Wealth effects on the MPC (Marginal Propensity to Consume)

Economists have different theories to explain how wealth changes people’s marginal propensity to consume (MPC). In other words, how do people’s current wealth level affect the proportion of additional income that will be consumed vs saved?

The theories all seem logical, but mutual contradiction means that there can only be one truth. How well do these theories hold up to empirical evidence?

This is actually a very important question, as it would mean radically different optimal solutions to how Governments should act in during times of recession, how they should tax citizens, how economic incentives should be structured, and big implications on the optimal levels of wealth inequality.

Keynesian Theory

In 1936, economist John Maynard Keynes published The General Theory of Employment, Interest and Money, which introduced the concept of the marginal propensity to consume (MPC) to the world. According to Keynes, consumption spending is a linear function of a person’s income. Hence, we can use this relationship to predict that if that individual’s income increases by a dollar, this will also increase marginal consumption spending as a fraction of that newfound income, with that fraction being known as the MPC.

According to this initial formulation, the MPC is constant in the short run, and can be represented by a formula such as C = a + bY_d, where ‘a’ represents some necessary subsistence level of consumption, Yd represents income, and b being the marginal propensity to consume variable.

In his magnum opus, Keynes justified this formulation due to “fundamental psychological laws”, and reasons that the MPC must be greater than 0 but less than 1. He also argues that the MPC is constant in the short run.

The MPC was a critical part of Government policy because Keynesians, including Keynes himself, used it to justify stimulatory Government fiscal spending in periods of recession. As each unit of additional income is further spent as consumption, which becomes someone else’s additional income, and so on, there is a multiplier effect where each additional dollar of Government spending can have a disproportionately significant outsized effect in stimulating the economy and raising GDP, if it is performing under-capacity during a recession.

Economists such as Tobin have altered the consumption equation to include wealth as a variable. In this formulation, C = a + dW + bY_d, where W represents wealth and d is the marginal propensity to consume out of household wealth. (Duca 2013)

Friedman’s Monetarist Theory

However, according to economist Milton Friedman and his influential publications, such as A Theory of Consumption Function (1957), this Keynesian perspective is flawed. Households are able to borrow or save over their life cycle, and thus they will choose to alter their consumption patterns based on their expected lifetime income, with the aim of smoothing their consumption over time.

This was argued to be the expected behaviour of households due to this being the utility-maximizing strategy, because of diminishing marginal utility as a function of consumption costs. In other words, because the welfare value of the first thousand dollars of income in a time period, which may be essential for life, food, sustenance, and other basic needs, for example, exceeds the welfare gained from the next additional thousand dollars, given a limited budget constraint, households are better-off when they spread out their consumption over time. This critique is known as the permanent income hypothesis (PIH).

Hence, according to Friedman, temporary economic stimulus via fiscal policy or tax cuts, for example, will have a much lower effect than expected, since the marginal propensity to consume is also partly determined by that person’s expected lifetime wealth, of which a sudden Government stimulus would only be a small part of. In other words, the MPC will be lower than what Keynes predicts, since people would choose to save most of the stimulus in anticipation of future tax increases to pay for this Government spending.

The PIH seems to have some empirical validity. Studies from economists such as Piella and Pistaferre (2017), as well as Christelis et al. (2015) do show that people react more strongly to permanent than temporary shocks, in line with Friedman’s predictions.

The seemingly incorrect, overly optimistic estimation of the MPC is one major factor that led Friedman to conclude that fiscal stimulus often does more harm than good, since the multiplier effect is minimal, and that the solution is instead to target the growth rate of the money supply, with the goal to have currency with mostly low and steady inflation in proportion with the level of real output. This view is known as Monetarism. (Friedman 2008)

Due to the widespread influence of Friedman, this critique of the MPC has become popular, and is one important factor leading toward the “Monetarist revolution” in economics. As a result, this can be considered the “orthodox” view at the start of the 21st century, despite the Keynesian view being the original “orthodox” view in the 1960s, when Friedman began to challenge it.

