Skip to main content
Category: Effects
Type: Consumption that rises with income and does not fall by the same amount when income falls. Other people’s spending is a separate postulate
Origin: James S. Duesenberry, Income, Saving and the Theory of Consumer Behavior, Harvard Economic Studies Volume 87, Harvard University Press, Cambridge, 1949. He names the ratchet effect on pages 114–116. The saving-ratio line is fit on pages 90–91
Also known as: Irreversibility of consumption relations. A Song-dynasty sentence recorded by Sima Guang names the same household asymmetry. A 1980 ratchet principle about work targets is a different claim
Quick Answer — The Ratchet Effect is a living standard that moves up with income and does not come down by the same amount when income falls. James S. Duesenberry named it in 1949 and tied the saving ratio to current income over the previous peak. The stickiness is heaviest in a short drop. A continued fall, on his account, lets the peak fade.

​
What is the Ratchet Effect?

The Ratchet Effect is a living standard that rises with income and resists an equal cut when income falls.
The irreversibility of income consumption relations produces a sort of “ratchet effect.”
That sentence is in the conclusions of Duesenberry’s 1949 book, at the opening of the section on saving and the business cycle. The rejected idea of reversibility, on page 3, is that the change in spending from a fall in income has the same absolute size as the change from a rise. When income is below the previous peak, his account puts the adjustment on the saving ratio. People defend the higher standard by saving less. A socket wrench with a ratchet turns one way and holds on the return stroke. A household that learned a higher grocery list and a higher rent in a good year does not walk those lines back by the same dollars when the paycheck drops. That picture is an illustration. It is not one of the counts below. The other rejected assumption, on page 1, is that each person’s consumption is independent of everyone else’s. That interdependence is a different postulate. Diderot Effect is a new possession that makes the rest of a set look wrong, so spending rises. Loss Aversion is a loss judged against a reference. The object on this page is the saving ratio and the previous peak income.

​
The Ratchet Effect in 3 Depths

  • Beginner: A raise becomes the level that feels normal. When pay falls, the first move is often to cut saving, and to leave the new standard in place.
  • Practitioner: Write this year’s income, the previous peak income, and the saving line. A spending cut that matches the income drop is the reversible path. A smaller spending cut is this effect.
  • Advanced: The peak weighs most in a short drop. If income keeps falling, he says the peak fades. The 78 percent below is what the fitted line says when saving is zero and investment is not negative. It is one equation’s floor, for that fit.

