Reasons for the fall of the saving rate

From The saving decline: Macro-facts, micro-behavior (PDF) by David Bunting in the Journal of Economic Behavior & Organization:

Between 1952 and 1984 the aggregate personal saving rate as calculated from National Income and Product Account (NIPA) data by the Bureau of Economic Analysis (BEA) averaged 9.0 percent... However, after 1984 the personal saving rate collapsed as the BEA rate fell nine percentage points ...

Reasons for this rise and fall remain controversial. While relatively steady saving rates facilitated development of “permanent” spending theories during the 1950s and 1960s (Modigliani and Brumberg, 1954; Friedman, 1957), these theories have been unable to explain shifts in saving behavior in the 1980s and the collapse in saving rates thereafter.

Thought it was interesting: Bunting opens a door to the rejection of Friedman 1957.

But before that, he writes: "Reasons for this rise and fall [of the saving rate] remain controversial." And that brings me to this resolution of that controversy:

From American Standards of Living: 1918-1988 (Google book) by Clair Brown, describing differences between 1973 and 1988:

As earnings failed to continue their rapid upward climb, families sought to augment income through more hours of employment, especially for wives... Because spending rates were lower [and saving rates were higher] in 1973 than in 1988, emulation of 1973 consumption patterns would have required families to reduce their consumption in many areas, including shelter, household operations, and food (per capita).

Instead, families increased their spending rate rather than reduce their consumption standards. Overall, expenditures rose faster than incomes... The difference was financed by reduced savings and a smaller family size along with employer-provided health insurance covering more health care expenses and energy-saving advances in fuel and utilities.

A google search turned up the two links and I happened to read them one after the other. They seem to fit together like question and answer.

Tweakin' the bull

From the Congressional Research Service: The Fall and Rise of Household Saving, a PDF by Brian W. Cashell, Specialist in Macroeconomic Policy:

From Do the Rich Save More?, a PDF by Dynan, Skinner, and Zeldes:

From Rama Bijapurkar, a breakdown of income, saving, and spending in India:

Household income and savings distribution, from NDSSP (adjusted), 2003 -4, and expenditure distribution
from NSS

Deciles	Income Distribution(%)	Saving Distribution(%)	Expenditure Distribution(%)
1 (lowest)	2.0	0.6	2.5
2	3.2	1.4	3.8
3	4.1	2.2	4.7
4	5.4	3.6	6.0
5	6.2	4.5	6.8
6	8.8	7.3	9.3
7	8.4	7.1	8.9
8	11.9	11.1	12.2
9	15.8	16.8	15.5
10 (highest)	34.1	45.3	30.4
Total	100	100	100
Top 20%	49.9	62.1	45.9
Top 5%	22.7	31.4	19.9
Top 1%	8.6	12.6	7.3

The higher the income per capita per household, the higher their savings as well. 34% of total income is earned by top 10% of the households. When we look at the saving pattern across economic status, it is found that as high as 45.3% of the savings comes from the richest 10% of the households.

The poorest 20% of the households contribute just 5% to India's total personal disposable income. In case of household saving also, only 2% of the savings come from these households

How I interpret the numbers: The first decile receives 2.0% of the income, does 0.6% of the saving, and does 2.5% of the spending. The second decile... etc.

Note that the lower 8 deciles all spend more and save less than their share of income. The top two deciles spend less and save more than their share.


But I guess no matter how much data you gather in support of the Marginal Propensity to Consume, none of it matters because as Paul Krugman says, "It’s true that at any given point in time the rich have much higher savings rates than the poor. Since Milton Friedman, however, we’ve know that this fact is to an important degree a sort of statistical illusion."

To an important degree.

To what degree. Half? Let's say half. So then half of the evidence supporting the MPC is a statistical illusion. It has no significance.

So the other half is *not* illusion. It is clear, it has significance, and it stands in support of the MPC.

It's all in how you tweak the bullshit, isn't it?

Why is there a discrepancy, and where did the money go?

Following McConnell mostly (though McConnell does not use inflation-adjusted values), Graph #1 shows Personal Consumption Expenditures versus Disposable Personal Income as a scatter plot. The blue dots represent Personal Consumption Expenditures. The block dots show where the blue dots would have been, if none of disposable personal income had been saved.

Graph #1

Here's the Google spreadsheet.

To me, this graph is not a good way to see whether the Marginal Propensity to Consume falls as income rises. It shows the average consumption of everybody, rich and poor alike. To see differences of income, it depends on the passage of time. But surely, conditions were much different in the 1960s than they were 40 years later!

To me, Graph #1 makes it difficult to see the relation between the two sets of numbers. If the black dots were not there, could you really tell that the blue dots are slightly below the 45-degree line? When I first did this graph I had the axes reversed, and no black dots. The blue dots were running above the 45-degree line, but I couldn't tell.

I prefer to look at the two sets of numbers as a ratio, like this:

Graph #2: Real Personal Consumption Expenditures relative to Real Disposable Personal Income
Click Graph for FRED Source Page

Here you can see an unusual down-loop created by World War Two spending in the 1940s. You can also see, pretty unusual too, personal consumption expenditures reaching 100% of disposable income just at the end of the Great Depression in the 1930s. (The end of the gray bar part of it, I should say.)

Other than that, there seems to be a pretty strong downtrend until the mid-1970s, and a pretty strong uptrend since the early 1980s. The downtrend of consumer spending (relative to income) implies an increase in saving in the early years. The uptrend of spending in the later years implies a decrease in saving. And sure enough:

Graph #3: Two Versions of the Saving Rate

The "saving rate" runs clearly up till the mid-1970s, and clearly down after the early 1980s. Until the crisis, anyway. Even the late-1970s hump on Graph #2 is matched by an inverted hump on Graph #3!

Does this have anything to do with the Marginal Propensity to Consume? Seems to me it has more to do with general economic conditions and economic policies.

Clearness of mind on this matter is best reached, perhaps, by thinking in terms of decisions to consume (or to refrain from consuming) rather than of decisions to save.

- J.M. Keynes

I thought this post was finished before the Keynes quote above. But during repeated proofreads, those last two graphs fascinated me repeatedly. See how the one line goes down gradually, then up fast? The other one goes up gradually and then down fast. Like mirror opposites. Intriguing.

True: When spending goes up as a share of income, saving must go down. And when spending goes down as a share, saving must go up. Nonetheless.

Consumption is a high percentage of income, up near 100%. And the saving rate is a low percentage of income, down near zero. Still, there is something about the changes that cries out to me.

I went back to FRED and put the line from Graph #2 and the red line from Graph #3 together on a new graph. Chopped off everything before 1947 to get rid of the World War Two woopsie there, and to have the two lines start and end pretty much together. And I used annual data for both, to eliminate a distraction.

So, I put 'em on a graph. At first glance, then, the consumption-relative-to-income line (from Graph #2) was up high on the new graph and the saving rate (from Graph #3) was down low, with a lot of air between the two. Plus, they went in opposite directions.

To make the saving rate go in the same direction as the consumption-to-income line, I decided to use the minus of the saving rate. And I wanted to add about 100 to this mirrored saving rate, to move it up and get it closer to the consumption/income line. So what I graphed for saving was: 100 minus the saving rate.

It came out close. But not close enough. I played with it a bit and subtracted some from the hundred I added. The two lines are now very close:

Graph #4: Comparison of Personal Consumption Expenditures relative to Disposable Personal Income (ratio in blue) and Personal Saving as a Percentage of Disposable Personal Income (red) with saving inverted and shifted up to get it close to the blue line.

Two observations, maybe three.

First: Yes, certainly we should expect to see this similarity. We are looking at what we spend and what we save, as percentages of the income that we choose either to spend or save. Of course. Still, actually seeing the similarity is way better than just expecting it to be there. (I'm not big on expectations.)

Second: Close as they are, the red line gains on the blue line for the whole period shown. Or for sixty years anyway, from 1947 to the crisis. The red line starts out below the blue line. It slowly closes the gap but remains below until about 1980. After 1980 the red line runs above the blue consistently, to the crisis. And the gap widens.

Yeah, the lines cross at 1980, but that's not because of anything Reagan. I could make the lines cross earlier or later by subtracting a little more or a little less than 3 (see the last term in the second formula line, in the blue border across the top of the graph).

Forget 1980. What this graph shows is that the red line gains on the blue line for the whole 60 years between the Second World War and the Global Financial Meltdown.

What does that mean? I dunno, but I'm thinkin about it. The red line is gaining on the blue: The saving rate is gaining on the consumption rate. Except that the saving rate here is inverted, because I minus'd it.

Odd, isn't it? Consumption and saving together add up to disposable income. Our only choices are to spend the income or not spend it, to consume or to refrain. You would think that if the one goes up by n% of income, then the other has to go down by the same n% of income.

If consumption goes up, saving should go down. If consumption goes down, saving should go up. Invert saving, and the two should move together. And, yes, they do move together. But the red line is gaining on the blue.

Fascinating.

Here: Before 1980, consumption (blue line) decreased. The red line went down more slowly, so the gap between red and blue disappeared. But the red line is inverted saving. Before 1980 the red line went down slower than consumption; this means that saving went up slower than consumption went down.

Something similar happened after 1980. And all the while, the red line was gaining on the blue.

Okay, let's put numbers on it. From 1947 to 1981 the blue line fell from 92% of disposable personal income to 86%. That's a drop of 6% of income.

During those same years the red line fell from 90% of income to 86%, That's a drop of 4% of income and it means that saving increased by 4% of income.

Fine. But consumption (the blue line) fell by 6% of income while saving increased by only 4% of income. What happened to the other 2%?

More numbers: From 1981 to the 2001 recession the blue line went from 86% to 93%, an increase of 7% of income. During those same years the red line went from 86% to 94%. The red line went up one percent (of income) more than consumption went up. That means saving went down one percent (of income) more than consumption went up. We lost another one percent!

We lost two percent of disposable personal income in the years before 1981 and another 1% in the years after 1981. That's three percent of income, total, that's missing.

Pretty weird.

One thing that should be obvious in all this: There was no great change around 1980. Income disappeared in the years before 1980, and income continued to disappear in the years after 1980.


Okay. What Graph #4 shows is that the red line gains on the blue line for the whole 60 years. The red line is saving, inverted. It's an indicator of income not saved.

The blue line is income spent.

If the red line gains on the blue line for 60 years, it means that "income not saved" gained on "income spent" for 60 years.

Why is there a discrepancy, and where did the money go?


Wait a minute. One more graph. I took the saving rate and added it to consumption as a percent of income. The money we spend plus the money we don't spend, together. It should equal 100% of the income we started with. But it doesn't:

Graph #5: Personal Consumption Expenditures as a Percent of Disposable Personal Income, plus the Saving Rate, together add up to less than 100% of Disposable Personal Income, and fall from 99% to 96% between 1947 and 2007.
Click Graph for FRED Source Page

Why is this line less than 100%. And why is it falling?

Is Wikipedia pulling my leg?

Scott Sumner says "give up on that MPC stuff, it was discredited decades ago".

Paul Krugman says

It’s true that at any given point in time the rich have much higher savings rates than the poor. Since Milton Friedman, however, we’ve know that this fact is to an important degree a sort of statistical illusion. Consumer spending tends to reflect expected income over an extended period.

So that gets us to Milton Friedman.

From the Handbook of consumer finance research, edited by Jing Jian Xiao, this note:

So that gets us to the permanent income hypothesis.

The Friedman reference is to: A theory of the consumption function. Princeton, NJ: Princeton University Press.


Note that the excerpt from the Handbook of consumer finance research is based on "perceived future income"... On expectations. I'm not big on expectations.

Anyway, this is some of what Wikipedia has to say on the Permanent Income Hypothesis:

The permanent income hypothesis (PIH) is a theory of consumption that was developed by the American economist Milton Friedman...

Friedman concluded that the individual will consume a constant proportion of his/her permanent income; and that low income earners have a higher propensity to consume; and high income earners have a higher transitory element to their income and a lower than average propensity to consume.

1. "low income earners have a higher propensity to consume"
2. "high income earners have ... a lower than average propensity to consume"

Sounds to me like it supports the idea of the Marginal Propensity to Consume.

Is Wikipedia pulling my leg?

Interest Income by Quintile

Didn't have to pull too many teeth to find this PDF from the BLS. (Data for 2009.)

Under "sources of income" the PDF includes "Interest, dividends, rental income, other property income" for "All consumer units" and by quintile.

Under "Consumer unit characteristics" the PDF provides the total income before taxes for each quintile.

I took a look at the ratios of those numbers:

Graph #1

Here's the spreadsheet.

Considering interest, dividends, rental income, and other property income as returns to various forms of saving, I am calling the total "interest income" and using it as a proxy to estimate various forms of savings by income quintile. For what it's worth.

McConnell discredited?

Two excerpts from Malcolm R. McConnell's 1975 textbook, and a contradiction I let slide in yesterday's post:

First, from the text above McConnell's Figure 11•2:

Many economists now feel that the MPC and MPS for the economy as a whole are relatively constant. Statistical data such as those of Figure 11•2 are consistent with this position.

In other words, statistical data supports the view that the MPC does NOT change.

Second, from the text below Mcconnell's Figure 11•2:

[The figure] suggests that households spend a larger portion of a small income than they do of a large income.

In other words, statistical data supports the view that the MPC *DOES* change.

Wow. Talk about seeing both sides of an issue!

MPC discredited?

Steve Roth ponders the Marginal Propensity to Consume (MPC) here:

...since those workers have a high propensity to spend their income, all things being equal that distributional shift should mean there’s ... a virtuous cycle.

and again here:

Sumner, Drum, and Krugman all seem to think that the distribution/MPC/velocity argument has no legs. They’re quite categorical about this.

It is a big deal, because a lot of things fall into place if the MPC idea holds good. If it doesn't hold good, people should be a little more explicit about how it fails than Paul Krugman is:

...that the rich spend too little of their income. This hypothesis has a long history — but it also has well-known theoretical and empirical problems.

If it doesn't hold good, people should be a little more explicit about how it fails than Scott Sumner is (at the first link):

you really need to give up on that MPC stuff, it was discredited decades ago.

The quest for a better explanation of what's wrong with the MPC idea is Roth's focus in comments at the first link (in response to Sumner)...

I’ve poked around a lot looking for a straightforward, cogent refutation (that doesn’t make assumptions that I consider to be questionable). Any leads for me?

... and at the second link, in its entirety. Roth writes:

Can folks (especially those who don’t believe this argument) point me to what might be considered definitive takedowns? I have notions about what they might say, but want to see the best argument(s) out there.

I want to see too, Steve. But I suspect they're a lot like the arguments for free trade:
1. "It's obvious"
2. "We'll get to that later" (and later, they only say "It's obvious"), or
3. something incomprehensible, just to shut us up.


Roth is looking for "the slam-dunk argument that has Krugman and especially Drum convinced that the MPC argument doesn’t hold water".

You know what I like? I like it that even though the MPC explains a lot in a most satisfactory way, Roth is willing to accept that the idea is garbage -- *IF* someone can make the convincing argument. That's the right way to do economics: "always approaching" the best answer.

Anyway, I decided to look through some textbooks for views on the Marginal Propensity to Consume. Maybe that would be the easy way to find the concept discredited, I thought. I went first to McConnell's 1975 edition of Economics:

The proportion, or fraction, of any change in income which is consumed is called the marginal propensity to consume (MPC). Or, alternatively stated, the MPC is the ratio of a change in consumption to the change in income which brought the consumption change about; that is:

This is a bit different from Steve Roth's (admittedly brief) "poorer people spend a larger share of their income/wealth than richer people." And maybe that opens up a door. Because even of the marginal propensity to consume *is* junk, Roth's non-marginal observation may still be true.

In his 1975 textbook McConnell wrote:

Economists are not in complete agreement as to the exact behavior of the MPC and MPS as income increases. For many years it was presumed it was presumed that the MPC declined and the MPS increased. That is, it was felt that a smaller and smaller fraction of increases in income would be consumed and a larger and larger fraction of these increases would be saved. Many economists now feel that the MPC and MPS for the economy as a whole are relatively constant. Statistical data such as those of Figure 11•2 are consistent with this position.

These "feelings" that McConnell describes -- is that all there is?

McConnell's Figure 11•2:

Source: Economics by Campbell R. McConnell (Sixth Edition) Page 229

Reiterating the notes below the figure, a dot would be on the 45-degree line (the black line) if every dollar of disposable income were spent. Where a dot falls below the black line it indicates that some part of disposable income has been saved. The farther the dot falls from the black line, the greater the amount that was saved.

McConnell says the brown line "suggests that households spend a larger portion of a small income than they do of a large income." I agree: The brown line is not a trend-line for the dots. The brown line is a "constant percentage of income" line. For the years since 1946 through 1966 the dots are generally above the brown line. For the years after 1966 the dots are generally below the brown line. At the higher incomes, saving was generally greater than at the lower incomes.

But somehow, that doesn't feel like a slam-dunk argument.