## _wp12199 — Distributional Effects of Monetary Policy Shocks (selected sections)

## Source details

**Canonical URL:** [_wp12199 — Distributional Effects of Monetary Policy Shocks (selected sections)](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12199.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12199.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12199.pdf.json)

---

### Survey, data, and measurement (Sections 2.1–2.3; Measuring Inequality)
- Data source and sample construction
  - Consumer Expenditure Survey (CEX), Interview Survey component only; quarterly aggregation since 1980.
  - About 1,500-2,500 households surveyed per month; each household interviewed once per quarter for at most five consecutive quarters (first interview dropped).
  - CEX covers up to 95 percent of typical household consumption; does not include the very upper end (top 1 percent).
  - Non-durable consumption, services, durables, and a broad total expenditures measure (adds mortgage/rent, health, education) constructed.
  - Income asked in interviews 2 and 5; financial data only in interview 5; nominal variables deflated using the CPI-U.
  - Data cleaning procedures eliminate about 7 percent of initial observations.

- Inequality measures used
  - Three measures for each outcome: Gini coefficient (0–1), cross-sectional standard deviation of log levels, and (log) 90th percentile minus (log) 10th percentile (zero observations dropped where needed).
  - Income forms: labor earnings and pre-tax total income (after-tax used in robustness checks).
  - Consumption forms: narrow "consumption" (non-durables + services + some durables) and broad "total expenditures" (includes large durables, mortgage, tuition, health).
  - Baseline measures are raw (no household controls); controls used in robustness checks.

- Key unconditional statistics (1980Q1–2008Q4)
  - High correlations across income inequality measures: correlations of 0.89 or above.
  - Example correlations (Table 1 Panel A): Income Inequality Corr(SD,Gini) 0.91; Corr(SD,90th-10th) 0.93.
  - Cross-form correlations vary: Corr(Income, Earnings) by Gini = 0.82; by SD = 0.22; by 90–10 = 0.12.
  - Volatility (standard deviations, Table 1 Panel C): Income Inequality SD = 0.05 (SD-measure), Gini = 0.02, 90th-10th = 0.10; Consumption Inequality SD = 0.01, Gini = 0.01, 90th-10th = 0.03.
  - Unconditional correlations with macro variables (HP-filtered): correlations with inflation and interest rates are very small and somewhat negative in many cases; expenditure and consumption inequality are more negatively correlated with unemployment (e.g., Expenditures Inequality Corr(UE,SD) = -0.32).

---

### Identification of monetary policy shocks and sample extension (Section III; shock construction)
- Shock identification
  - Primary shocks: Romer and Romer (2004) innovations to the Federal Funds rate (FFR) orthogonal to Fed Greenbook forecasts.
  - RR residuals estimated using real-time Greenbook forecasts of real output growth, GDP deflator inflation, and current-quarter unemployment; residuals defined as monetary policy shocks.

- Extension and quarterly aggregation
  - RR shocks extended to December 2008 by adding FOMC meeting actions, extending Greenbook forecasts to December 2006, and using Blue Chip Economic Indicators when Greenbook forecasts unavailable (2007–2008).
  - Quarterly shock series constructed by averaging orthogonalized FFR innovations in each quarter; plotted starting 1980Q1.
  - Sample endpoints and lag choices:
    - Full sample: 1969:Q1 to 2008Q4.
    - Restricted sample: since 1980Q1.
    - Over full sample use J = 8 and I = 12; over restricted sample use J = 4 and I = 8.
  - Standard errors: Newey-West (1987); confidence intervals via bootstrap.

- Macroeconomic responses to contractionary shocks (1969:Q1–2008Q4)
  - Contractionary shocks: lower real GDP, consumption and investment; raise unemployment.
  - Short-term and long-term interest rates rise immediately; inflation declines after a two-year lag.
  - Asset-price responses: real stock prices decline on impact then eventually rise; real housing prices decline gradually after one to two years.

---

### Main empirical findings on inequality (Section III.2 and 3.3)
- Baseline result
  - A 1 percentage point (100 b.p.) contractionary monetary policy shock raises observed inequality across households in:
    - total income,
    - labor earnings,
    - total expenditures,
    - consumption.
  - Effects are statistically significant and highly persistent for consumption and expenditure inequality; point estimates for expenditures consistently larger than for other forms.

- Percentile-level dynamics (P10, P25, P50, P75, P90)
  - Labor earnings:
    - After about two years, upper-end wage earnings rise while lower-end earnings fall.
    - Median response essentially zero; earnings at the 90th percentile rise by almost the same percentage as the decline at the 10th percentile.
    - Example long-run magnitudes: labor income at P90 rises ~4 percent; at P10 falls ~5 percent (consumption responses closely mirror these magnitudes).
  - Total income:
    - Effects present but smaller than for earnings; the 90th percentile shows the largest positive response starting two years after the shock; the 10th percentile is lowest.
    - Difference between 90th and 10th percentiles in total income is about half of that implied by earnings differentials.
    - Transfers partially offset declines in income at the bottom: transfers drop in the first year, then rise approximately two years after the shock (temporarily over the full sample, persistently since 1980).
  - Consumption and expenditures:
    - Consumption percentile responses align closely with earnings percentiles and appear earlier than wage responses.
    - Expenditures at P90 show a very large increase after contractionary shocks—much larger than income or earnings changes—explaining large and persistent rises in expenditure inequality.
    - Examination of expenditure composition does not find that upper-end households’ spending composition explains their large expenditure response.

- Income composition by consumption quintile (Table 3, 2000s)
  - Labor income shares: Quintile 1 = 0.435; Quintile 3 = 0.740; Quintile 5 = 0.812.
  - Other income share (transfers and similar): Quintile 1 ≈ 0.460; Quintile 5 ≈ 0.065.
  - Business income share rises from 0.019 (Q1) to 0.051 (Q5); financial income share falls from 0.086 (Q1) to 0.071 (Q5).
  - Implication: bottom quintile relies much more on transfers; top quintiles derive larger fraction from labor and business/financial sources.

---

### Mobility and short-run composition effects (Section 3.4)
- Mobility trends (1980Q1–2008Q4)
  - Quarterly transition probabilities constructed for consumption quintiles show declining mobility over time for all but the bottom quintile.
  - Example: middle quintile persistence rose from ~35 percent in 1980 to ~45 percent in 2008.
  - Bottom quintile upward mobility to middle or above has been about 5 percent per quarter.

- Effect of monetary shocks on mobility
  - Monetary policy shocks increase short-run movement within the distribution: frequency of remaining in same quintile falls for all quintiles immediately after shocks.
  - Effects on transition probabilities dissipate after about two years; this short-run mobility contributes to volatility of percentile impulse responses in the first two years.

