Sunday, May 18, 2014

The Nonexistent Rise in Household Consumption

Did you know that about 10 percent of private consumption in the US consists of Medicare and Medicaid? Despite the fact that these are payments by the government to health care providers, they are counted by the BEA both as income and consumption spending for households.

I bet you didn't know that. I bet plenty of people who work with the national income accounts for a living don't know that. I know I didn't know it, until I read this new working paper by Barry Cynamon and Steve Fazzari.

I've often thought that the best macroeconomics is just accounting plus history. This paper is an accounting tour de force. What they've done is go through the national accounts and separate out the components of household income and expenditure that represent cashflows received and made by households, from everything else.

Most people don't realize how much of what goes into the headline measures of household income and household consumption does not actually correspond to any flow of money to or from households. In 2011 (the last year covered by the paper), personal consumption expenditure was given as just over $10 trillion. But of that, only about $7.5 trillion was money spent by households on goods and services. Of the rest, as of 2011:

- $1.2 trillion was imputed rents on owner-occupied housing. The national income and product accounts treat housing on the principle that the real output of housing should be the same whether or not the person living in the house happens to be the same person who owns it. So for owner-occupied housing, they impute an "owner equivalent rent" that the resident is implicitly paying to themselves for use of the house.  This sounds reasonable, but it conflicts with another principle of the national accounts, which is that only market transactions are recorded. It also creates measurement problems since most owned residences are single-family homes, for which there isn't a big rental market, so the BEA has to resort to various procedures to estimate what the rent should be. One result of the procedures they use is that a rise in hoe prices, as in the 2000s, shows up as a rise in consumption spending on imputed rents even if no additional dollars change hands.

- $970 billion was Medicare and Medicaid payments; another $600 billion was employer purchases of group health insurance. The official measures of household consumption are constructed as if all spending on health benefits took the form of cash payments, which they then chose to spend on health care. This isn't entirely crazy as applied to employer health benefits, since presumably workers do have some say in how much of their compensation takes the form of cash vs. health benefits; tho one wouldn't want to push that assumption that too far. But it's harder to justify for public health benefits. And, justifiable or not, it means the common habit of referring to personal consumption expenditure as "private" consumption needs a large asterix.

- $250 billion was imputed bank services. The BEA assumes that people accept below-market interest on bank deposits only as a way of purchasing some equivalent service in return. So the difference between interest from bank deposits and what it would be given some benchmark rate is counted as consumption of banking services.

- $400 billion in consumption by nonprofits. Nonprofits are grouped with the household sector in the national accounts. This is not necessarily unreasonable, but it creates confusion when people assume the household sector refers only to what we normally think of households, or when people try to match up the aggregate data with surveys or other individual-level data.

Take these items, plus a bunch of smaller ones, and you have over one-quarter of reported household consumption that does not correspond to what we normally think of as consumption: market purchases of goods and services to be used by the buyer.

The adjustments are even more interesting when you look at trends over time. Medicare and Medicaid don't just represent close to 10 percent of reported "private" consumption; they represent over three quarters of the increase in consumption over the past 50 years. More broadly, if we limit "consumption" to purchases by households, the long term rise in household consumption -- taken for granted by nearly everyone, heterodox or mainstream -- disappears.


By the official measure, personal consumption has risen from around 60 percent of GDP in the 1950s, 60s and 70s, to close to 70 percent today. While there are great differences in stories about why this increase has taken place, almost everyone takes for granted that it has. But if you look at Cynamon and Fazzari's measure, which reflects only market purchases by households themselves, there is no such trend. Consumption declines steadily from 55 percent of GDP in 1950 to around 47 percent today. In the earlier part of this period, impute rents for owner occupied housing are by far the biggest part of the difference; but in more recent years third-party medical expenditures have become more important. Just removing public health care spending from household consumption, as shown in the pal red line in the figure, is enough to change a 9 point rise in the consumption share of GDP into a 2 point rise. In other words, around 80 percent of the long-term rise in household consumption actually consists of public spending on health care.

In our "Fisher dynamics" paper, Arjun Jayadev and I showed that the rise in debt-income ratios for the household sector is not due to any increase in household borrowing, but can be entirely explained by higher interest rates relative to income growth and inflation. For that paper, we wanted to adjust reported income in the way that Fazzari and Cynamon do here, but we didn't make a serious effort at it. Now with their data, we can see that not only does the rise in household debt have nothing to do with any household decisions, neither does the rise in consumption. What's actually happened over recent decades is that household consumption as a share of income has remained roughly constant. Meanwhile, on the one hand disinflation and high interest rates have increased debt-income ratios, and on the other hand increased public health care spending and, in the 2000s high home prices, have increased reported household consumption. But these two trends have nothing to do with each other, or with any choices made by households.

