What Is the Marginal Propensity to Consume (MPC)? Formula, Example, and Limits

In a simple consumption function, C = a + cYd, C is consumption, Yd is disposable income, a is consumption not explained by current disposable income in that stripped-down model, and c is the marginal propensity to consume. In a smooth model, c is the change in C associated with a small change in Yd. With observed finite changes, an analyst often calculates ΔC ÷ ΔYd over a stated interval. These are connected definitions, but the finite ratio is an average response over that interval, not proof that every additional dollar has the same effect.

The letters need definitions before the ratio can be interpreted. An analyst might study one household or a group; use a one-week, one-year, or longer response window; and measure all consumption or only selected purchases. A study that looks at retail-card transactions over two weeks asks a different question from one that tracks household consumption over a year. The acronym alone does not specify those choices.

For U.S. national accounts, the Bureau of Economic Analysis defines disposable personal income (DPI) as personal income less personal current taxes and describes it as income available to persons for spending or saving. It defines personal consumption expenditures (PCE) as goods and services purchased by or on behalf of U.S. residents. Those are useful U.S. aggregate concepts, but a household survey, a model, or another country's accounts may define income and consumption differently. A careful estimate names its variables rather than assuming every source uses BEA's labels. {source:beaPersonalIncomeOutlaysDefinitions}

MPC is about a change-to-change relationship. It is not the fraction of a household's entire income that it spends, and it does not say whether spending happened immediately or later. A small MPC over the first month can coexist with a larger cumulative response over a year if part of the added income is spent later. Conversely, a temporary jump in purchases can be followed by a quieter period if people move planned purchases forward. State whether a figure is short-run, cumulative, or annualized.

Use the same unit, population, and time window for both changes:

Finite-change MPC = change in measured consumption ÷ change in measured disposable income

Suppose a simplified household has disposable income of $3,000 and consumption spending of $2,400 over a chosen period. In a hypothetical comparison, income rises to $3,500 while spending rises to $2,580 over the same kind of period. The changes are $500 and $180, so the finite-change MPC is $180 ÷ $500 = 0.36. In this example, consumption is $0.36 higher per additional dollar of income within the stated window.

All numbers are invented, use one consistent measurement window, and hold the comparison's other influences aside for teaching arithmetic. They are not a measured household response or a current economic estimate. If income and consumption are recorded in different currencies, price bases, or periods, the ratio may be meaningless until they are aligned. For a change across time, researchers may deflate dollar amounts to separate spending volume from price increases; when the price basis is nominal, label it.

The example stipulates that the $180 increase is the consumption response. In real data, two observations moving together do not show that the income change caused the spending change. Employment, prices, credit, household composition, expected future income, or an unrelated event may also have changed. A causal MPC estimate needs a credible comparison for what consumption would have been without the income change. The arithmetic is simple; identifying the numerator attributable to the shock is the harder part.

The remaining $320 is not automatically “saving.” In a bare classroom budget with only income, consumption, and saving, the identity would put the remainder into saving. Actual household outlays also include items such as interest payments and transfers, and families can use money to repay debt or acquire assets. BEA's personal-saving measure is based on DPI less personal outlays, not DPI less PCE alone. The MPC numerator is defined consumption; it does not classify every other dollar. {source:beaPersonalIncomeOutlaysDefinitions}

Average propensity to consume (APC) compares levels: consumption divided by disposable income. MPC compares changes. In the example, the initial APC is $2,400 ÷ $3,000 = 0.80, or 80%. The after-change APC is $2,580 ÷ $3,500, about 73.7%. The MPC for the change is 0.36, or 36 cents per additional dollar. These numbers differ because they answer different questions: the APC describes a period's average spending share, while the MPC describes a response to a specified income change.

A straight-line consumption function makes the distinction visible. If C = a + cYd, then APC = C ÷ Yd = a ÷ Yd + c, while the slope c is the MPC. Unless the autonomous term a is zero, the average ratio is not the same as the slope. This equation is a teaching model, not a claim that a single straight line describes every household or period. Spending may respond differently at different income levels, and the observed ratio can also reflect wealth, credit, expectations, and timing.

A related classroom concept is the marginal propensity to save (MPS). In the narrow model where every extra dollar of disposable income is either consumed or saved, MPC + MPS = 1. That complement does not carry over automatically when “saving” is measured with a different accounting definition or when the calculation separates consumption from debt repayment, interest, taxes, transfers, or purchases of assets. Define the budget identity before using the complement.