---

### Contribution of monetary policy shocks to inequality dynamics (Section 3.5)
- Forecast-error variance decompositions
  - Monetary policy shocks account for most median estimates lying between 10 percent and 20 percent of the variance of inequality measures at different horizons (Figure 7).
  - Appendix results: comparable 10–20 percent contributions of monetary shocks to variances of many macro variables (except short-term rates and financial income where contribution differs).

- Historical counterfactual (predicted vs actual; 1985–2005; Figure 8)
  - Monetary shocks explain little of income, earnings, and expenditure inequality variation until the mid-1990s (with some higher-frequency early-1980s effects).
  - Since mid-1990s, monetary shocks can account for gradual changes in income and earnings inequality (e.g., rise in 2005–2006, decline into 2008).
  - For consumption inequality: monetary shocks explain many short-lived deviations in the 1980s and can almost fully account for behavior since 2000.
  - Expenditure inequality: monetary shocks account for little of historical variation except for decline since 2005.

---

### Wealth- and net-worth–related channels (Section 4 and 3.3 caveats)
- Data limitations
  - CEX lacks reliable household wealth and cash-holdings measures; cannot directly identify financial-market-connected households.
- Saver–borrower (high vs low net-worth) exercise (definitions used)
  - Low net-worth proxy: age 30–40, white male head, no financial income, positive mortgage payments.
  - High net-worth proxy: age 55–65, white male head, positive financial income, no mortgage payments.
- Findings (Figure 9)
  - Little evidence of differential labor earnings or total income responses across these groups.
  - High net-worth households display larger increases in consumption and total expenditures relative to low net-worth households after contractionary shocks—consistent with redistribution of nominal wealth boosting high net-worth spending and reinforcing increases in inequality.

---

### Permanent inflation-target changes (Section 5)
- Alternative measures of the Fed’s inflation target
  - Coibion and Gorodnichenko (2011) (CG): time-varying Taylor-rule reaction function estimated with Greenbook/Blue Chip forecasts.
  - Ireland (2006): small-scale New Keynesian model estimated by maximum likelihood to recover implied inflation target.
  - Both measures: rising targets in 1970s (peak ~8 percent), rapid declines in early 1980s (Volcker), settling near 2 percent by 2005.

- Effects of a 1 percentage point decrease in the inflation target (cross-sectional standard deviations; Figure 11)
  - Both measures: disinflations raise long-run levels of income and consumption inequality.
  - Quantitative size: inequality of each type rises approximately 2 percent.
  - Forecast-error variance contributions of inflation-target shocks ≈ 10–20 percent across horizons; even higher for consumption in some specifications.
  - Historical contribution:
    - Early-1980s disinflation explains a very large, persistent increase in consumption and expenditure inequality; Ireland’s measure can fully account for the early-1980s large increase in income inequality, CG accounts for a smaller fraction (timing differences).
  - Distributional detail: inflation-target disinflation effects concentrate more on households in the bottom half of the income distribution (relative decline in P10 vs median), differing from broader RR shocks where upper-end dynamics dominate.

---

### Robustness and alternative specifications
- Robustness checks produced qualitatively similar results:
  - Excluding Volcker period (start in 1985Q1) or excluding recession quarters: results largely unchanged.
  - Longer lag structures (J = 8, I = 12) strengthen expenditure and consumption inequality effects.
  - Direct moving-average estimation and VAR approaches confirm contractionary shocks raise inequality; VAR shows particularly strong expenditure/consumption effects.
  - Controlling for technology shocks, oil shocks, and tax shocks contemporaneously increases estimated effects.
  - Controlling for household size and observables often strengthens estimates; restricting to full-time households increases significance of earnings, expenditure, and consumption inequality responses.

---

### Interpretation, policy implications, and recommended research (Conclusion, Section VI)
- Main conclusions
  - Contractionary monetary policy shocks increase income, labor-earnings, consumption, and total expenditures inequality.
  - Monetary policy shocks do not explain the long-run trend increase in income inequality since early 1980s, but they play a significant role in cyclical fluctuations and some longer-run deviations from trend (notably in consumption since 2000).
  - Transfers and fiscal/transfer policy buffer low-income households against some monetary shocks.

- Mechanisms emphasized
  - Labor-earnings channel: differential wage/earnings responses across distribution (upper rises, lower falls).
  - Household balance-sheet / wealth-transfer channel: unexpected changes in interest rates and inflation transfer wealth and affect expenditures, especially at the upper end.
  - Portfolio and financial-segmentation channels are plausible but appear quantitatively smaller in the CEX-based evidence.

- Policy and modeling implications
  - Need for heterogeneous-agent models with incomplete insurance and household balance-sheet heterogeneity to evaluate distributional consequences of monetary policy.
  - Costs of the zero lower bound may be larger when accounting for distributional effects; representative-agent models may understate welfare costs of zero-bound episodes.

- Recommended avenues for future research
  - Identify causal drivers of differential labor-income responses (capital–labor complementarities, age, supply elasticities).
  - Develop heterogeneous-agent monetary models that incorporate incomplete insurance and balance-sheet heterogeneity.
  - Use richer data that identify household net-worth to assess portfolio and financial-segmentation channels quantitatively.

*Italic: Source — IMF Working Paper _wp12199 (selected sections: 2.1; III; IV; V; VI; Figures and Tables summaries).*

### 2.1  The Consumer Expenditure Survey .........................................................9

### 2.1  The Consumer Expenditure Survey .........................................................9

### Survey and measures (Sections 2.1–2.3)
- 2.1  The Consumer Expenditure Survey .........................................................9
- 2.2  Measures of Inequality ...........................................................................11
- 2.3 Unconditional Properties of Inequality Measures ...................................13

### Effects of monetary policy shocks (Section III)
- III. Effects of Monetary Policy Shocks on Inequality .............................................................14
  - 3.1 The Identification of Monetary Policy Shocks .......................................14
  - 3.2 The Effects of Monetary Policy Shocks on Inequality ...........................17
  - 3.3 Why Does Inequality Increase After Contractionary Monetary Policy Shocks? ............................................................................................................20
  - 3.4 Distributional Mobility after Monetary Policy Shocks ...........................24
  - 3.5 How Important Is The Contribution of Monetary Policy Shocks to Inequality?........................................................................................................25

### Wealth redistribution and permanent policy changes (Sections IV–V)
- IV. Wealth Redistribution in Response to Monetary Policy Shocks .......................................26
- V. Permanent Changes in Monetary Policy .............................................................................28

