Tuesday, January 09, 2007

The underground economy and the poor

Neil Reynolds takes a look at the role that the poor play in the hidden economy.

The Globe and Mail (Jan. 5) ($):

How do poor families spend so much more money than they earn? By one measure — the National Council of Welfare — the average poor Canadian family spends $4,855 a year more than the $14,366 it receives as income, a difference of 33 per cent. By another measure — the Fraser Institute — the average poor Canadian family spends $9,370 more than the $9,114 it receives as income, a difference of more than 100 per cent.

* * * * * *

How does Statscan determine the income of the poor? It asks them. How does it determine the spending of the poor? It asks them. What's the source of the "bonus bucks" that the poor spend? Perhaps, in one of its surveys, Statscan should ask them. We can, meantime, only speculate.

Off-the-table earnings. Wanton use of credit cards. Gifts from more affluent family members. Academic scholarships. (Many postgraduate students are, by LICO logic, poverty-stricken.) But Canada's basic information on poverty remains dubious.

No one knows whether the poor, in their reports, minimize the money they either earn or otherwise get. It shouldn't be surprising if they do. Everyone else does it all the time.

I want to make two points. First, nobody is implying that the underground economy is strictly measuring income concealment, as some detractors seem to be charging. In a 1992 report published by Statistics Canada, "the underground economy" is defined as the mean economic activity that is not measured in the system of national accounts. In the past I may have failed to mention other sources of the "hidden economy" when referring to income concealment, but that’s because I was being sloppy and perhaps aiming for brevity.
Secondly I’ve read quite a few papers recently about different approaches to income concealment (ie. ignoring the rest of the hidden economy). The expenditure approach, developed by Pissarides and Weber (1989), seems to be the most well-received, at least according to the literature that I’ve read (please share your opinion here if you have one!). For example, Pissarides and Weber look at the relationship between income and food expenditures for salary and wage earners to evaluate the "normal" relationship between the two. They then compare this relationship to the income/food expenditure relationship of the self employed. If food expenditure appears to be incredibly high relative to the income level for the self employed, involvement in the hidden economy is assumed.
A major assumption is that the wage and salary earners (in comparison to the self employed) have very little ability to conceal their income (again, never mind their total involvement in the hidden economy), because employers document employees' earnings on their T4 slips. Thus, data from wage and salary earners is assumed to be actual. In relation to Reynold's article, this implies that poor families spend more money than they earn not because they are concealing income, but because of credit, loans, etc.

I have some criticisms of the expenditure approach, only one of which is relevant to Reynold’s article. The expenditure approach often depends on the use of surveys, for lack of other data. But to what extent can we rely on surveys? I have in mind a paper by Elffers, Weigel and Hessing (1987), who found zero correlation between survey results and audits for Dutch taxpayers. If there is zero correlation, even for wage and salary earners (this surprises me), this is a violation of a key assumption of the Pissarides and Weber model.

Further to Elffers et al’s findings, Andrew Jackson at the RPE blog has this to say:
Statistics Canada’s main surveys of consumption patterns are not very reliable, particularly when it comes to measuring the consumption of the very poor. Household surveys (formerly the SCF and the SLID) have been shown by Statistics Canada to produce significantly lower estimates of the incidence of low income than Census and tax data, likely because of under-sampling at the low end of the income distribution. (See Marc Frenette, David Green and Garnett Picot "Rising Income Inequality in the 1990s" in David Green and Jon Kesselman (Eds) Dimensions of Inequality in Canada, UBC Press, 2006.)
While this is a major blow to the expenditure approach (and there are others I won’t bore you with), I have to agree with Mr. Reynolds: i) the data simply does not tell what is hidden; and, ii)if other wage and salary earners are in some way involved in the hidden economy, why assume that the poor are an exception? Jackson’s critical piece on Reynold’s article can be found here.
By the way, if it's ever the case that you're itching to read an available-by-subscription-only article that I refer to, I don't mind emailing it by request.

No comments: