Marginally confusing

When the efficacy of (unconventional) monetary policy is discussed, the point that is often raised is that QE is ineffective because it directs money to the rich, who have a lower marginal propensity to consume than the poor. People argue that if only we could direct money to the poor - perhaps by throwing money out of helicopters into deprived areas - monetary stimulus would be far more effective.

The same argument appears in debates about the effect of inequality on growth. Consumption drives growth (well, it's not quite that simple, but bear with me); inequality concentrates income in the hands of a few at the expense of the many; the rich few have a lower marginal propensity to consume than the (relatively) poor many, so spending is less than it would be if inequality were lower. Therefore rising inequality impedes growth.

But here is John Cochrane:
"Didn't Milton Friedman demolish the whole concept of "marginal propensity to consume" 70 years ago?"
Well, did he? This is serious. If the concept of "marginal propensity to consume" is as dead as the dodo, how come people keep relying on it to support arguments that inequality reduces growth and monetary stimulus is ineffective?

In his Permanent Income Hypothesis, Friedman argued that rational agents would use saving and borrowing to smooth their consumption over their lifespans, eliminating income shocks and enabling them to maintain a stable level of consumption. Transitory changes in income would not affect people's long-term consumption path. The marginal propensity to consume is consequently zero.

Friedman assumed no liquidity constraints: agents are always free to borrow. He also implicitly assumed that employment and income is stable over the long-term. But people have no ability whatsoever to plan for the long-term when they are unable even to secure their income in the short-term. They may have dreams, but dreams are not rational expectations.

Before I unpick this any further, though, I need to distinguish between the marginal propensity to consume and the average propensity to consume. They are quite different concepts that are frequently confused. The average propensity to consume is the proportion of long-run average income that goes to consumption spending: the marginal propensity to consume is the proportion of any INCREASE in income that will go to consumption spending. So for example, if my expected long-run average monthly income is £1000 and my expected long-run average monthly consumption spending is £600, my average propensity to consume is 600/1000 = 0.6. If I receive a bonus of £100, and I spend £50 of that on a new pair of shoes that I would not otherwise have bought, my marginal propensity to consume is 0.5. Note this is a one-off (i.e. transitory) income increase. This is important: a permanent income increase changes the average propensity to consume. I shall return to this when considering the inequality question.

To illustrate this, consider two individuals. One is CEO of a bank, earning a telephone-digit salary, mega-bonuses and stock options. The other is a labourer on the minimum wage.

The CEO is able to meet all his everyday expenses comfortably from his income. Indeed, he can meet all exceptional needs and wants from his income too, without borrowing. He may like to have a bigger bonus, but he doesn't need it, and receiving a bigger bonus will make absolutely no difference to his spending patterns. Because of this, as his income rises his average propensity to consume decreases. And his marginal propensity to consume is zero, since no income increase will make him spend any more.

So it is not that the rich choose to save more of their income when it rises. It is just that increasing their income makes absolutely no difference to their spending, and therefore they save a higher proportion of their enlarged income. I hope this is clear.

Now let's consider our minimum-wage labourer. Let's suppose that he is able to meet his everyday living expenses from his wages. But he can't cover exceptional items. So he has two choices: he either saves routinely from his wages, or he borrows to meet unexpected expenses. Or possibly both, since unexpected expenses like the washing machine breaking down occur at random and therefore may or may not be covered by the savings built up through routine saving. If he is always able to borrow to meet spending needs, then under Friedman's hypothesis, his marginal propensity to consume is also zero, and his average propensity to consume falls as either his income or his debt rises (since debt must be serviced and therefore the proportion of income that goes on consumption in the future must fall).

I have two issues with this. Firstly, I question whether a rise in income and a rise in debt can realistically be regarded as equivalent. Debt involves incurring a future cost in order to meet consumption needs now, whereas a higher income means that consumption needs can be met now without future cost. Another way of looking at this is as a difference in price: goods bought with debt are more expensive than goods bought from income. When credit is widely used to substitute for income, there is actually consumer price inflation because goods and services are purchased with an additional debt interest charge*. This doesn't show up in any measure of consumer price inflation. Perhaps it should.