​
Origin

The book is Harvard Economic Studies Volume 87, Harvard University Press, Cambridge, copyright 1949, by James S. Duesenberry. The preface is dated Cambridge, June 1948. It says the original version, under the title The Consumption Function, was a doctoral dissertation at the University of Michigan in February 1948. The same preface says Chapter V republishes material from Income, Employment and Public Policy: Essays in Honor of Alvin H. Hansen (W. W. Norton, 1948). That 1948 essay was not re-read. This page uses the 1949 book. Page 1 says the book began as a critique of the Keynesian consumption function, because two assumptions of aggregate demand theory are invalid. The first is that every individual’s consumption behavior is independent of every other individual’s. The second is that consumption relations are reversible in time. He then names three sets of facts that conflict if those assumptions are kept. Simon Kuznets’s aggregate saving and income for 1869–1929 show no tendency for the proportion of income saved to rise with income. Budget studies for 1935–36 and 1941–42 show that the saving ratio rises with income. Yearly aggregate figures since 1929, from the Department of Commerce, also show the saving ratio varying with income over the trade cycle, and he says those numerical results are not the same as the budget studies. The interdependence postulate, in Chapter III, is a different theorem. For a given relative income distribution, the percentage of income a family saves tends to be a unique, increasing function of its percentile position, and independent of the absolute level of income. That theorem is his way of holding the Kuznets result together with the budget studies. It is the neighbor’s place in the distribution. It is the other half of the book. The ratchet is the time half. On pages 3–4 he says the remaining conflict, between the yearly data and the long period, rests on reversibility. In a depression, the saving ratio can be treated as a function of current income over the highest income previously attained. He calls that an approximation that seems to work well. Past incomes lower than the current one, he argues on page 86, are not very relevant. Families whose incomes rose to a given level saved about the same amount as families already at that level the year before. That sentence is his reading of the 1941–42 budget figures. This page did not re-read the table behind it. On page 90 the period 1929–1940 covers less than one major cycle, so he does not estimate the secular piece and the cycle piece in one regression. He writes the saving ratio as a linear function of current income over the previous peak, and he fits 1929–1940. The fit gives 0.25 and 0.196. The conclusions, on page 116, write the saving ratio as 0.25 times that income ratio, minus 0.196. The minus is the sign that turns his own substitutions into the figures he prints. Putting the ratio at 1.03, his stand-in for slowly rising income, produces the .06 he prints. Putting it at 0.95 produces the 4.1 percent in the 1947 check below. The correlation for the fit is 0.9. He says the correlation is as good as the usual saving-and-income relations, and that the correlation is not the test of the hypothesis. The footnote says the computations use income and saving in the July 1947 Supplement to the Survey of Current Business, put on a per capita basis and deflated by the Bureau of Labor Statistics consumer price index. Figure 4, discussed on pages 114–116, is a diagram. The long-run line OB is drawn with a slope of .9. The short-run curves and the lettered points are not counts. In a gradual recovery, consumption climbs the short-run curve back to the old peak and then moves along OB. In a very rapid recovery, consumption can leave that path while income is rising fast. He then says income cannot keep rising at that pace once full employment is reached, because only productivity raises income. Eventually a full adjustment to the new living standard takes place, and consumption may again be read off OB. That sentence is about the return to the long-run line after a fast rise. The same pages say the ratchet keeps the economy from slipping back all the way and losing all the gains in income acquired during the preceding boom. If net investment is zero, saving must be zero, so 0.25 times the income ratio equals 0.196, and the ratio equals 0.78. Provided investment does not become negative, income will not fall below 78 percent of the previous peak. That 78 percent is a property of the fitted line. It is not a trough he measured in a table. An older sentence names the same household asymmetry and prints no ratio. Sima Guang’s Xun Jian Shi Kang records Zhang Wenjie: it is easy to go from thrift into luxury, and hard to go from luxury back to thrift. The text read here is the Chinese Text Project transcription. Zhang’s worry, in that essay, is that a family used to a high salary will not be able to cut at once if the salary ends. The essay is not the 1949 regression.

​
Key Points

Four readings keep the past peak, the neighbor, and the fitted line apart. A diagram slope is not a household rule.
1

Put the neighbor on the other line

Interdependence says the share a family saves rises with its percentile place, at a given distribution, and does not depend on the absolute income level. The ratchet says the saving ratio, in a depression, depends on current income over the previous peak. A neighbor’s new car can be the first postulate. It becomes this page when the question is what happens to your own standard after your income falls. Veblen Effect is extra demand because a good bears a higher price. A higher price tag is a third line.
2

Read the saving ratio

The 1949 claim is not that consumption is frozen. On pages 86–87, high-income families do cut some lines, particularly durable goods, and they absorb a large income cut by saving less. The fitted object is the saving ratio. Mental Accounting is money kept in labeled jars. Defending a standard by emptying the saving jar can look like a jar choice. The equation here is the aggregate saving ratio and the previous peak.
3

Give a short drop a heavier peak

On pages 87–89, a fall of a couple of years, or a fall that then sits on a low plateau, leaves the peak year with a very heavy weight. The peak does not fade unless income continues to fall steadily. A second fall makes the intermediate income matter as well as the old peak. A budget that treats a one-month gap and a three-year decline as the same case has skipped his split.
4

Leave 78 percent on the fitted line

Page 116 gets 78 percent by setting saving to zero in the 1929–1940 line, and only if investment does not become negative. The drawn long-run slope of .9 is a different mark on Figure 4. Status Quo Bias is a preference for the current arrangement. A floor implied by a national saving equation is not that preference, and it is not a rule for one paycheck.

​
Applications

Write the previous peak, this year’s income, and the saving line before you decide which dollars move. Then keep the 1949 percents on that fit.