### Conclusion (Section VI)
- VI. Conclusion .........................................................................................................................31

### Tables (listed with titles and page ranges)
- 1. Correlations and Volatilitiese of Inequality Measures .........................................................47
- 2. Correlations of Inequality Measures with Macroeconomic Variables ................................48
- 3. Decomposition of Income by Quintile .................................................................................49
- 4. Decomposition of Expenditures and Consumption by Quintile ..........................................50

### Figures (listed with titles and page numbers)
- 1. Inequality in Total Income, Labor Earnings, Expenditures and Consumption in the U.S. ..36
- 2. Monetary Policy Shocks ......................................................................................................37
- 3. Effects of Monetary Policy Shocks on Macroeconomic Variables .....................................38
- 4. Response of Economic Inequality to a Contractionary Monetary Policy Shock .................39
- 5. Distributional Effects of Contractionary Monetary Policy Shock by Percentiles ................40
- 6. Time-Varying Probabilities of Transitioning Between Consumption Quintiles..................41
- 7. Contribution of Monetary Policy Shocks to Forecast Error Variance of Inequality............42
- 8. The Contribution of Monetary Policy Shocks to Historical Variation in U.S. Inequality ...43
- 9. Income and Consumption Responses of High and Low Net-Worth Households ................44
- 10. Historical Estimates of the Federal Reserve's Target Rate of Inflation .............................45
- 11. Response of Inequality to Permanent Increases in the Inflation Target .............................46

### Appendix figures
- Appendix Figures 1. Robustness of Baseline Inequality Results to Sample and Lags ..........................................51
- Appendix Figures 2: Robustness of Baseline Inequality Results to Econometric Approach ................................52
- Appendix Figures 3. Robustness of Baseline Inequality Results to Controlling for Household Size, Observables, and Hours .................................................................................................................................53
- Appendix Figures 4. Robustness of Earnings Responses by Percentiles ..............................................................54
- Appendix Figures 5. Contribution of Monetary Policy Shocks to Variance of Macroeconomic Variables .........55

*Source: _wp12199 - 2.1  The Consumer Expenditure Survey (PDF chapter/section).*

### 6. Distributional Effects by Percentile of Inflation Target Increases .......................................56

### 6. Distributional Effects by Percentile of Inflation Target Increases .......................................56

### Introduction and motivation
- Recent public concern (e.g., Occupy Wall Street) has focused attention on the role of monetary policy and the Federal Reserve in inequality; quote from Daron Acemoglu and Simon Johnson: “Who Captured the Fed?” 3/29/2012.
- Contrasting views exist on whether monetary policy increases or decreases inequality (Austrian view, Galbraith 1998, mainstream views focused on technology, trade, institutions).
- The paper studies whether U.S. monetary policy contributed to historical changes in consumption and income inequality and through which channels, using household-level data from the Consumer Expenditures Survey (CEX) since 1980.

### Channels through which monetary policy can affect inequality
- Income composition channel:
  - Heterogeneity in primary income sources (labor vs. business and financial income).
  - If expansionary shocks raise profits more than wages, owners (typically wealthier) benefit disproportionately.
- Financial segmentation channel:
  - Agents with earlier or better access to financial markets gain from changes in money supply.
  - If financially active agents are higher-income on average, consumption inequality may rise.
- Portfolio channel:
  - If low-income households hold relatively more currency, inflationary policy transfers wealth from low- to high-income households (Erosa and Ventura 2002; Albanesi 2007).
- Savings redistribution channel:
  - Unexpected increases in interest rates or decreases in inflation benefit savers and hurt borrowers (Doepke and Schneider 2006), potentially increasing consumption inequality if savers are wealthier.
- Earnings heterogeneity channel:
  - Labor earnings respond differentially across the distribution (e.g., unemployment or wage rigidity disproportionately affecting low-income households—Galbraith; Heathcote, Perri and Violante 2010).

### Data and empirical approach
- Inequality measures constructed from Consumer Expenditures Survey (CEX), available quarterly since 1980.
  - CEX includes consumption purchases plus other expenditures such as mortgage payments, auto purchases, and education expenses.
  - CEX does not include the very upper end of the income distribution (i.e., the top 1 percent).
- Monetary policy shocks identified using Romer and Romer (2004) innovations to the Federal Funds rate uncorrelated with Fed Greenbook forecasts prior to FOMC meetings.
- Also consider more specific shocks: permanent changes in the Federal Reserve’s inflation target identified as in Coibion and Gorodnichenko (2011) and Ireland (2006).

### Main empirical findings
- A contractionary monetary policy shock raises observed inequality across households in:
  - income,
  - labor earnings,
  - expenditures, and
  - consumption.
- Robustness:
  - Results robust to dropping the Volcker disinflation period or all recession quarters.
  - Results similar when using VAR’s or controlling for other macro shocks.
  - Largely invariant to controlling for household size and household observables (age, education, hours worked).
- Magnitude and importance:
  - Forecast error variance decompositions suggest the contribution of monetary policy shocks to inequality is of the same order of magnitude as their contribution to other macro variables like GDP and inflation.
  - Monetary policy shocks can account for a surprising amount of historical cyclical changes in income and consumption inequality, particularly since the mid-1990s.

### Evidence on channels and distributional detail
- Percentile responses in labor earnings:
  - Contractionary monetary policy shocks are followed by higher earnings at the upper end of the distribution but lower earnings for those at the bottom.
  - Lower labor earnings at the low end appear even for households reporting full-time work—indicating heterogeneity in wage responses beyond unemployment.
  - High-earnings households (e.g., the 90th percentile) earn more labor income after contractionary shocks.
  - Long-run responses of labor earnings and consumption for each percentile align almost one-for-one, indicating a close link between earnings and consumption inequality responses.
- Income composition evidence:
  - Aggregate labor earnings respond little on average to monetary shocks.
  - Aggregate financial income rises sharply while business income declines after contractionary shocks.
  - CBO finding: the top 1 percent received approximately 30 percent of their income from financial income—suggesting total income for the top 1 percent likely rises more than for most CEX households after contractionary shocks.
  - Because CEX excludes the top 1 percent, baseline results on income inequality are likely a lower bound.
- Role of transfers:
  - Lower labor earnings at the 10th percentile (and to a lesser extent the 25th percentile) after contractionary shocks are offset in total income by transfers.
  - Transfers rise (with delay) after contractionary monetary shocks, insulating incomes of many bottom-distribution households.
  - Thus, dynamics of total income inequality primarily reflect fluctuations at the upper end of the distribution, dominated by their labor earnings.
- Wealth-related channels (limitations and suggestive evidence):
  - CEX lacks reliable household wealth measures; cannot directly quantify portfolio channel or financial market segmentation channel.
  - Baseline results showing inequality rising after contractionary shocks suggest these wealth-based channels, if present, are small relative to others.
  - Using Doepke and Schneider (2006) characterization of high vs. low net-worth households:
    - High net-worth defined as older, homeowners, receiving financial income.
    - Low net-worth defined as younger, fixed-rate mortgage holders, receiving no financial income.
    - Consumption and, to a lesser extent, total expenditures rise significantly more for high net-worth households than low net-worth households after contractionary shocks—suggestive of wealth-transfer effects.