However, one important logically necessary caveat Friedman forgets to explicitly mention in A Theory of Consumption Function is that this hypothesis is only true when financial frictions are sufficiently low for households to be able to access credit to borrow or save.  (Meghir 2004) If households are credit constrained, or if financial frictions or interest rates are sufficiently high, then smoothing one’s consumption function via borrowing becomes difficult.

In such a case, the MPC may be higher than Friedman’s critique predicts, since the household may desire to consume more but is unable to. We note that there is at least one particular situation during which the two factors may coincide —  when there are both high amounts of credit constraints or financial uncertainty and frictions, and when an individual’s desired consumption is higher than their current income, and that is when they are unemployed during a recession, and as in the recent 2007 recession, possibly with a foreclosed house which otherwise could be used as collateral to secure loans. This is the precise situation in which fiscal stimulus is said to be necessary and where the MPC would come into play.

Which explanation of the MPC is more accurate, and under which contexts and situations? For example, we can imagine that there can be a great deal of heterogeneity of households, some of which are credit constrained and behave more similarly to Keyne’s formulation, and some of which are not, and behave more like Friedman’s version of the PIH.

Evaluation

It is with this key question in mind that we examine the work of Dimistris Christelis et al. (2019) from the National Bureau of Economic Research (NBER), where the authors survey a representative sample of 1,264 Dutch households to find out the effects of a sudden 10% “shock” to the value of their home on their marginal propensity to consume. Housing was used as it is the primary store of wealth for most Dutch, as ⅔ of them are homeowners.

The study differs from many other studies as it accounts for individual heterogeneity of household responses. These surveys collected demographic information, income, financial wealth and debt, asking them how much they would increase or cut spending for the next year.

Through the empirical results of these papers, we can contrast them to the predictions of the PIH and MPC to verify if the permanent-income hypothesis holds for these Dutch households. In order to do this, the paper recognizes two different possible effects.

Although the PIH takes into account lifetime wealth in determining current consumption, the initial level of wealth does not change the marginal propensity to consume itself, since in all cases of an income shock, consumption will increase or decrease proportionately by the annuity value of the shock in lifetime wealth regardless of initial wealth, in this case represented by housing prices. The paper refers to this as the wealth effect.

To illustrate this technical jargon, suppose a person has 10 days left to live, and has $1000 in lifetime wealth, with the interest rate being r = 0 for simplicity. The PIH predicts that they would split this wealth evenly across time, so they would consume $100 per day.

If this person receives an extra $100, they would apply the same process and choose to consume $10 extra per day, so the MPC is $10 / $100 = 0.1, with a total consumption of $110 per day. Note that the MPC would remain 0.1 regardless of the initial level of wealth, since the calculations can be performed separately, and that the only factor changing MPC is the expected duration left to live, or age.

In addition to this wealth effect, the paper also refers to a liquidity effect: this is the previously mentioned phenomenon where liquidity and access to credit markets can affect MPC by restricting intertemporal smoothing. The value of the house can be used as collateral, which can affect the ability to borrow. If households become credit constrained and unable to borrow, they may decrease consumption in order to build up precautionary savings in the event of an emergency. The opposite effect can also happen, when an increase in housing value can reduce the need for precautionary savings and increase access to liquidity.

The authors find that there is significant variance in how different individuals respond: over 90% of households claim that their consumption would not change, while 8.4% would increase consumption due to a positive shock, and 9.4% would reduce consumption from a negative shock.

The high rates of no response to higher wealth may also be problematic for the PIH, since the hypothesis would predict that agents would increase their consumption slightly because of higher lifetime wealth. However, it may also be explained by several factors; the expected price of the good being consumed (housing) has also increased in proportion, so the budget constraint they face remains the same. The illiquidity of the house can also make it difficult to translate wealth into consumption.

The average response of the MPC is between 4.7% for positive shocks and 2.1% for negative shocks, but because the average household does not respond to the change, the ones that do change, do so greatly; 56% for positive shocks and 21% for negative shocks.

If this asymmetric result is true and informative, then this would also contradict the PIH, since the PIH predicts that a 10% shock of wealth would have a symmetrical effect on both losses and gains. Going back to our previous example, an individual with $1000 to spend over 10 days would cut daily spending by 10% to $90 or increase it by 10% to $110 in response to a 10% decrease or increase in budget – having $900 or $1100 respectively.

This asymmetric result can only make sense in the context of some effect not captured by the PIH, such as a wealth-related liquidity effect, IE: individuals being credit constrained. As the paper argues, a positive wealth shock can alleviate some previous borrowing constraint and “pent-up consumption”, allowing these households to increase their spending, thus explaining the higher MPC change for positive shocks.

However, the authors were not able to reject the null hypothesis for this particular issue at the 5% level, due to sample sizes, so it is difficult to draw any conclusions from this result alone. Despite this, it seems that this result is plausible, as other authors have reached similar conclusions, such as the work of Pialle and Pistaferre (2017) whose research displays a similar pattern of asymmetry in the same causal direction of affecting MPC. Likewise, Fuster (2021) et al’s work also finds an asymmetry between losses and gains.

Christelis et al. have also found that there is a negative relationship between MPC from wealth shocks and “cash-on-hand”. In other words, households which are cash poor have higher MPC. This relationship can be observed in the graph below, where the previously mentioned relationship between age and MPC can also be observed.

Figure 1 Four Graphs showing the relationship between the MPC, Cash-on-Hand, and Age, as a result of a positive or negative wealth shock.
Christelis, Dimitris (2019)

As previously explained, the PIH predicts that MPC will stay constant regardless of the initial level of wealth, but empirically this is not the case. The negative correlation between cash-on-hand and MPC suggests that many households in the lower cash-on-hand brackets are credit constrained. This explanation seems to fit the empirical data that households who are at the very top bracket are not credit constrained have almost zero MPC, where individuals behave more like Friedman’s formulation and prediction of low MPC.

There are possible complicating factors or issues with the methodology that the authors attempt to address. For example, the authors use regression dummies to control for the effects of income, housing wealth quartiles, financial literacy, expectations of future house prices, bequest motive, and life expectancy, finding that the addition of these controls do not change the observed result by a statistically significant margin.

It can be also argued that the MPC will be affected by the additional burden of an increased property tax due to rising property values, but the authors believe any possible effect is small enough to ignore, since the change in tax burden is “about 100 euro per year”.

Kueng

There are similar studies which seem to contradict the PIH’s predictions. For example, Kueng (2018) uses data from a personal finance website (PFW) and Consumer Expenditure Survey (CE) from the Bureau of Labour Statistics to find that consumption smoothing does not happen as predicted by Friedman’s permanent income hypothesis (PIH). As previously stated, the PIH predicts that the timing of a wealth transfer does not matter in determining consumption if such a transfer is expected, since households will choose to smooth their consumption regardless.

However, according to his study, households in Alaska who receive regular payments of an average of $1,650 per person or $4,600 per household from the Alaska Permanent Fund Dividend (PFD) have spending patterns that vary based on the timing of receiving the payment. Because the timing of these payments are known long in advance, the PIH would predict that households would not significantly vary their consumption patterns if they are not credit constrained.

However, Kueng finds that empirically, there is “excess sensitivity” to payment timings, where households significantly increase their consumption when this payment is received. For example, the MPC is found to be 25% for “nondurables and services”. Households also increase their MPC spending by 12% instantaneously in October when the payment is received, falling back to a MPC of 25% after one quarter.

Kueng argues that this “excess sensitivity” is a contradiction of the PIH, as the households studied are not credit constrained and hold substantial reserves of “liquid assets”, and thus can easily smooth their consumption despite choosing not to.

Kueng further attempts to demonstrate that these results cannot be explained by alternate factors such as “inattention, information costs, or durability of expenditures”. For example, information failures are ruled out as a possible explanation by Kueng, as consumers are responsive to the actual PFD payment rather than the difference between the PFD forecast and the actual payment. Consumers also do not respond to announcements confirming the exact value of the PFD payment in September, only to the payments itself.

