Published May 2, 1994
Fairness is one thing, but transfer payments always reduce real labor income: evidence from the U.S. and the U.K.
An international consensus is emerging on the problems of chronic unemployment and lagging real wages. It has won the backing of powerful politicians like Jacques Delors and President Clinton’s economic advisers. Unfortunately, this new consensus is contradicted by standard theory and by labor-market reality.
One expert (Adrian Wood) recently found general agreement on three main points. First, that the main cause of chronic unemployment is a fall in the quantity of labor demanded due to “labor market institutions that put a floor under unskilled wages.” Second, that “removing this floor would not solve the problem, just change its form (from European-style open unemployment into American-style working poverty or withdrawal into crime).” Third, that better education and training will help but will “only work slowly — over decades, not years” (Financial Times, 10 March 1994).
A Doleful Choice?
If this is the prevailing view, then it must be challenged. The first point is quite correct — and it reflects a remarkable change: until recently it was widely argued that unemployment benefits are an “automatic stabilizer” which increases the demand for labor. Nor can one doubt the third: education and training are fine things — but they were fine things back when such policies were fewer and unemployment was much lower.
It is the second point that makes no sense. How often we hear that we face a doleful choice between an “American” model of low benefits, low unemployment and low real wages, or a “European” model of high benefits, high unemployment and high real wages. But the facts show that both the faster rise of unemployment in Europe and the faster relative decline of take-home pay in America are due to expanded social benefits. Transfer payments may be defended for the sake of fairness, but in every case reduce real labor income.
Relative Price of Labor
In standard theory, unemployment is associated with a rise in the relative price of labor. At a higher price, firms hire fewer workers (reducing the quantity of labor demanded) while more people seek jobs (increasing the quantity supplied). In the 1930s and 1940s this was known as “Rueff’s Law,” because Jacques Rueff demonstrated, as early as 1925 for the U.K., that unemployment directly parallels changes in the relative price of labor.
But what, exactly, is the relative price of labor? Obviously it has to do with the level of labor compensation. But like all prices, wages have a meaning only in relation to other prices. A wage of $3 an hour isn’t much when a sandwich costs $3 — but it was a decent wage back when a sandwich sold for 25 cents. So the relative price of labor has to take both pay and prices into account.
But the cost of labor is also affected by labor’s productivity. If we could double the goods produced with an hour of labor, while holding wage rates and prices constant, it would effectively cut the cost of labor in half. But in a competitive market, all units of labor (and capital) are paid incomes equal to what the last unit adds to output. With doubled productivity, real wage rates normally double, leaving the relative price of labor unchanged.
Thus, to measure the relative price of labor, we need to adjust labor compensation for both prices and productivity. (This is sometimes called the “efficiency wage.”) In doing this calculation, we find that the relative price of labor is the same as labor’s share of net national output or income (see Appendix 1). This is a great convenience, since it means that we can measure labor’s relative price without actually knowing the average hourly wage rate, the number of hours worked, or the level of productivity — as long as we know total labor income and total national income.
But there is one more step: we must take into account all taxes and transfer payments affecting labor. This step is usually omitted from labor-cost calculations, to reduce their complexity. But taxes and benefits should be included, because they affect people’s behavior. Rather than just using gross labor compensation, we should subtract taxes on labor and add transfer payments to labor.
International Evidence for Rueff’s Law
I recently calculated labor’s share of national income in this way, back to 1929 for the United States, and back to 1960 for the United Kingdom. (An attempt to do the same for France has so far been defeated by poor data.) Exactly as theory and common sense predict, the relative price of labor is closely tied to the unemployment rate. Whenever labor’s share of income rises, unemployment goes up; whenever labor’s share of income falls, unemployment goes down.
In the U.S., for example, labor took more than 90% of national income at the depths of the Depression, and unemployment hit 25%. During the World War II boom, labor’s share of income hit an all-time low of 64% — corresponding to an all-time low in unemployment of 1%. Since the Second World War, there has been a steady uptrend in labor’s share of national income, and a parallel uptrend in the minimum unemployment rate. On average, each 1% rise in labor’s share of national income has been accompanied by about a 0.8% drop in employment. (See Graph 1.)