There's a common trope in left and heterodox circles that macroeconomic developments in recent decades have been shaped by "financialization." In particular, it's often argued that the development of new financial markets and instruments for consumer credit has allowed households to choose higher levels of consumption relative to income than they otherwise would. This is not true. Rising debt over the past 30 years is entirely a matter of disinflation and higher interest rates; there has been no long run increase in borrowing. Meanwhile, rising consumption really consists of increased non-market activity -- direct provision of housing services through owner-occupied housing, and public provision of health services. This is if anything a kind of anti-financialization.

The Fazzari and Cynamon paper has radical implications, despite its moderate tone. It's the best kind of macroeconomics. No models. No econometrics. Just read the damn tables, and think about what the numbers mean.

15 comments:

  1. Michael Mandel pointed this out many years ago along with growth of debt before the bust, but no longer actively blogs.

    ReplyDelete
    Replies
    1. Mandel blogs, but it's more sporadic at his new place.

      Delete
  2. I can't remember Mandel actually blogging specifically about that. His issue was more about how the trade and productivity statistics make the US seem more productive than it actually is.

    ReplyDelete
    Replies
    1. That's a rather different issue, tho also important. One thin I've been thinking about for a while, but haven't yet figured out how to approach systematically, is how the different approaches quality adjustment taken by national statistical agencies will show up in the international accounts.

      Delete
  3. ISTR Doug Henwood pointed this out some time ago, at least the healthcare bit.

    ReplyDelete
    Replies
    1. What I recall Doug pointing out is that health care spending accounted for the whole rise in consumption. I don't think he said that it was specifically public health spending.

      Delete
  4. This doesn't seem quite right to me. We need to be careful of what we mean by "growing". So it looks like private consumption has been stable relative to GDP over the past 30 years or more. This means that private consumption per capita has been growing relative to inflation. But I thought wages were not growing relative to inflation, and were therefore shrinking relative to GDP? This would mean consumption was growing relative to wages. How was that achieved if not through debt?

    ReplyDelete
  5. Of course it doesn't sound right -- it's a very different story from what we're used to, which is what makes it interesting.

    You are right that the wage share of GDP has been falling. So a stable consumption share implies rising consumption relative to wages. But there's nothing strange about that -- people consume out of capital income and transfers as well as labor income. As far as I can tell, consumption inequality has increased more or less in line with income inequality. So the answer to your question is that wage earners have been consuming less, and profit recipients have been consuming more. Credit financed consumption has not been macroeconomically important, ever. Even the rise in consumption in the late 90s seems to have been about wealthy households reducing their holdings of financial assets, as opposed to increased borrowing.

    ReplyDelete
    Replies
    1. I should clarify that while I don't think credit-financed *consumption* has ever been important, credit-financed household demand in the form of housing investment has been, especially in the 1950s and 2000s.

      Delete
  6. Maybe they should just rename it "personal consumption" (which is what they seem to be attempting to measure) and have another (smaller) measure that's actual personal consumption expenditures. (I guess it matters: does the modifier "personal" apply to "consumption" or "expenditures"?)

    Similarly, when talking about total national consumption versus investment, gross and net, why does nobody ever talk about Gross Consumption, which would be Consumption (expenditures?) plus consumption of fixed capital.

    Economists are guilty of very, very sloppy terminology and accounting definitions in most of their discussions. (Don't even get me started on "financial capital.") All hail Cynamon and Fazzari (and Josh Mason).

    ReplyDelete
    Replies
    1. Steve-

      Half of me wants to say: There's nothing wrong with the current PCE series except that it doesn't mean what people think it means. We just need a second series that captures consumption spending by households. But the other half of me wants to say that the choices made by the BEA are not innocent, they are based on some strong theoretical commitments.

      Re consumption of fixed capital, the C&F adjustments do add that back into income (since it's not a subtraction from cash income in the first place) in the two places it's subtracted: owner-occupied housing and proprietor's income.

      Delete
  7. JW: "a very different story from what we're used to, which is what makes it interesting."

    Yup.

    ReplyDelete
  8. Thank you for sharing valuable information. Nice post. I enjoyed reading this post. The whole blog is very nice found some good stuff and good information here Thanks..Also visit my page Rental House Accountant There is great satisfaction in knowing we've done our job well and served our clients' interests. It gives us particular satisfaction when our clients take the time to acknowledge their satisfaction by providing Rental Accountants with their testimonials .

    ReplyDelete