### Sensitivity to the nature of monetary policy innovations
- Romer and Romer (2004) shocks aggregate various types of Fed innovations (preferences, objectives, political constraints); these can be transitory or persistent.
- Inflation-target shocks:
  - Permanent decreases in the inflation target (as identified by Coibion and Gorodnichenko (2011) and Ireland (2006)) systematically increase income and consumption inequality.
  - Forecast error variance decompositions for inflation-target shocks produce contributions in line with baseline Romer and Romer shocks.
  - Inflation-target shocks imply larger historical contributions of monetary policy to consumption and expenditure inequality, particularly in the early 1980s (Volcker disinflation accounts for large, persistent increases in consumption and expenditure inequality).

### Implications and relevance
- Monetary policy may have played a more significant role in historical inequality patterns in the U.S. than previously recognized.
- Findings provide stylized facts on conditional responses of income, earnings, and consumption across households to monetary policy shocks—useful for calibrating heterogeneous-agent macro models and assessing monetary transmission.
- Transfers and fiscal/transfer policy play an important buffering role for low-income households against monetary shocks.
- The connection between inequality and macroeconomic stability (credit booms, financial crises) highlights potential feedbacks from inequality to macro outcomes.

*Source: _wp12199 - 6. Distributional Effects by Percentile of Inflation Target Increases .......................................56*

### section 5 considers the implications of changes in the inflation target. Section 6 concludes.

### _wp12199 - section 5 considers the implications of changes in the inflation target. Section 6 concludes.

### Measuring Inequality — data source and sample construction
- Data source: Consumer Expenditure Survey (CEX), Interview Survey component only.
- Coverage: provides information on up to 95 percent of the typical household's consumption expenditures.
- Frequency and panel design:
  - Monthly rotating panel available on a continuous basis since 1980.
  - About 1,500-2,500 households surveyed in any given month.
  - Each household interviewed once per quarter, for at most five consecutive quarters (first interview used for pre-sampling and not available for analysis).
  - Reference period for expenditures in each interview covers the three months prior to the interview month; household time aggregated to quarterly expenditures for analysis.
- Measurement and aggregation:
  - Non-durable consumption includes food, alcohol and tobacco, and gasoline and other fuel.
  - Service consumption includes household utilities, household operations, service charges, recreational services, public transportation, and personal care services.
  - Household consumption = non-durables + services + expenditures on durable goods (e.g., furniture and furnishing, jewelry and watches, recreational goods, personal care durables).
  - Broad measure (total expenditures) adds mortgage and rent payments, health expenditures, education spending and other expenses to household consumption.
- Data cleaning and adjustments:
  - Corrected sample breaks for food at home (1982Q1-1988Q1), personal care services (2001Q2), and occupation expenditures (2001Q2).
  - Dropped interviews with more or less than three monthly observations; households reporting zero food or total expenditures; observations with negative expenditures where inappropriate.
  - Summed expenditures reported in the same month but in different interviews (BLS recommendation).
  - These procedures eliminate about 7 percent of observations in the initial sample.
- Income data specifics:
  - Income asked in the first and last interview (interviews 2 and 5 in CEX terminology); financial data only asked in the last interview.
  - Reference period for income flows covers the twelve months before the interview.
  - All nominal variables deflated using the CPI-U.
  - Survey sample weights used to ensure comparability across sub-samples and with aggregate studies.

### Measures of inequality — definitions and implementation
- Three primary measures used for each form of inequality:
  - Gini coefficient of levels (value between 0 and 1).
  - Cross-sectional standard deviation of log levels (requires dropping zero observations).
  - Difference between the 90th percentile and the 10th percentile of log levels (requires dropping zero observations).
- Income measures:
  - Two forms: labor earnings inequality (wages and hours) and total income inequality (labor earnings + financial income + business income + transfers).
  - Income in the CEX is asked only in interviews 2 and 5; therefore income-based inequality measures use only individuals reporting income in each survey quarter (a subset of the survey population).
  - Baseline focus on pre-tax total income measures; after-tax measures also presented as robustness checks. All labor income is pre-tax.
  - Robustness checks restrict the sample to full-time working individuals in some specifications.
  - Tax burdens for after-tax measures computed following Kueng (2012) using the TAXSIM calculator of the NBER.
- Consumption measures:
  - Narrow measure (referred to as consumption inequality) follows Parker (1999): non-durables + services + some durables (household appliances, entertainment goods like televisions, furniture); excludes large durables (houses, cars).
  - Broad measure (referred to as total expenditures) includes narrow consumption plus mortgage payments, car purchases, medical supplies and services, tuition and books, and other items.
  - Consumption and expenditure data measured every period; inequality measures use the entire survey population each period (subject to survey caveats).
- Baseline measures are raw (do not control for household characteristics). Controls for household observables are included in robustness checks to avoid masking channels (e.g., redistribution between borrowers and savers related to age).

### Empirical correlations, trends, and unconditional properties
- Correlations among income inequality measures (1980Q1 to 2008Q4):
  - Different measures of income inequality are highly correlated with one another, with correlations of 0.89 or above over the entire sample from 1980Q1 to 2008Q4.
- Earnings measures:
  - Correlations between different measures of earnings inequality are generally lower, particularly for the Gini coefficient.
  - Correlation between the standard deviation and the 90th-10th percentile measures is 0.85 for earnings.
- Expenditure and consumption correlations:
  - Correlations across different measures of expenditure inequality range from 0.75 to 0.89.
  - Correlations across different consumption measures range from 0.80 down to 0.45.
- Cross-form correlations (selected examples):
  - Correlation between income and earnings inequality using Gini coefficients: 0.82.
  - Correlation between income and earnings inequality using the standard deviation: 0.22.
  - Correlation between income and earnings inequality using the 90th-10th percentiles: 0.12.
- Volatility patterns:
  - Income and earnings inequality measures have approximately the same volatility.
  - Volatility in expenditure inequality tends to be higher than that of consumption inequality.
- Time series patterns (Figure 1 observations):
  - Total income inequality measures trend up over time.
  - Labor earnings inequality trends up when measured by Gini but not when measured by cross-sectional standard deviation nor 90th-10th percentile differential.
  - Sharp increase in all forms of inequality in the early 1980s.
  - Income inequality rises over the 1990s; little such movement for consumption or expenditure inequality measures.
  - Noticeable decline in expenditure inequality over the 2000s despite no such decrease in income inequality.
- Unconditional correlations with macro variables (Table 2, HP-filtered series):
  - Correlations of different forms of inequality with the inflation rate are very small and somewhat negative.
  - Similar small/negative results obtain with interest rate correlations.
  - Labor earnings inequality is weakly positively correlated with the unemployment rate and negatively with inflation.
  - Expenditure and consumption inequality are more strongly negatively correlated with the unemployment rate.
  - Overall, unconditional correlations do not point to very strong links between business cycles and inequality patterns.