In the United Kingdom, the relationship is very similar, except that, until recently, labor’s share of income led unemployment by a year or two. Labor’s share remained below 69% in the 1960s, when unemployment was below 3%. After labor’s share rose to 76% in 1980, unemployment exceeded 12%. Subsequent swings in labor’s share were followed by commensurate swings in unemployment. On average, each 1% rise in labor’s share of national income was accompanied by about a 0.9% drop in employment. (See Graphs 2 and 3.)
Labor’s Share Up — Labor Income Down
Because there is a strict relation between employment and net national output or income, a rise in the relative price of labor must be associated not only with a rise in unemployment but also with a fall in real national income. In both the U.S. and the U.K., each 1 percentage-point rise in labor’s share of national income is associated with about a 2% decline in real national income — measured, for example, by the gap between potential and actual output or income (see Appendix 2). This means that when labor’s income share rises 1 percentage point, real labor income falls 1 percentage point. (See Graphs 4 and 5.)
Why does real labor income fall when labor’s share of income rises? We recall that all units of labor earn a wage based on what the last unit adds to output. For any given equipment, each extra labor unit has less equipment to work with, so adds less to output. Therefore labor must receive a lower “efficiency wage” — a smaller share of total income — as unemployment falls. But real national income and real labor income rise with employment and output, e.g., World War II. If real labor income did not rise as unemployment fell, it would mean that workers as a group offered to work an extra hour even though their real income would drop if they did. This is inherently implausible.
Labor’s share of income must stop falling at full employment since, if no more labor is forthcoming, labor’s contribution to extra output can’t decline any further. Similarly, labor’s income share rises with unemployment, because the last unit of labor hired has more capital to work with; but real labor income falls, because employment and national income are cut back, e.g., the Depression.
Solving The Benefit Puzzle
We can see the effect of government benefits by comparing labor’s share of income with the share received by employed workers. The difference is transfer payments to persons who are not employed — that is, income earned by the effort of employed workers but paid to persons who are either unemployed or outside the labor force. Such benefits represent, in effect, a purchase of labor services by the government.
The effect on unemployment cannot depend merely on the size of benefits. Transfer payments to the non-employed are much higher in the U.S. than in the U.K. — about $3100 vs. $2350 per capita in 1992, measured at purchasing power parity — yet unemployment is far lower. Moreover, such benefits expanded to almost exactly the same degree in both countries between 1960 and 1992– just over 11 percentage points of national income. But in the U.S., labor’s share of income rose less than 3 percentage points, while employed workers’ share fell more than 8 percentage points. In the U.K., labor’s share of income rose more than 6 percentage points, while employed workers’ share fell by 5 percentage points. The larger rise in the U.K. relative price of labor explains why unemployment rose more sharply in the U.K. than in the U.S. But what determines whether benefits raise labor’s share of income or reduce employed workers’ share?
Theory suggests that benefits to the unemployed create a wage floor, thereby raising labor’s share of income and unemployment. But this is not the case with benefits to persons outside the labor force. Since, for example, most workers cannot qualify for old-age benefits, such benefits do not create a wage floor and therefore are financed by reducing employed workers’ take-home pay. This reduces labor force participation and employment without increasing unemployment. (Benefits to employed workers are, in effect, paid by other employed workers, also reducing work effort but without affecting labor’s or employed workers’ shares of national income.)
A closer analysis of the U.S data confirms this. The rise in labor’s share of income is just equal to the rise of benefits to the unemployed — mostly welfare to the able-bodied — while the fall in take-home pay is equal to the rise of benefits to persons outside the labor force — mostly transfers to the aged or disabled. (For a detailed discussion, see “How Can Wages Fall While Unemployment Rises?” March 11, 1994 LBMC Report). I was unable to make a similar breakdown of U.K. benefits — perhaps some enterprising Brit will do so — but common sense suggests that benefits to the unemployed have expanded faster than benefits to persons outside the labor force.
Thus the new conventional wisdom is mistaken on two key points. First, it confuses labor’s share of income with real labor income. Benefit (or minimum-wage) laws which create an above-market wage floor do increase labor’s share of national income, but force a cutback in employment which unambiguously lowers real labor income. Second, the main difference between the United States and Europe is not the level but the kinds of social benefits. Since the 1960s, benefits to the unemployed have increased more sharply in Europe — increasing unemployment faster — while benefits to persons outside the labor force have increased more rapidly in the U.S. — causing a sharper decline in take-home pay as a share of national income.
