In response to some vigorous discussion in the comments last week, the other Bears and I decided to host a blog seminar in ‘Savings 101,’ and I was duly nominated (i.e. drafted) to get the ball rolling. This installment will primarily cover the definition of personal savings and major features of how it’s measured for the purposes of the U.S. National Income and Product Accounts (NIPA). I liberally and without other citation draw on a number of methodology papers posted by the Bureau of Economic Analysis, get them here. A second installment will discuss rationales for savings and financial “innovations” that may affect savings behavior. As demand warrants, later installments may discuss savings of firms and governments.
We’ll get the ball rolling after the jump.
1. Defining Personal Savings
A simple definition of personal savings, which also happens to be the NIPA definition, is whatever’s left over from personal income after taxes and consumption:
Savings = Personal Income – Personal Taxes – Personal Consumption Expenditures [PCE]
From the NIPA accounting perspective, this definition relates sources of personal income with uses of the income; “savings” is the term makes the accounts balance. At least in the short run, PCE may be greater than income after taxes and savings will be negative. To make a little more sense out of it than the accountant’s desire to balance ledgers, net savings may also (and relatedly) be considered as a flow of funds to or from stores of wealth. Indeed, the BEA handily provides a comparison of NIPA and flow-of-funds savings measures, and they track each other pretty well. (The latter series is a lot noisier.)
In addition to personal savings, NIPA also counts savings by businesses in the form of retained earnings and by government in the form of unspent tax dollars. In 2006, essentially all of the net savings recorded in NIPA, around $250 billion, came from businesses’ retained earnings of some $400 billion. That year, personal savings were $39 billion, and the government sector dissaved to the tune of nearly $196 billion thanks mostly to the federal government’s unified budget deficit. (State and local governments, which commonly operate under balanced-budget constraints and also may maintain small rainy-day funds, were net savers.) You may compare these to 2006 GDP of $13,195 billion, personal income of $10,983 billion, business income of $5,814 billion, and government revenues of $3,935 billion as you’d like.
2. The NIPA Measurement of Personal Savings
The basic relationship defining personal savings above may be simple enough on its face, but it does beg the questions of just what constitutes personal income (PI) and what is included in PCE (and how it’s measured).
PI is a bit more straightforward than PCE. By far the biggest component of PI is wages, salaries, and benefits. Next is dividend and interest income. Then comes transfers, mostly from government benefits programs. Then comes income from business proprietorship, and finally a small amount of rental income.
Contributions to government social insurance programs (read, Social Security and Medicare) are subtracted from PI and added to government receipts. So accumulation of social insurance trust funds is considered savings in the NIPAs, though government rather than personal savings. Of course, as seen above the Social Security program’s savings currently serve to offset in part dissavings elsewhere in the government sector, i.e. the “on-budget” federal deficit.
The other big question is, what is included in PCE? PCE tries to measure personal consumption quite broadly, covering consumption of durable goods (which have an expected useful life of three years or longer), non-durable goods, and services. Non-durable goods and market services are straightforward, at least in theory, in that those basically involve current expenditures for current consumption. For housing and other durables, the NIPAs offer something to make everyone a little mad, since those measurements involve a variety of assumptions and imputations that aren’t necessarily innocent.
The big imputation relates to owner-occupied housing, a/k/a “owner’s equivalent rent.” The idea here is that houses and other residential structures are forms of fixed capital and the consumption is of shelter services produced using that capital. Renters pay for those services in market transactions, but owners — while they consume similar shelter services to renters of comparable dwellings — don’t literally charge themselves rent. However, the rent that could hypothetically be obtained by renting out a house instead of occupying it is an opportunity cost of owner-occupation. That opportunity cost amounts to owners’ implicit expenditures on shelter services. Importantly, there’s no reason to view owners’ mortgage payment(s), or lack thereof, as having any particular relationship to that implicit shelter expenditure.
The rent imputation helps make the measurement of housing services less invariant to how people obtain them. Think of two identical houses next door to each other, one of which rents for $10,000/year and the other of which is occupied by an owner who owns the house outright. In the absence of the imputation, PCE would measure only the $10,000 in market rent. If instead the owner rented the house next door, and rented out her or his house for the same $10,000, measured PCE would increase to $20,000 even though all we’ve done is move families between two identical houses.
Finally, and importantly for savings measurement, the imputation of rent effectively splits housing outlays into savings and consumption components. To the extent actual outlays exceed the imputed rent, the excess is treated as saving.
Economists of a reasonably orthodox bent shouldn’t take issue with owner’s equivalent rent at a conceptual level, but the actual measurement involves a fair amount of sausage-making. In particular, the owners’ rent is extrapolated from market rents. While the rental-extrapolation method is based on the presumed existence of “well-developed” housing rental markets, in many cases those markets are quite thin in units comparable to the owner-occupied housing stock.
If you think the PCE methods are too economically pure for housing services, the BEA’s methods are more expedient than correct for other consumer durables (e.g. autos). PCE simply includes the full selling prices of durables at the time of durable-goods sales. This would tend to make durable-goods PCE more volatile than actual durables consumption. This may also make NIPA savings appear relatively low during durables booms and high during durables busts.
A final large category of imputed consumption is for financial services provided without explicit charges. These services aren’t free, of course; the associated expenses are usually subtracted off investment income before it’s paid by the financial services providers. In this case, the NIPA methodology imputes both the value of the services and an addition to personal income for the subtraction, so it’s on both sides of the savings equation and doesn’t affect the measured level of personal savings.
A lot of criticism of the standard definition of personal savings boils down to potential differences between aggregate personal savings and an equation that shows the direction of personal assets:
Change in Assets = Net Cash Flow + Investment Earnings + (Unrealized) Gains or Losses from Asset Price Changes
The NIPA methodology explicitly declines to include the last term in income, regarding it as “reflect[ing] a change in wealth rather than a change in productive activity.” However, changes in and levels of wealth are widely regarded as interacting with expenditures and savings behavior, so it’s hardly irrelevant. We will turn to that in Savings 101b.