Alaskans also greatly increase the rate of searching up “Permanent Fund” in Google in the months right before the payment is due, but this anticipatory effect does not lead to changes in consumption. Finally, Kueng points out that the cost of acquiring information is low, as the size of the next dividend is frequently and accurately reported by the media, as well as published on a public website.

Similarly, this phenomenon cannot be explained by durability causing indifference in timing, as the goods purchased are often nondurable. In addition, this PFD wealth effect does not coincide with other recurring annual expeditions such as taxes or Christmas.

This study also demonstrates a liquidity effect, as having few liquid assets is associated with a higher MPC for lower income households. The MPC also varies across households, but is overall “monotonically” increasing with income, with the highest income households displaying a MPC of over 50%. This result is somewhat paradoxical, as we have previously expected and found lower-income households to have a higher MPC due to “pent up consumption”, a general wealth effect, or some other factor relating to the liquidity effect. Along this line, this result is also noteworthy because several other papers have predicted and empirically found MPC to be higher for lower income groups.

 This result is explained by Kueng to be the result of lower-income households having the most to lose should they choose not to smooth consumption. He suggests that mental accounting may cause richer households to feel less guilty about “squandering” this windfall, and it is richer households that can afford to spend more lavishly on such occasions. 

Hence, according to Kueng, this result makes sense, as households that would lose the most “violate the PIH the least”, while households that would lose less also violate the PIH the most, with efficiency losses being small of around 0.1%. More research can be performed to explain the source of this discrepancy.

Regardless, this paper shows that the PIH has some difficulty fully predicting the observed empirical results, due to households not smoothing consumption in accordance with the hypothesis.

Wealth and MPC

Although this Kueng’s paper predicts that higher income households have a higher MPC in this particular situation, the vast majority of papers, including the one studied previously, show the exact opposite. There are many reasons to believe that the true relationship is likely to be that wealth and MPC are inversely correlated.

Firstly, there is the aforementioned liquidity effect. Secondly, the very wealthy often plan on leaving sizable bequests to their descendants, which incentivizes a very high savings rate and low MPC. As the poorest either do not plan on or cannot afford leaving bequests, they necessarily have to spend more on their own consumption. Thirdly, because of diminishing marginal utility of consumption and the hedonic treadmill, the wealthy often spend a smaller percentage of their income on consumption compared to the poorest. (Arrondel 2015, King 1994, Carroll 1996)

When the consumption goods in question are essential for life, the poor will be forced to purchase it, or else perish. This inelastic demand also creates a high MPC. Some economists such as Almeida (2022) and Arrondel (2015) argue that precautionary savings also leads to suppressed consumption by the poor, which can be relieved through small cash transfers, leading to a high MPC. There are other possible lines of reasoning similar to these that can lead to the same conclusion.

The empirical results generally seem to fit this reasoning. For example, a paper by Arrondele et al. (2015) from the European Central Bank (ECB) uses French household surveys such as the French Wealth Survey (Enquête Patrimoine) and the Household Budget Survey (2006) to conclude that there is a negative correlation between MPC and wealth, with The MPC for financial wealth being 11.5% at the bottom of the wealth distribution to near zero at the top.

They have also found that MPC is also much less sensitive to housing prices except for individuals at the very top of the wealth distribution, which confirms the previously mentioned results by Christellis that most households do not react to housing prices. This finding makes sense, as those at the lower end of the wealth distribution consume housing as necessary personal shelter, while those at the top end own are more likely to own housing for investment purposes, which means that this wealth is more liquid, as they are able to sell the property without incurring the costs of having to move.

Average wealth increases more than average consumption as we go up wealth brackets. However, the effect is not strictly decreasing. Households that are somewhat indebted have a high MPC, whereas households that are the most indebted have a low MPC, as they would rather pay off their debts than increase their consumption.

The ECB tries to isolate the effects of wealth on consumption, or whether it can also reflect higher levels of confidence, whose effects are temporary. The paper concludes that even after adjusting for a household’s subjective expectations as a control variable, the negative correlation of wealth on MPC remains significant.