APC can exceed 1 in a period if spending is financed from borrowing or previously accumulated resources. Likewise, an empirical finite-change MPC is not mechanically confined to a value between zero and one: timing, measurement error, offsetting income changes, a purchase of a durable good, or the chosen comparison window can produce a negative estimate or one above one. A simple model may impose bounds as assumptions. An estimate should be interpreted under its data and model, not checked against a supposed universal range.

The denominator may be a change in disposable income, a one-time transfer or tax refund, a paycheck-timing shift, or a change in wealth. These are not interchangeable. An MPC out of income asks how consumption changes per dollar of income change; an MPC out of wealth asks how it changes per dollar of asset-value change. The latter can involve an asset that is not immediately spendable, and the response may unfold over a different horizon. A report that says only “the MPC is 0.2” leaves out what changed and for whom.

The expected duration of the change can matter. A temporary payment and a lasting increase in income do not necessarily alter a household's view of resources over its lifetime by the same amount. CBO explains that, in its fiscal-policy analysis, a temporary tax cut generally has a smaller effect on household purchases than a permanent cut because it changes lifetime disposable income by less. CBO also says the direct spending effect depends on the affected households' financial circumstances. Those are modeling and evidence-based relationships used in CBO's analysis, not a promise that every temporary policy will have a specific MPC. {source:cboFiscalPolicyEffects}

Timing and predictability matter for interpretation too. Income received unexpectedly may be treated differently from income that was known in advance; an anticipated payment can be partly reflected in spending before it arrives. A household that uses a payment to pay a bill has changed its balance sheet, but that payment is not necessarily a new purchase of goods and services in the same way that PCE records consumption. Researchers therefore have to specify whether a response means purchases, broad outlays, debt repayment, or reported intentions. These are different outcomes.

For example, an analyst might separately estimate a two-week response to an EITC refund, a one-year response to a one-time rebate, and a long-run response to a permanent wage change. Those estimates should not be ranked as though they share a denominator and horizon. Name the shock, the recipient group, the spending categories, and the observation period before comparing their MPCs.

An aggregate MPC is usually a ratio of the total consumption response to the total income change for the defined group and period. If household i receives an income change ΔYdᵢ and responds with an MPCᵢ, the aggregate finite-change ratio is Σ(MPCᵢ × ΔYdᵢ) ÷ ΣΔYdᵢ, assuming the household changes are measured consistently. That is an income-change-weighted response, not necessarily the simple average of each household's MPC.

Consider a hypothetical transfer: household A receives $100 and has an MPC of 0.80, so consumption rises $80. Household B receives $900 and has an MPC of 0.20, so consumption rises $180. Total income rises $1,000 and consumption rises $260, giving an aggregate MPC of 0.26. The unweighted average of 0.80 and 0.20 is 0.50, but it does not describe this distribution of dollars. If a policy targets different recipients, the aggregate response can change even when each modeled household's MPC remains the same.

The example assumes both households receive positive income changes and uses the same consumption definition and horizon. With mixed positive and negative shocks, different population weights, or households that enter and leave the sample, the aggregate denominator and interpretation require more care. A number averaged across people can answer a descriptive question; it need not predict the response to a policy that sends a different amount to a different group.

CBO notes that, in its analysis, added disposable income is likely to boost purchases more for lower-income households than for higher-income households, while it also considers whether a tax change is temporary and the affected households' finances. This supports asking who receives an income change; it does not establish a universal ranking for every group, policy, or consumption category. The recipient mix is part of the measurement and policy question. {source:cboFiscalPolicyEffects}

An MPC describes a consumption response to an income change. A fiscal multiplier compares a broader output response with a defined policy impulse over a chosen horizon. The MPC can be one behavioral input into a simplified macroeconomic model, but the multiplier also depends on what starts the process, who receives income, imports, taxes, prices, production capacity, interest rates, policy reactions, and the way output is measured.

In the simplest closed-economy classroom model, let consumption be C = a + cY, with a fixed MPC c; let output satisfy Y = C + I + G; and hold investment, taxes, imports, prices, interest rates, and capacity conditions fixed. If government purchases G rise by ΔG, then ΔY = ΔG + cΔY because the first round of extra output becomes income and a fraction c returns as consumption. Rearranging gives ΔY ÷ ΔG = 1 ÷ (1 − c). If c = 0.60, the model's spending multiplier is 2.5. This is a conditional algebra result, not a forecast.

The assumptions do the work. The model treats a new round of output as available when demand rises, holds prices and financing conditions fixed, excludes taxes and imports, and assumes a stable marginal consumption response. If proportional taxes or imports absorb some income, the equation changes. If production is constrained, demand can affect prices rather than quantities. If monetary policy responds, interest-sensitive spending may be offset. A student should not use 1 ÷ (1 − MPC) as a direct estimate of a real-world GDP effect without checking the model's structure.