Set the peak beside this year's pay

When a household income falls, write the previous peak income and the current income as a ratio before you cut the grocery list or the rent. His page 4 approximation is that the saving ratio in a depression follows that ratio. A cut that copies the whole income drop is the reversible path he rejected.

Let saving move, and date a second look

On his account, the first defense of a high standard is a lower saving ratio, not a full retreat in consumption. Name the saving line that absorbs the first months, and set a date to look again. Pages 87–89 give the peak its heaviest weight when the drop is short. A continued fall is the case where he says the peak fades.

Mark a raise before it becomes the peak

A new salary can reset the standard the next drop will defend. Before the raise is spent into rent, school, or a subscription, write which of those lines would still be there if income returned to the old level. The Song sentence is the household form of that worry. It still prints no saving ratio.

Keep a rising quota off this page

A target that uses this year’s result as the floor for next year’s target is Martin Weitzman’s ratchet principle, in the Bell Journal of Economics 11(1), Spring 1980, pages 302–308. The abstract says current performance is a partial basis for future targets. That object is a quota. The object here is a consumption standard and a saving ratio.

​
Case Study

Duesenberry fit the saving ratio to United States aggregate income for 1929–1940. The series is the July 1947 Supplement to the Survey of Current Business. He puts the data on a per capita basis and deflates them by the Bureau of Labor Statistics consumer price index. Current income and the previous peak are disposable income. The period, he says on page 90, covers less than one major cycle. The question is whether the saving ratio in that window moves with current income relative to the previous peak. The fit gives 0.25 and 0.196, and the conclusions write minus 0.196. The correlation is 0.9. He says that correlation is not the test. The tests he names are the budget studies, which this page does not reopen cell by cell, and predictions from the line. One prediction looks backward at a slow rise. He substitutes a ratio of 1.03 and prints .06 as the saving ratio for a period of slowly rising income. For 1880–1930 he says Kuznets’s personal saving was somewhat less than 10 percent, and that he cannot be more exact because he can only guess at the size of Kuznets’s overestimate. The .06, he says, cannot be very far off, and the facts are not inconsistent with the hypothesis to any serious extent. The intro’s Kuznets window is 1869–1929, and it is described only as no tendency for the saved proportion to rise with income. The two windows stay separate. Kuznets’s own tables were not re-read. The forward check is 1947. He takes 1947 real income per capita at 5 percent below 1946, so the ratio is 0.95. The line then gives a saving ratio of 4.1 percent. The actual figure was 5.1 percent. He says the prediction is a success if 1947 can be treated as a normal year and if non-income factors were unchanged from before the war, and that there is doubt on both counts. Some consumer goods were still in short supply, which he expects to raise the saving ratio. Income was somewhat more equally distributed, which would ordinarily lower it in the short run, though much of the lower group’s gain went to farmers, who tend to save more, and he thinks the net effect of the size distribution was probably small. Rank shifts work against saving: people whose incomes fall probably cut their saving by more than those who gain. Holdings of government bonds and bank deposits rose more than in proportion to income, while the stock market kept private securities from rising in proportion. On balance he thinks the asset-to-income ratio did not change very much. Expectations of a depression, common during much of 1947, would favor saving. Shortages and those expectations push saving up. Rank shifts and the asset side push the other way. He says there is no reason to suppose the forces balanced, and that the success may or may not have been invalidated. A real test, he writes, waits on a few more years. The lesson is the gap he does print: 4.1 percent from the peak ratio, against an actual 5.1 percent, one point apart, with those doubts left open. The 78 percent is not this result. It is the ratio on page 116 at which the same line makes saving zero, if investment is not negative.