### Effects of monetary policy shocks on inequality — identification and data extension
- Identification strategy:
  - Follows Romer and Romer (2004, RR) to identify policy innovations orthogonal to the Fed’s information set (Greenbook forecasts).
  - RR construct a historical measure of changes in the target Federal Funds rate (FFR) at each FOMC meeting from 1969 until 1996.
  - Using real-time forecasts in the Greenbooks (denoted F), RR estimate policy-change residuals that are orthogonal to Greenbook forecasts of real output growth, GDP deflator inflation, and the current quarter’s average unemployment rate; the estimated residuals are defined as monetary policy shocks.
- Extension of RR shocks in this study:
  - RR dataset extended until December 2008 by:
    - Incorporating more recent changes in the target FFR decided at regular FOMC meetings.
    - Extending Greenbook forecasts until December 2006 (the most recent period through which the Federal Reserve has released them).
    - Using consensus forecasts from the Blue Chip Economic Indicators in place of Greenbook forecasts for FOMC meetings in 2007 and (implicitly) later meetings up to 2008 where Greenbook forecasts are not available.

### Themes signaled for subsequent analysis (as presented)
- Baseline results quantifying effects of monetary policy shocks on different forms of inequality in the U.S.
- Robustness checks on identification and inequality measures (including after-tax income, sample restrictions, and controls for household characteristics).
- Analysis of how monetary policy shocks affect different components of the distributions and mobility within distributions.
- Assessment of the economic importance of monetary policy for inequality dynamics.

*Italic: Source — _wp12199 - section 5 considers the implications of changes in the inflation target. Section 6 concludes.*

### 2008.  The  dataset  therefore  extends  until  the  zero-bound  on  interest  rates  became  binding  in

### 2008. The dataset therefore extends until the zero-bound on interest rates became binding in December 2008.

### Construction of the monetary policy shock series
- Estimated the same specification as RR on data since January 1969 to obtain a sequence of monetary policy shocks at the frequency of FOMC meetings.
- Constructed a quarterly measure of monetary policy shocks by averaging the orthogonalized innovations to the FFR from each meeting within a quarter.
- Resulting shock series plotted starting in 1980Q1; shocks particularly large and volatile in the early 1980s (Volcker disinflation) and distinctly positive in 1994-1995 (“pre-emptive strike”) and 2005-2006; the 2000–2004 period suggests more expansionary policy, consistent with Taylor (2007).
- Sample endpoints and lag choices:
  - Full sample: 1969:Q1 to 2008Q4.
  - Restricted sample: since 1980Q1.
  - Over the entire sample set J = 8 and I = 12 as in RR.
  - Over the more restricted sample use J = 4 and I = 8.
  - All estimates are at the quarterly frequency.
- Standard errors: Newey-West (1987). Confidence intervals for impulse responses constructed by bootstrap drawing repetitively from the estimated distribution of coefficients of equation (2).

### Macroeconomic and financial responses to contractionary monetary policy shocks
- Over 1969:Q1 to 2008Q4, contractionary monetary policy shocks:
  - Lower real GDP, consumption and investment.
  - Raise unemployment.
  - Both short-term and long-term interest rates rise immediately.
  - Inflation declines after a two-year lag.
- Impulse responses point to smaller real effects of monetary policy shocks since 1980, in line with Boivin and Giannoni (2006), though differences are not statistically significant.
- Asset price responses:
  - Real stock prices (Dow Jones Industrial Average deflated by the GDP deflator) decline on impact but eventually rise (as in Bernanke and Kuttner 2005).
  - Real housing prices (Case-Shiller price index deflated by the GDP deflator) decline gradually after one to two years.
- Distributional implication: differential asset exposures imply monetary policy can raise wealth for households with more stock holdings and lower wealth for households whose wealth is primarily in housing.

### Income components and channels
- Responses of income components to contractionary shocks:
  - Real wages: not statistically different from zero over either sample.
  - Business income: drops rapidly and significantly.
  - Financial income: rises significantly and persistently over both the entire period and the period since 1980Q1.
  - Total income from transfers: drops over the first year, then rises approximately two years after the shock (temporarily over the entire sample, persistently since 1980).
- Interpretation: heterogeneity in income sources leads to distributional consequences—contractionary monetary policy tends to raise incomes for those with substantial financial income and lower incomes for business owners.

### Effects of monetary policy shocks on inequality
- Estimated equation (2) for inequality measures of total income, labor earnings, total expenditure and consumption using three measures each:
  - Cross-sectional standard deviation (of logged values).
  - Gini coefficient.
  - (Log) 90th percentile minus (log) 10th percentile.
- Lag structure used for inequality measures: J = 4 and I = 8 quarters.
- Main findings (1980Q1 to 2008Q4; one standard deviation confidence intervals):
  - Income and labor earnings inequality: statistically significant effects; higher long-run levels of inequality after contractionary shocks (statistical significance varies by measure).
  - Consumption and expenditure inequality: especially strong evidence—each measure points to a statistically significant and highly persistent increase in inequality after a contractionary shock.
  - Point estimates for expenditures are consistently larger than for other forms of inequality, indicating disproportionately large effects on expenditure inequality.
- Conclusion: across all forms of inequality and measures, contractionary monetary policy shocks are associated with higher levels of economic inequality.