BEA explains that the size of macroeconomic multipliers is closely linked to MPC because income changes can lead to diminishing rounds of new spending and leakages through saving or purchases outside the relevant economy. Its paper is about the assumptions and use of regional input-output multipliers, and it explicitly distinguishes them from national macroeconomic multipliers. Do not substitute a regional project-impact multiplier for a national fiscal multiplier or treat either as an MPC. {source:beaMpcAndMultiplierRelationship}

CBO likewise separates direct effects on spending from indirect effects summarized by a demand multiplier. It uses evidence on similar policies for direct spending and macroeconomic models for indirect effects, and reports a range because estimates are uncertain. That separation is why an individual household's MPC, a policy's first-round consumption response, and total output response should not share one label. For the policy-level ratio and its counterfactual, see What Is a Fiscal Multiplier? Spending, Taxes, and GDP. {source:cboFiscalPolicyEffects}

One Federal Reserve Board working paper by Sahm, Shapiro, and Slemrod studies the 2008 U.S. one-time stimulus rebates. Its survey-based method translates reported spending categories and timing into an estimated aggregate MPC of about one-third after a year. The authors explain that this translation depends on assumptions about how “mostly spend” answers map to a distribution of individual MPCs, and that the no-rebate counterfactual cannot be observed with certainty. The paper addresses direct spending and debt responses; it does not estimate general-equilibrium or fiscal-multiplier effects. It is one historical U.S. episode, not a universal household parameter or current estimate. {source:fed2008RebateMpc}

A separate Federal Reserve Board note examines the policy-driven two-week delay in 2017 EITC refund issuance. The authors estimate that recipients spent about 14 to 15 cents of each refund dollar at retail stores and restaurants within two weeks. Their measure covers a subset—about one-third—of aggregate PCE, and the window is narrow. They note that scaling to total spending would imply a larger response, but that extrapolation is not directly measured by the retail-and-restaurant estimate. The result applies to that group, policy timing shift, dataset, and short window; it is not a full-year MPC for all households. {source:fed2017EitcMpc}

The two findings are not competing estimates of one identical quantity. The 2008 paper reports a survey-based one-year response to a one-time rebate across a broad recipient population; the 2017 note uses transaction data around a timing delay for EITC recipients and selected merchant categories. Time horizon, shock, sample, consumption coverage, and estimation method differ. Their value here is methodological: an MPC has to be read with its denominator and measurement window attached.

An earlier Federal Reserve working paper compares survey responses to a one-time payment and extra pay delivered through lower tax withholding. The authors report different “mostly increase spending” response shares across the 2008 and 2009 programs, while discussing the changing macroeconomic conditions and the different policy designs. Those response shares are not MPCs by themselves. This is another reason not to translate a survey answer directly into dollars of aggregate consumption without an explicit method. The FEDS paper is preliminary research by its authors, not a Federal Reserve policy statement. {source:fedStimulusDeliveryMpc}

Before using a quoted MPC, identify five items: what income or wealth changed; which households received the change; whether consumption means all PCE, a subset of purchases, or reported intentions; the time horizon and whether the figure is cumulative; and the method used to separate the shock from the counterfactual. Check whether the numerator and denominator use matching prices, currency, units, and periods. A missing answer to any of these questions can make two decimal estimates look comparable when they are not.

Next, distinguish an observed association from a causal response. A household with higher income may also have different assets, household size, employment stability, or future expectations. A payment can coincide with seasonal spending or another policy. A carefully designed study may use a natural experiment or other identification strategy, but its estimate still belongs to the study's population, shock, and time window. The 2017 EITC paper, for instance, uses a statutory delay and state/time variation but measures card-based retail and restaurant spending around refund receipt, not every form of consumption. {source:fed2017EitcMpc}

Then ask what conclusion the number supports. An MPC may help describe the first-round household consumption response in a model or policy analysis. It cannot by itself say how much GDP will change, whether inflation will rise, whether the policy pays for itself, or whether a household is financially secure. Those questions need additional models, outcomes, and evidence. The fiscal multiplier guide covers the separate output-response question; Retail Sales vs. PCE Consumer Spending: Why They Differ explains why a retail-sales measure is not the full national-accounts consumption measure. For inflation-adjusted comparisons, see CPI vs. PCE vs. the GDP Deflator: How U.S. Inflation Measures Differ.

MPC is most useful when it is treated as a defined response, not a free-standing constant. Write down the change, recipient group, consumption measure, price basis, horizon, and comparison path. Only then can the figure be compared with another estimate or used as an input to a broader macroeconomic model.