If debt and income are not equivalent, it is not reasonable to assume that someone on a low income, such as our labourer, is indifferent to how his consumption is financed. When ordinary people doing TV quiz shows are asked what they will do with the winnings, invariably they say "I will have a holiday". That is a holiday that otherwise they would have borrowed to afford (thereby paying more for it), or saved up for (thereby cutting other spending), or not taken at all. Their marginal propensity to consume is not zero. It is actually positive (if they have more income, they will spend some or all of it).

And there is another problem, too. Friedman's assumption that a labourer on a low wage can always borrow is very evidently false. People on very low incomes often can't borrow at all, or only at cripplingly high rates. And even people on higher incomes may have limits on how much they can borrow. When the limit is an affordability constraint (debt service as a proportion of income), lowering interest rates may encourage people to borrow. But then so would raising wages. Borrowing is not necessarily a response to low income - it may be a response to rising income, or expectation of rising income in the future.

So our labourer not only doesn't really want to borrow (he'd rather have a pay rise), he may not be able to. Either way, his marginal propensity to consume is positive. Give him more money, he will spend it, mostly.

There are however a couple of special cases where the marginal propensity to consume is generally zero for people on low wages. The first of these is people near retirement and anyone else who faces a sharp drop in income in the near future. Friedman argues that people plan their saving and consumption patterns over their lifetimes to ensure that they don't face a cliff edge at retirement. At the time that Friedman was writing, this was quite possibly true: people generally stayed in the same job all their lives and were therefore able to predict with reasonable certainty how their income would fluctuate over their lifespan. This sort of stability encourages both borrowing for consumption smoothing and long-term saving. But that is not how people live now. In our quest for flexibility in labour markets, we have destroyed the job security and income stability that enabled people to plan for the future in this way. Few people now stay with the same company for long, and an increasing number - particularly among those on low wages - are doing temporary, casual and insecure work. The result is a much more short-term outlook. This may seem perverse: if people don't know how long they will have their current job, we might expect them to increase precautionary saving. And some people do. But for many, chronic insecurity seems to have the opposite effect: they spend the money while they have it, and rely on state safety nets and/or debt to see them through the difficult times.

The second special case is people who are close to insolvency. These people are likely to spend any additional money they get on paying down debt (economically, this is equivalent to saving). Their marginal propensity to consume may therefore be close to zero, and their average propensity to consume falls as their income rises until they have reduced debt to a less scary level. Again, this may seem perverse, since rising income makes debt more affordable. But highly indebted people have severe liquidity constraints: being unable to borrow because you are already very highly leveraged means that you have no way of smoothing consumption, and the cost of debt service may make it impossible to save on a regular basis. Easing the liquidity constraint is most easily done by reducing debt and restoring creditworthiness. And debt is not called a "burden" for nothing: very high debt feels like a millstone round the neck. No wonder people want to get rid of it if they possibly can.

But Friedman - and Cochrane - was of course looking not at these microeconomic cases, but at the macroeconomic aggregates. So can it be true that ignoring distributional factors, overall the marginal propensity to consume is zero? I would have to say "it depends".

When there is full employment and job security, and incomes are high enough for debt to be used only to cover unexpected expenses and not to compensate for chronic funding shortfalls, then I could believe that the aggregate marginal propensity to consume would indeed be at or close to zero. And this was generally true, I think, at the time at which Friedman was writing.

But it is not true now. A large number of people are in unstable employment, and real incomes** for the majority of people have been falling for six years and, in the US, stagnating for two decades. The value of qualifications is debased, and certain skill sets are becoming obsolete, making it hard for people to predict their future employment path. Debt levels are generally far higher than they were in Friedman's time, and credit availability is more restricted than it was six years ago. A lot of people therefore face liquidity constraints. As I discussed above, when there are liquidity constraints the marginal propensity to consume is usually positive. It seems highly unlikely that the combination of the rich, those close to retirement and those close to bankruptcy would be sufficient to reduce this to zero in aggregate.

And this of course explains why people still rely on the "marginal propensity to consume" to explain the negative effects of rising inequality on growth and the ineffectiveness of monetary policy that benefits the rich. It works.