### Robustness checks and alternative empirical procedures
- Lag length sensitivity:
  - Using longer lags (J = 8 and I = 12) yields qualitatively unchanged effects for total income and labor earnings inequality; effects for expenditure and consumption inequality are strengthened and become statistically significant at longer horizons.
- Sensitivity to Volcker period:
  - Results starting in 1985Q1 are qualitatively unchanged, indicating Volcker disinflation does not drive results.
- Excluding NBER-dated recession quarters:
  - Results are qualitatively unchanged; estimated effects on consumption inequality less precise, effects on total income and labor earnings inequality are larger.
- Direct moving-average estimation (equation (3) with N = 20):
  - Point estimates for total income and labor earnings inequality similar but with larger standard errors (marginally significant).
  - Effects for consumption and expenditure inequality more pronounced.
- Controlling for other macro shocks (equation (4)): technology shocks (Gali 1999), oil supply shocks (Kilian 2009), tax shocks (Romer and Romer 2010) included contemporaneously and with two lags:
  - Estimated effects of contractionary monetary policy on inequality are larger than baseline.
- VAR approach:
  - VAR including real GDP, unemployment rate, price level, cumulative sum of monetary policy innovations, and a measure of inequality; identification via Choleski decomposition; 4 lags; 1980Q1–2008Q4.
  - Results point to contractionary shocks raising all measures of inequality; effects on total expenditure and consumption inequality noticeably stronger than baseline.
- Household characteristic adjustments:
  - OECD equivalence scale (1.0 head, 0.7 each additional adult 17+, 0.5 each child): effects stronger than baseline.
  - Controlling for observables (age quartic, number of adults, number of children, race, education, etc.) by regressing logged household variables on observables and defining inequality as cross-sectional standard deviation of residuals: estimates qualitatively unchanged and often stronger, except consumption inequality.
- Controlling for labor supply responses:
  - Restricted to households with either both members working 35–45 hours/week all year or one member 35–45 and the other <5 (single-adult households ~40 hours).
  - Constructed inequality measures controlling for household size and observables for full-time households: responses of earnings, expenditure and consumption inequality larger and more statistically significant than baseline; total income inequality slightly smaller but still points to higher inequality after contractionary shocks.

### Summary conclusion
- The increase in economic inequality (income and consumption) after contractionary monetary policy shocks is robust:
  - Holds across different inequality measures (Gini, percentile differentials, standard deviations).
  - Not driven exclusively by the Volcker disinflation or recessions.
  - Recoverable using multiple empirical methods (RR baseline, moving-average, augmented regressions, VAR).
  - Persists when controlling for household characteristics and labor supply decisions.

*Source: _wp12199 - 2008. The dataset therefore extends until the zero-bound on interest rates became binding in December 2008.*

### 3.3 Why Does Inequality Increase After Contractionary Monetary Policy Shocks?

### 3.3 Why Does Inequality Increase After Contractionary Monetary Policy Shocks?

### Mechanisms examined
- The analysis focuses on how contractionary monetary policy shocks affect consumption and income at different percentiles (90th, 75th, 50th, 25th, 10th) of the distributions.
- Construction of percentile series:
  - For income and labor earnings: percentiles constructed each quarter from the distribution of households reporting income/earnings that quarter; composition changes across quarters because income/earnings over the last twelve months are collected only in the first and fourth quarter of household participation.
  - For consumption and expenditures: households are ranked each quarter by consumption or expenditures; those near each percentile are tracked to construct percent changes controlling for composition effects.
- Responses of each percentile to contractionary monetary policy shocks are recovered from estimating equation (2) for percentile changes.

### Key empirical findings on earnings and income percentiles
- Labor earnings:
  - After about two years, wage earnings for those in the upper end of the salary income distribution rise while those at the lower end see their earnings fall.
  - The 75th percentile response lies between the 90th and 50th; the 25th lies between the 10th and 50th, indicating heterogeneous effects across the distribution.
  - Responses are highly symmetric: the median response is essentially zero, while earnings at the 90th percentile rise by almost the same percentage as the decline in earnings of the 10th percentile.
  - Restricting the sample to households where either both adults report full-time work or one reports full-time work and the other reports no working hours lowers the long-run effects of monetary policy shocks for each percentile by approximately 1 percentage point such that real wages weakly fall for all percentiles but does not affect their relative magnitudes.
  - Controlling for household observables (and for observables plus hours worked) reduces quantitative differences but leaves the relative ordering of earnings responses by percentile unchanged. The extent to which differential earnings responses reflect education, age, race, other observables, or labor supply is unclear.

- Total incomes:
  - Effects on income inequality are present but smaller than for wage income.
  - The 90th percentile of the income distribution shows the highest response starting two years after the shock; the 10th percentile is lowest.
  - The difference between the 90th and 10th percentiles in total income is about half of that implied by earnings differentials.
  - The 25th percentile of the income distribution rises relative to both the 50th and 75th percentiles, so monetary policy shocks have less monotonic effects on total income than on wage incomes.

- Decomposition of total income (Table 3, by quintile measured by consumption of nondurables and services):
  - In the 2000s:
    - Labor income accounted for over 80 percent of total income for the highest quintile.
    - Labor income accounted for 74 percent for the middle quintile.
    - Labor income accounted for only slightly over 40 percent for the bottom quintile.
  - The largest contributor to total income for the bottom quintile is the “other income” category (approximately 50 percent), which includes unemployment insurance, Social Security and pension payments, welfare, worker’s compensation, and other transfer programs.
  - For the second quintile, other income accounts for approximately 25 percent of total income.
  - For the top 2 quintiles, other income is less than 10 percent of total income.
  - Business income share: rises from 2 percent of total income for the bottom quintile to 5-9 percent for the top quintile.
  - Financial income share: falls from 9-12 percent at the bottom quintile to approximately 7 percent for the top quintile.

- Interpretation of income responses:
  - Because lower-income households rely less on salary income and more on transfer and financial income, the decline in their labor earnings is partially mitigated in total income by other income sources and by rises in financial income after monetary policy shocks.
  - Therefore, while labor earnings respond symmetrically (upper rises, lower falls), total income responses are asymmetric: increases in inequality primarily reflect changing incomes at the upper end, while incomes at the bottom differ little from the median due to compensating transfer income.
  - The results likely provide a lower bound on the effects of monetary policy shocks on income inequality because the top 1 percent (not captured in the CEX) derive a larger share of income from financial and business income and could experience larger net gains.

### Consumption and expenditures by percentile
- Consumption:
  - Responses of consumption by percentile closely mirror those for labor earnings, but differential behavior between high-consumption and low-consumption households appears much earlier than for labor earnings.
  - Long-run responses of consumption percentiles are almost identical to those for earnings:
    - Labor income at the 90th percentile rises by 4 percent; consumption at the 90th percentile rises by 4 percent.
    - Labor income at the 10th percentile falls by 5 percent; consumption at the 10th percentile falls 6 percent.
    - Responses for the 25th, 50th and 75th percentiles of salary and consumption are also remarkably close to one another.