The logic of labor-market policy is exactly the same in the U.S. and in Europe. The only difference is which policy change is more urgent. Trimming excessive unemployment or welfare benefits would reduce unemployment and necessarily raise real labor income. Cutting back benefits to persons outside the labor force would not affect unemployment, but would increase labor-force participation and reverse the decline of take-home pay as a share of national income. Only benefits conditioned on holding a job can offer a “living family wage” without raising unemployment or reducing employed workers’ share of income.
Appendix 1: Why Is the Relative Price of Labor the Same as Labor’s Share of Income?
We observe in the text that the relative price of labor is derived by dividing labor compensation (adjusted for all taxes and transfer payments affecting labor) by both product prices and labor productivity. Let W be labor compensation per hour, L the number of hours worked, P the index of product prices and Q net output. Then the “product wage” is W/P, and productivity (output per hour) is Q/L. So the relative price of labor is (W/P)/(Q/L) = WL/PQ. But WL is total labor compensation, and PQ is the value of net output.PQ (net of capital consumption and indirect taxes) is also equal to national income. Therefore the relative price of labor is the same as labor’s share of national income. As long as we know the aggregate values WL (labor compensation) and PQ (national income), we can measure the relative price of labor without actually knowing W, L, P or Q.
Appendix 2: A Note on the “National Income Gap”
For any given equipment and organization, there is a strict relation between employment and output, and therefore between employment and national income. Therefore a rise in unemployment must be associated with a decline in national output and income.
But national output and income can also change with equipment and organization, or with the labor force or education. We can isolate the effect of changes in unemployment by focusing on “the national income gap” — the difference between actual national income and potential national income at some specified unemployment rate. Both potential and actual output or income should move in parallel except for changes in unemployment.
This idea is usually expressed as the “GDP gap” — the difference between actual and potential gross domestic product. But for our purposes, we wish to focus on national income — the proceeds of GDP actually paid to either labor or capital.
For the United States we use the Congressional Budget Office’s estimate of potential GDP, adjusted to a national income basis instead of GDP. CBO’s potential GDP is not based on zero unemployment, or even on a constant unemployment rate, but rather upon an unemployment rate which ranges slowly between 5% and 6%, based on CBO’s estimate of the unemployment rate at which inflation begins to accelerate. This explains why actual national income sometimes exceeds potential national income. Since we are interested in changes in the national income gap, rather than its absolute level, these anomalies do not greatly affect our results. The CBO output gap goes back to 1949, but we have reconstructed it back to 1929 using the same relationship (see Graph 6).
For the United Kingdom, a useful series for the output or income gap back to 1960 was not available, so we used the divergence of real national income per capita from the 1960-74 trend (chosen because the unemployment rate was fairly constant during that period). U.S. data show that such a measure behaves very much like the “national income gap.” Graph 7 shows U.K. real national income, labor income and take-home pay relative to this trend.
Appendix 3: Measuring Labor’s Share of Income for the U.K.
As explained in the text, labor’s share of national income equals total labor compensation, minus taxes on labor, plus transfer payments to labor. The calculation for the United States was described in “How Can Wages Fall While Unemployment Rises?” (March 11, 1994 LBMC Report, Appendix 1).
The figures for the U.K. come from OECD National Accounts, which differ slightly from the U.S. National Income and Product Accounts. For example, we must subtract indirect taxes net of business subsidies from the OECD measure of national income to approximate national income under the U.S. definition used in this study. These items are from “Main Aggregates,” as is gross labor compensation, which for the U.K. includes an estimate for the self-employed.
Net transfer payments are from “Accounts for General Government,” and include “other current transfers” except “transfers to the rest of the world” under “Disbursements,” less “imputed unfunded employee pension and welfare contributions” under “Receipts.” The item “other current transfers, except imputed” under “Receipts” refers to the poll tax, which is treated separately.
Payroll taxes are from “Accounts for General Government.” Direct personal taxes and fees come from “Accounts for Households.” Labor’s share of personal taxes is assumed to be proportional to labor’s share of national income as defined above. (This method agrees closely with estimates of personal taxes on labor for recent years from IMF Government Statistics Yearbooks.) The poll tax is attributed entirely to labor: 75% to employed workers and 25% to recipients of transfer payments, reflecting the share of the population which is either unemployed or older than working age.
Labor’s share of national income = (labor compensation – payroll taxes – poll tax – labor’s share of personal taxes + net transfer payments)/(national income – indirect taxes + business subsidies).
Employed workers’ income = labor income – net transfers + 25% of poll tax.