Well, it works for monetary policy, anyway. Helicopter drops or fiscal easing that targeted the poor would be a more effective stimulus than QE. Even if people used the money to pay off debt it would STILL be a more effective stimulus than putting yet more money in the hands of those who already have more than they know what to do with. I refuse to accept that distributional matters are of no consequence.

However, I have to agree with Cochrane that the marginal propensity to consume is not relevant in the inequality debate. Inequality is not a short-term problem: it cannot be fixed by transitory stimulus. It is therefore the average, not the marginal, propensity to consume that matters. And the reality is that the rich save a far higher proportion of their incomes than the poor do. Indeed at the bottom end of the income distribution, the entire income may go on consumption spending - not luxuries, just basic needs. Life is a considerable struggle for many people.

But this isn't fundamentally a problem of inequality, although inequality may be an aggravating factor. The real problem is still as it has been for thousands of years: poor people have too little money. If ALL people had the means to live, how much money the rich had would be much less important. I wonder if the present focus on inequality is a distraction from the real issue, which is the slow descent into poverty facing all too many people in the Western world.

Related reading:

The Permanent Income Hypothesis - Milton Friedman
Sacred cows and the demand for money
Capital in the 21st Century - Thomas Piketty (book)

* If the rate of interest on consumer lending were the risk-free rate, then the present value of the debt plus all future interest payments would be the same as the current price of the good or service. A risk-free borrower should therefore be indifferent as to whether purchases are financed from income or debt, although there may be tax advantages from debt financing. But for most borrowers, interest rates are far higher than the risk-free rate - the riskier the borrower and the longer-term the loan, the higher the rate. The present value of the debt plus all future interest payments is therefore considerably more than the current price of the good or service.

** I'm aware that I have only discussed income effects in this piece: wealth effects matter too, but for wealth to influence consumption significantly there must be some way of converting wealth into income. Borrowing against assets is the usual way of doing this for most people.


  1. Compared to centuries gone by, global media and celebrity culture means that the wealth of the 1% is now far more visible - to the extent that they (or rather their offspring) are flaunting it 'Harry Enfield Loadsamoney' style. Think rich kids of instagram to the features in the mainstream newspapers.

    Looking back on the economics I was taught at college and university & comparing it with my situation today, I see this huge gap between the models & assumptions used vs 'life on the front line'. Furthermore, I see a huge number of contradictions in what we were taught in the late 1990s/early 2000s. We were taught that flexible labour markets were good - without any thought for the wider consequences. Perfectly flexible labour markets where firms can hire and fire at will (for the low paid) mean they cannot plan for the future. Therefore borrowing against future income becomes a very strong assumption. But is that factored into the standard models?

    In your final couple of sentences on inequality vs long term descent into poverty, are they part of the same wider problem?

  2. Some empirical evidence on propensities to consume here:

  3. Who are these mythical "rational agents" of whom Friedman speaks? In their absence his entire premise is rendered moot.

  4. Acknowledging your second footnote, I still think it's crucial to differentiate between MPC out of income, and MPC out of wealth (net worth). A useful consumption function would incorporate both, with mutual interactions.

    Analysis using such a function would of course require a model that can incorporate heterogenous agents, with their consumption determined by their income and their wealth (and age, and...). A representative-agent model is by its very nature incapable of incorporating such a function.

    Related: Thinking about the velocity of money -- annual spending relative to the stock of wealth -- what's important is not how much of wealth is "saved" each year (whatever that could possibly mean), but how much of it is spent each year. That includes both consumption spending and investment spending -- the relative proportion of those two being a secondary consideration.

    1. I'm planning to write more about this "consumption out of wealth" lark. I don't believe it. If you don't have liquidity, you can't consume, however much wealth you have in the form of illiquid assets such as housing. People can only "consume out of wealth" if they convert that wealth somehow into income, either by selling assets or by borrowing against them. Mostly, it's the latter.

  5. I would split the house into the value of improvements, buildings, crops, etc and land/location value. Ignorance of Georgism and land value taxes are major problems.

  6. So in this way a house is not just a "hugeee consumer good."


Post a Comment

Popular posts from this blog

The FTX-Alameda nexus

WASPI Campaign's legal action is morally wrong