understanding of what the standard of living in a poor country means in concrete terms, instead of numbers that have slipped out of their range of relevance.

  4. let me add a few thoughts, examples (you do not even have to go as far as the drastic but valid Ghana / Afghan comparison):
    your US poverty line is typically somewhere around 40 % median, EU definition is 50 %. If you crank up social support and child support measures to 55 %, like in Germany, nobody is “poor” anymore, “the poor” get a lot more babies, and those with a job “caring about the poor” have nothing to do / talk about anymore ? Of course not, they invented “threatened by poverty” and fix that at 60 % of median. So you have again 10 – 15 % of the people to talk about. After the 2nd or 3rd time, or in citing somebody else, these people conveniently forget the “threatened of” part before “poor”.
    In a recent comparison they forgot to incorporate some small additional subsidy, and suddenly Germany was with some 16 % at the top of the EU list of “poor kids”. After adding this small part, the number literally halved and Germany was now at the bottom. Needless to say, depending on e.g. how you incorporate availability and cost of child care, full day school, etc. you can mix those lists any way you want. One last example, in the deep black south of Germany, Bavaria, the most conservative and nowadays the most successful part, unemployment is now down to 3.5 % (statewide), and folks with the german minimum standard do feel poor in their environment, have a solid incentive to better their live. In the eastern parts, which still have for the median only 70 % of the west, 60 % of Bavaria, they still have 10 – 15 % unemployment, and the “standard minimum” represents 70 – 80 % of the median in their neighborhood. that means that some 5 – 10 % are actually somewhat stupid to work 40 hours a week, instead of tending to their garden, doing a little on the side, etc., if you look at it just from the monetary perspective. Kids of 6 year grade working engineers have here often even less space than those with the minimum entitlements, but in their own houses (tiny in comparison to the US, I lived for many years in East, West, and the US). Adding any more (especially financial) public “support” to these 15 % “poor” is extremely counter productive, since it perpetuates the welfare mentality, I know some of these folks. But on paper “objectively” the situation is the same for East and West.
    Soo, if you want to write a lot of papers with dramatic story lines, just use/invent some grades (see above) which give you rising numbers above 15 %

  5. addon:
    This east/west germany comparison might actually be a good teaching example.
    It is far enough away to stir not any sensitivities to the present disucssions in the US.
    But it is close enough (in terms of median GDP per capita, practically the same) to be relevant,
    in constrast to the Ghana / Afghanistan example above.

    There are no racist/ nationalist undertones, no arguments with climate and fauna, different industrialization, mineral riches, whatever.
    Just precisely one difference: one part was impoverished by 40 years of socialism and the other one not.

    I discuss it here from a perspective: how much does it cost and do I give the right incentives.
    I give a few numbers in the beginning of the discussion, just to make sure to you, that there are real numbers and models behind it, which do not require more than high school math, and your sociology students should be able to not only follow, but to actually work with !!

    Nominally social minimum payments (my acronym: SMP) are exactly the same in East and West, in terms of buying power I would argue that means for the east about 5 % more (primarily due to surrounding wage level).
    Rates of “threatened by poverty (60 % of national median)” are a factor of 2 higher in the east (roughly 16 vs 8 %). Ooooogh, lets do something for them …. bullshit.
    Roughly the West (of Germany) has already transfered about 100 % of its GDP in the last 20 years (so about 5 % per anno). So obviously you can squeeze this out of people to for example pay down public debt of the same size : -)
    The US did the same after WWII, and UK twice that.

    First, if I would increase social minimum payments in Bavaria by relatively 10 % (of 45 % of the LOCAL (defined as state level (Bavaria vs … other) median, which is around 70 % LOCAL mean),
    It will increase the number of people from 3.5 (unemployed) + 4.5 % (other) to maybe 5.0 + 5.0 % after a few years.
    (I have calculations with number background !)
    (10.5 % 45 % * 1.1 * 70 %) – (8 % * 45 % * 70 %) ) * 60 % (Consumption/ GDP) = 0.59 % GDP, ridiculously low compared to 5 % West-> East transfer, 1-2% for our lovely european friends, and 5 % current account surplus
    We could do twice of that without significant problems, …. from a strictly LOCAL bavarian monetary perspective.