- Expenditures:
  - Expenditures by percentile show substantial heterogeneity in responses to monetary shocks.
  - The largest observed effect is a very large increase in expenditures by those at the 90th percentile of the expenditure distribution; this increase dwarfs changes in income or earnings and likely explains the significant and persistent rise in expenditure inequality after monetary policy shocks.
  - Investigation of expenditure composition (Table 4):
    - Upper-end households consume relatively more durables and devote more spending to interest-sensitive expenditures (mortgage payments, auto purchases), but the differences across quintiles are small.
    - The ratio of spending on expenditures to consumption of nondurables and services is not rising across quintiles.
    - Impulse responses of average spending by category across quintiles show no evidence that upper-end households increased their expenditures more than others after contractionary shocks.
  - Thus, the greater response of expenditures for the 90th percentile cannot be explained by composition of spending across quintiles.

### Distributional mobility after monetary policy shocks (Section 3.4 summary)
- Mobility trends (1980Q1–2008Q4):
  - Time-varying quarterly transition probabilities are constructed for each quintile of the consumption distribution (fraction of consumers within each quintile who end up in another quintile next quarter).
  - Mobility has declined over time for each quintile other than the bottom one.
  - Example: middle quintile frequency of remaining in the same quintile from one quarter to the next went from approximately 35 percent in 1980 to nearly 45 percent in 2008.
  - Mobility at the bottom end is limited: the frequency at which people in the lowest quintile move to the middle quintile or above has been approximately 5 percent each quarter over this period.

- Effects of monetary policy shocks on transition frequencies:
  - Equation (2) in levels is estimated for each series using squared monetary policy innovations to identify whether positive or negative shocks increase movements across the distribution.
  - Two findings:
    - Little persistence in the effects: after two years, almost none of the estimates for transition probabilities are different from zero.
    - Monetary policy shocks cause some increased movement within the distribution in the short run: the frequency of households remaining within the same quintile declines for all quintiles.
  - Implication: significant movement within the distribution in the quarters immediately following monetary policy shocks likely contributes to the volatility of impulse responses for percentiles of total income and labor earnings in the first two years. As increased mobility fades after two years, percentile responses converge to more stable outcomes.
  - Consistent with this, percentile responses for expenditures and consumption (which control for composition effects) are much more stable through the first two years than those for earnings and income.

*Source: _wp12199 - 3.3 Why Does Inequality Increase After Contractionary Monetary Policy Shocks?*

### 3.5 How Important Is The Contribution of Monetary Policy Shocks to Inequality?

### 3.5 How Important Is The Contribution of Monetary Policy Shocks to Inequality?

### Variance decompositions: share of inequality variance accounted for by monetary policy shocks
- The fraction of the variance in inequality at different horizons accounted for by monetary policy shocks is recovered from estimates of equation (2).
- Results (using Gini coefficients, cross-sectional standard deviations and 90th–10th percentiles for inequality) are presented in Figure 7 for total income, labor earnings, total expenditures, and consumption inequality.
- One-standard confidence intervals and median estimates are reported.
- Most estimates lie between 10 percent and 20 percent.
- Appendix Figure 5 shows equivalent variance decompositions for macroeconomic variables over the same period; with the exception of short-term interest rates and financial income, the contribution of monetary policy shocks to the variance of these variables is also in the 10-20 percent range for most forecasting horizons.
- Conclusion: the contribution of monetary policy shocks to inequality dynamics is in line with its effects on other macroeconomic variables.

### Historical contribution: predicted vs actual dynamics since 1980
- Method:
  - Use predicted changes in income, salary, expenditure and consumption inequality due to monetary policy shocks from estimates of equation (2).
  - Accumulate predicted changes over time and compare to actual evolution of inequality measures.
  - Normalize both predicted and actual series by the trend growth rate of each historical inequality measure to avoid assigning trend growth from estimated constants in equation (2) as a contribution of monetary policy shocks.
  - Average both actual and predicted variables over the previous and subsequent quarter values to downplay very high-frequency variation and emphasize business cycle variation. Detrending and averaging are done after estimation.
- Results (Figure 8; cross-sectional standard deviation measures; other measures qualitatively similar):
  - Monetary policy shocks account for very little of the variation in income, earnings and expenditure inequality until the mid-1990s, with the exception of some higher-frequency variation in income inequality in the very early 1980s.
  - Since the mid-1990s monetary policy shocks can account for some gradual changes in income and earnings inequality, including:
    - gradual reversion to trend until 2004;
    - increase in income and earnings inequality in 2005-2006 when monetary policy shocks were systematically contractionary;
    - decline in income and earnings inequality into 2008 while monetary policy shocks were distinctly expansionary.
  - Consumption inequality:
    - monetary policy shocks account for many short-lived deviations from trend in the 1980s;
    - monetary policy shocks can almost fully account for the behavior of consumption inequality since 2000.
  - Expenditure inequality:
    - monetary policy shocks account for very little of the variation over the sample, except for the decline since 2005.
    - For the rest of the sample there is very little correlation between actual movements in expenditure inequality and predicted changes due to monetary policy shocks.

### Wealth redistribution and other redistributive channels
- Three additional channels commonly discussed for distributional effects of monetary policy:
  1. Portfolio channel: differential asset holdings and protection against inflation surprises can reallocate wealth across agents (Erosa and Ventura (2002); Albanesi (2007)).
  2. Segmented financial markets: agents more connected to financial markets may be affected by changes in money supply earlier, redistributing wealth from connected to unconnected agents (Ledoit (2009); Williamson (2009)). If market participants have higher income/consumption on average, contractionary shocks should lower consumption inequality.
  3. Saver–borrower channel: contractionary shocks transfer from borrowers (low net-worth) to savers (high net-worth), potentially increasing expenditures of borrowers (Doepke and Schneider (2006)).
- Data limitations in the CEX:
  - No reliable data on household cash holdings.
  - No identification of households most connected to financial markets.
- Empirical test of saver–borrower channel using restricted samples:
  - Low net-worth definition: aged 30-40 year-old white households with a male head, no financial income, positive mortgage payments.
  - High net-worth definition: aged 55-65 year-old white households with a male head, positive financial income, no mortgage payments.
  - Construct mean (log) income and expenditures and subcategories; compute impulse responses via equation (2).
- Findings (Figure 9):
  - Little evidence that labor earnings or total income responses differ across the two groups.
  - High net-worth households display increases in consumption and total expenditures relative to low net-worth households, consistent with redistribution of nominal wealth.
  - Because high net-worth households are on average near the upper end of income and consumption distributions, this channel reinforces the positive effects of contractionary monetary policy shocks on income and consumption inequality.