​
Boundaries and Failure Modes

The Ratchet Effect is the wrong summary in three retellings. A short drop is treated as a permanent law. A food panel is treated as this national line. A work quota is treated as a living standard. The first boundary is inside Chapter V. The passage on pages 86–87, about absorbing an income cut by saving less and keeping the same kind of neighborhood, applies only to the upper-income groups. He says the upper 10 percent of the distribution produces almost all of the positive saving. The lower 90 percent, in his account, mostly lose income through unemployment and save very little even in good years. Families who stay out of work fade the prosperity standard as assets shrink. Families who are often in and out of work gradually cut consumption even in the months when they are employed, because assets fall and debt rises. Families with only two or three long spells of unemployment renew the peak each time full employment returns. A continued fall in income makes the peak fade. Sunk Cost Fallacy is spending more because a past outlay cannot be recovered. A living standard that fades when the drop continues is not that outlay. The second boundary is a later household test of a different habit claim. Karen E. Dynan, “Habit Formation in Consumer Preferences: Evidence from Panel Data,” American Economic Review 90(3), June 2000, pages 391–406, says a condition estimated on food consumption in the Panel Study of Income Dynamics yields no evidence of habit formation at the annual frequency. The paper also says the finding holds for several proxies for nondurables and services built from that study and from Consumer Expenditure Survey weights. The PDF was not re-read. No coefficient and no sample size from it are on this page. Yearly food in that panel is not the 1929–1940 aggregate saving ratio. The third boundary is the other English name. Weitzman’s 1980 abstract calls it an almost universal feature of economic planning that current performance is a partial basis for future targets, and he says this ratchet principle creates a dynamic incentive problem for the enterprise. A better year can tighten next year’s quota. Nothing in that abstract is a consumption standard or a saving ratio. The opening of the paper likens current performance to a notched gear that sets the departure point for the next target. That gear is the quota. It is not Figure 4. The export fails in the same place. The 0.25, the 0.196, the correlation 0.9, the .06, the 4.1 percent, the 5.1 percent, and the 78 percent are the 1929–1940 line, the Kuznets comparison he states, and one 1947 check. They are not a household, a shop, a firm, or another year. A family that cuts its saving because pay fell has not thereby been shown to sit 78 percent of the way down from its own peak.

​
Common Misconceptions

Three retellings travel with the name. One freezes consumption. One folds the neighbor into the peak. One promotes a fitted floor into a law of recessions, or into proof that the 1947 check was exact.
Pages 86–87 have high-income families cutting some spending, particularly durable goods, while they protect the rest by saving less. Pages 87–89 say the peak fades if income keeps falling, and that repeated unemployment cuts consumption as assets shrink and debt grows. The full-adjustment sentence on pages 115–116 is the return to the long-run line after a rapid rise, once only productivity can raise income. The Song sentence says a luxury habit is hard to leave. It does not say the habit never moves.
Keeping a place in the income distribution is the interdependence postulate. The percentage saved, in that theorem, rises with percentile position and does not depend on the absolute income. The ratchet is current income over the previous peak, and it is about the saving ratio when income has fallen. A neighbor’s visible purchase can reset what feels normal. It is this effect when your own past peak is what the next cut refuses to match.
The 78 percent is page 116 with saving set to zero, and only if investment does not become negative. It is not a measured trough, and it is not one household. The 1947 check is 4.1 percent against 5.1 percent, after he sets the income ratio at 0.95. He leaves the success open, because shortages, rank shifts, assets, and depression talk did not have to cancel. The correlation of 0.9 is the fit he says is not the test.
The Ratchet Effect sits beside a new possession, beside a loss judged from a reference, beside a preference for what is already in place, beside labeled jars of money, and beside a past outlay.

Diderot Effect

One new possession makes the rest of a set look wrong, so spending rises. The ratchet is what happens to a standard after income falls below a previous peak.

Loss Aversion

A loss judged against a reference. The reference here is a past peak income, and the measured object is a saving ratio.

Status Quo Bias

A preference for the current arrangement. The 1949 claim is about consumption and saving when income is below the previous peak.

Mental Accounting

Money kept in labeled jars. Defending a standard by cutting saving can look like a jar choice. The equation here is the aggregate saving ratio.

Sunk Cost Fallacy

Spending more because a past outlay cannot be recovered. A standard that fades when income keeps falling is not that outlay.

​
One-Line Takeaway

When income falls, write the previous peak beside this year’s income, and decide whether the first cut is saving or the living standard, before you treat the new level as permanent.