Economists such as Carroll et al. (2013, 2017) demonstrate empirical results that agree with this assertion. In his paper, he builds a model that accounts for individual heterogeneity as well as the level of wealth and tests it with empirical data, finding that the average MPC is between 0.2 and 0.4, with wealthier households having a lower MPC than poorer households.

In the same vein, Dynan, Skinner and Zeldes (2004) found that the MPC for the poorest quintile is nearly double than that of the richest, 0.9 vs 0.2. Tullio (2014) examines the Italian Household Survey to conclude that poorer households have a much higher MPC, with the average MPC being 48%. Mian (2013) pairs economic outcomes with ZIP codes to prove similar results.

Similarly, Fisher et al. (2020) studies households in the US from 1999 to 2013 and demonstrates that MPC is lower at higher wealth quintiles, with the bottom two quintiles having MPC around 0.24, while the top two quintiles being unresponsive to income shocks at all. He also argues against the PIH, corroborating similar empirical evidence with Kueng that households change their consumption even for predictable changes in income.

As previously explained, if this relationship is true, then this may contradict the PIH, as the PIH predicts that the initial level of wealth does not matter in determining the MPC.

Policy Implications

The empirical issues we have previously shown regarding the PIH imply that the MPC may be higher during a crisis than Friedman originally anticipated. Hence, the higher multiplier effect stemming from a higher MPC suggests that that Keynesian fiscal policy may be more effective than predicted by the Monetarist consensus.

Furthermore, the fact that the MPC might be magnitudes higher for people with lower wealth is significant because it calls into question the efficacy of the standard monetary policy response to recessions. If this relationship is true, then the higher multiplier from a higher MPC would imply that it is far more effective to give the poor fiscal stimulus and monetary support.

Despite this, common stimulatory monetary policies, such as Quantitative Easing (QE), are frequently criticised as disproportionately benefiting the wealthy rather than the poor, since it is the wealthy who have the most access and exposure to the bond or financial markets which QE targets. 

Numerous important economic influencers, such as the Bank of England (2014), who calculated that nearly half of the gains went to the top 5% richest households, as well as the Dallas Federal Reserve President Richard Fisher, have agreed with these assertions. (Beldevere 2013; Frank 2012)

Nineteen other influential economists, such as Steve Keen (2015) and Ann Pettifor, have also written an open letter denouncing such economic policies that “benefits the well-off, who gain from increasing asset prices, much more than the poorest”, and that alternatives such as fiscal spending or basic income would be more effective.

This is not to mention it is precisely the people at the bottom of the income distribution who are the most desperately in need of economic assistance in a recession, and for whom each marginal dollar improves welfare the most, due to diminishing marginal utility. For these reasons, it may be important to reassess how stimulatory policy is conducted.

Conclusion

In this post, we have given a brief overview of MPC theory, the most important historical perspectives being Keynes’ initial formulation and Friedman’s Permanent Income Hypothesis. We examined a variety of reasons and scenarios, under which the PIH fails to accurately predict the empirical data, due to reasons such as wealth, liquidity effects, or psychological framing. In such cases, the MPC may behave closer to Keynes’s formulation, which predicts a larger MPC.

We also examine some of the ways in which MPC and wealth can interact, one notable example is that there is a negative correlation between the two. As poorer households have a higher MPC, and hence a larger multiplier effect, as well as gaining more welfare per dollar spent due to diminishing marginal utility, we make the case for a re-examination of Government policy, advocating that there ought to be room for some level of stimulatory fiscal policy in recessions that target the poor, as an alternative to less effective purely monetary policy or QE that also exacerbates wealth inequality at the expense of the poor.

Bibliography

Arrondel, Luc, Pierre Lamarche, and Frédérique Savignac. “Wealth effects on consumption across the wealth distribution: empirical evidence.” European Central Bank. (2015) Available at https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp1817.en.pdf

Belvedere, Matthew J. “QE Halt ‘Too Violent’ for Market: Fed’s Fisher.” CNBC. CNBC, May 20, 2013. https://www.cnbc.com/id/100749826.