    But: social minimum payments in the East and the West Germany are the same. And suggesting any tiny difference (like 1 %) would wreck havoc to the social political fabric. I have seen personally the eye slits getting small and gort out fast, before “the people” lynching me.
    Increasing SMP in the east would tip the fraction of relevant folks to 2/3, kind of re-introducing socialism. it would not only cost a lot more, give all the wrong incentives, but even be morally wrong to the “poor” in the West.

    I can go into a lot more detail, if anybody is interested.

    Bottomline:
    – any reasonable, “objective”, scale of “poverty” would regularily lead to wrong policy conclusions.
    – this question is very relevant to present day political discussions.
    – The US has the better long term databases
    – the EU /DE has the more urgent decisions (Eurobonds, etc.) and data comparisons /examples with much less the emotional baggage, well, at least for you in the US : – )

    Very curious about feedback from Kenworthy and others : – )

    Maybe that discussion would lead to some better idea how to define “poverty”, I am somewhat sceptical.
    But it would lead very likely to a better understanding

  6. Why not focus on some measurement of consumption which is less prone to individual fluctuations?

  7. HERE WAS MY ATTEMPT AT A “MINIMUM NEEDS” LINE IN 2/18/2008 (“poverty” conjures up children with distended stomachs; not the extreme needed to make correctives) — which I hoped would finally discourage the parroting of the official federal poverty line calculated by the long brain-dead, mid-fifties formula of three times the price of an emergency diet — dried beans only, no canned! Seemed to work; haven’t seen much of the official line since. Wish I could come up with a cute sales trick for common sense notion.

    ******

    Approaching 37% of American families are below a more up-to-date poverty line?

    The 50 percentile American family income in 2005 was $56,277* (mean third-quintile in the Census tables).
    * http://www.census.gov/hhes/www/income/histinc/f03ar.html

    “Minimum needs” (table 3-2) on p.44 of the 2001 book Raise the Floor* maps out a very plausible poverty line for a family of three at $31,111 in 2005 dollars — assuming that health care is otherwise provided. Add $11,000 to purchase a family health plan and this plausible poverty line rises to $42,111 for a family of three (three years ago). The Raise minimum needs line is computed by totaling up a comprehensive list of actual needs — does not parrot the half-century old federal formula that simply multiplies three times the cost of a minimum food budget.
    * http://www.amazon.com/Raise-Floor-Wages-Policies-That/dp/0896086836/ref=pd_bbs_sr_1?ie=UTF8&s=books&qid=1202780080&sr=8-1

    (Raise’ provides extensive explanations for its minimum needs parameters in Appendix B — its tables cite Solutions for Progress. Average family size is 3.13* persons.)
    * http://www.census.gov/population/www/socdemo/hh-fam/cps2006.html

    The difference between second and third quintile average family incomes ($35,000 and $56,000) runs roughly $1,000/percentile. So, adding $7,000 to $35,000 (the 30 percentile mark) gets us to $42,000: and debarks 37% ** of American families as below minimum needs, at least without food stamps and other helps. Assuming that all families were covered by comprehensive health insurance would still leave 26% of families on the south side of Raise’s minimum needs line without helps. I do not know how many of those families between 26% and 37% are covered or by how much.

    However perfectly accurate Raise’s tables may or may not be, our media continue to report the decades old, mis-measured official federal poverty line of 12.5%* without qualification; which is like the press of Columbus’ era repeating without comment that the world is flat: it makes no waves; but informed folks know better. :-)
    * http://en.wikipedia.org/wiki/Poverty_in_the_United_States

    [** Raise’s tables allot $3,000 to yearly medical expenses for a family of three even if insured.]

  8. Still pretty irrelevant.

    Would you be richer if Bill Gates left the US?

    Nope, you would be poorer.

    First the 10% line would increase, but not your wages.

    Secondly, government would have to increase its taxes, making you absolutely poorer.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s