Carroll, Christopher, Jiri Slacalek, Kiichi Tokuoka, and Matthew N. White. “The distribution of wealth and the marginal propensity to consume.” Quantitative Economics 8, no. 3 (2017): 977-1020. Available at http://www.econ2.jhu.edu/people/ccarroll/papers/cstwMPC.pdf

Carroll, C. and Kimball, M., 1996 . On the concavity of the consumption function. Econometrica, 64(4), 981-992

Christelis, Dimitris, Dimitris Georgarakos, Tullio Jappelli, Luigi Pistaferri, and Maarten Van Rooij. Wealth shocks and MPC heterogeneity. No. w25999. National Bureau of Economic Research, 2019.

Duca, John V., and John Muellbauer. “Tobin LIVES: Integrating evolving credit market architecture into flow of funds based macro-models.” (2013).

ENGLAND, BANK OF. “Quarterly Bulletin.” 2014, Bank of England, London (2014).

Friedman, Milton, and Anna Jacobson Schwartz. A monetary history of the United States, 1867-1960. Vol. 14. Princeton University Press, 2008.

Friedman, Milton. “Introduction to” A theory of the consumption function”.” In A theory of the consumption function. Princeton university press, 1957.

Fisher, Jonathan D., David S. Johnson, Timothy M. Smeeding, and Jeffrey P. Thompson. “Estimating the marginal propensity to consume using the distributions of income, consumption, and wealth.” Journal of Macroeconomics 65 (2020): 103218.

Fuster, Andreas, Greg Kaplan, and Basit Zafar. “What would you do with $500? Spending responses to gains, losses, news, and loans.” The Review of Economic Studies 88, no. 4 (2021): 1760-1795.

Kueng, Lorenz. “Excess sensitivity of high-income consumers.” The Quarterly Journal of Economics 133, no. 4 (2018): 1693-1751.

Frank, Robert. “Does Quantitative Easing Mainly Help the Rich?” CNBC. CNBC, September 14, 2012. https://www.cnbc.com/id/49031991.

Mian, Atif, Kamalesh Rao, and Amir Sufi. “Household balance sheets, consumption, and the economic slump.” The Quarterly Journal of Economics 128, no. 4 (2013): 1687-1726.

Carroll, Christopher, Jiri Slacalek, and Kiichi Tokuoka. “The Wealth Distribution and the MPC: Implications of the New EU Wealth Survey Cross-Country Evidence.” (2013).

Keen, Steve. “Better Ways to Boost Eurozone Economy and Employment.” Better ways to boost eurozone economy and employment. Financial Times, March 2015. https://www.ft.com/content/7bc99348-d40b-11e4-99bd-00144feab7de#axzz3Va5I8EHR.

Kaplan, Greg, Giovanni L. Violante, and Justin Weidner. The wealthy hand-to-mouth. No. w20073. National Bureau of Economic Research, 2014.

Keynes, John Maynard. 1936. The General Theory of Employment, Interest and Money. Wordsworth Classics of World Literature. Herts, UK: Wordsworth Editions.

Meghir, Costas. “A retrospective on Friedman’s theory of permanent income.” The Economic Journal 114, no. 496 (2004): F293-F306.

Paiella, Monica, and Luigi Pistaferri. “Decomposing the wealth effect on consumption.” Review of Economics and Statistics 99, no. 4 (2017): 710-721.

Karen E. Dynan, Jonathan Skinner, Stephen P. Zeldes, 2004. “Do the Rich Save More?,” Journal of Political Economy, University of Chicago Press, vol. 112(2), pages 397-444, April.

Vanda Almeida “Rising Income Inequality and Aggregate Activity (Note).” BSI Economics. Accessed March 7, 2022. http://www.bsi-economics.org/628-income-inequality-activity.

Jappelli, Tullio, and Luigi Pistaferri. “Fiscal policy and MPC heterogeneity.” American Economic Journal: Macroeconomics 6, no. 4 (2014): 107-36.

King, Mervyn. “Debt deflation: Theory and evidence.” European Economic Review 38, no. 3-4 (1994): 419-445.

---