### Permanent changes in monetary policy: inflation target shocks
- Context on Romer and Romer (RR) shock measure:
  - RR uses changes in the FFR target decided at FOMC meetings, purged of anticipatory effects via real-time staff forecasts.
  - RR shocks reflect a composite of sources (evolution of Fed operating procedures, policymakers’ beliefs, objectives, political pressures, responses to other factors).
  - Different sources (permanent vs transitory) may yield very different economic responses; permanent changes could have more pronounced effects.
- Two alternative measures of the Fed’s target rate of inflation considered:
  1. Coibion and Gorodnichenko (CG) approach:
     - Postulate a time-varying Taylor-rule reaction function (equation (3)) where the central bank moves interest rates in response to perception of the natural rate of interest, deviations of expected inflation from a time-varying target, deviations of expected output growth from its target, and the output gap.
     - Time-varying coefficients follow random walk processes. Estimate coefficients using real-time Greenbook forecasts (and Blue Chip Economic Indicators for 2007–2008) for 1969–2008 at FOMC meeting frequency.
  2. Ireland (2006) approach:
     - Estimate a small-scale New Keynesian model with a Taylor rule and time-varying inflation target via maximum-likelihood using final data on output, prices, and interest rates; recover implied time path of Fed’s inflation target.
- Both measures (Figure 10):
  - Point toward rising inflation targets over the 1970s, peaking at approximately 8 percent.
  - Capture rapid declines in target inflation in the early 1980s (Volcker disinflation) and prolonged subsequent decline over the 1990s and 2000s, reaching 2 percent in 2005 in both measures.
  - Qualitative differences:
    - Ireland: rapid increase starting around 1973 to 8 percent in late 1974, declining to 6 percent in 1975; begins decline in 1981.
    - CG: only gradual increase in early 1970s; continues to rise until end-1982, then drops much more abruptly (3 percentage points over a few months).
- Effects of 1 percent point decrease in the inflation target (Figure 11; cross-sectional standard deviations):
  - Both inflation-target measures yield statistically significant responses: disinflations raise long-run levels of both income and consumption inequality.
  - Estimated effects quantitatively similar across income and consumption measures: inequality of each type rises approximately 2 percent.
  - Short-run and long-run dynamics similar for consumption and expenditure inequality across the two measures; short-run dynamics in income and earnings inequality differ more between measures.
- Forecast-error variance contributions of inflation-target shocks:
  - Around 10-20 percent at different horizons, in line with baseline RR-shock estimates, and even higher for consumption.
  - Suggests inflation-target shocks by themselves can account for a significant proportion of variation in inequality measures.
- Historical contribution of inflation-target shocks:
  - For consumption and expenditure inequality, both measures point to very important contributions:
    - The early-1980s disinflation generates a very large increase in these inequality forms and can account for persistently high levels of inequality (relative to trend) in the mid-1980s and 1990s.
  - For income and earnings inequality:
    - Ireland’s measure can fully account for the large increase in the early 1980s and its persistence.
    - CG measure accounts for only a fraction of the variation, likely because its dramatic decline in target inflation occurs too late to explain the early-1980s increase in income inequality.
- Distributional detail (Appendix Figure 6 referenced):
  - Consumption: both measures indicate the low end fares no worse than the median; the high end fares better.
  - Expenditures: both measures yield result that the bottom falls further below the median, but deliver conflicting results for the upper end.
  - Total income and labor earnings: both measures indicate the rise in inequality after contractionary shocks is driven by a relative decline in the 10th percentile relative to the median; the 90th percentiles fare little better than the median.
  - Implication: unlike the broader RR shocks, inflation-target disinflation effects on income concentrate disproportionately on households in the bottom half of the income distribution.

### Interpretation and caveats
- The positive finding that persistent disinflations raise income inequality in the U.S. contrasts with cross-country literature documenting a positive relationship between inequality and inflation (e.g., Easterly and Fischer 2001; Albanesi 2007; Romer and Romer 1998).
- Potential explanations for divergence:
  - Cross-country correlations may reflect confounding factors (e.g., more unequal countries adopting populist higher-inflation policies).
  - This analysis focuses on the conditional response of inequality to disinflationary policies in the U.S. since 1980, mitigating some confounding concerns.
  - Cross-country differences in financial development, labor market institutions, transfer systems, skill distributions, and other factors will affect the relative strength of portfolio, saver–borrower, and labor-earnings channels.
- Final caution: be wary of drawing broad generalizations from the U.S.-focused, post-1980 results given likely cross-country heterogeneity in channels and institutional settings.

*Source: IMF Working Paper section 3.5 (How Important Is The Contribution of Monetary Policy Shocks to Inequality?)*

### conclusions   about   whether   higher   inflation   necessarily   increases   or   decreases   economic

### _wp12199 - conclusions   about   whether   higher   inflation   necessarily   increases   or   decreases   economic

### Summary of main conclusion
- Contractionary monetary policy shocks appear to have significant long-run effects on inequality, leading to higher levels of income, labor earnings, consumption and total expenditures inequality across households.
- Monetary policy shocks cannot account for the trend increase in income inequality since the early 1980s, but they have played a significant role in cyclical fluctuations in inequality and some longer-run movements around the trends.
- Consumption inequality and expenditure inequality (after changes in the target inflation rate) are particularly affected and are likely the most relevant metrics from a policy point of view.

### Empirical findings
- Contractionary monetary policy shocks increase:
  - income inequality;
  - labor earnings inequality;
  - consumption inequality;
  - total expenditures inequality.
- The dynamic responses point to two empirically important channels (see Channels subsection).
- The response of consumption and expenditures by high net-worth households is larger than that of low net-worth households in the data, consistent with redistribution effects documented in Doepke and Schneider (2006).
- Data limitations of the CEX make identification of high and low net-worth households tentative.
- Counterfactual predictions from portfolio and financial market segmentation redistribution channels suggest these channels, if present, are quantitatively small relative to the other channels linking economic inequality and monetary policy in the U.S.

### Channels linking monetary policy shocks to inequality
- Labor earnings channel:
  - Labor income rises at the upper end of the distribution and falls at the lower end.
  - Since monetary policy shocks operate primarily through interest rates, this suggests modeling the complementarity/substitutability of capital with different forms of labor (such as skilled or unskilled) could help account for these patterns.
  - Heterogeneity could also reflect different age profiles and associated labor supply elasticities or complementarities between labor and consumption.
- Household balance-sheet / wealth-transfer channel:
  - Disproportionate increase in expenditures for those at the upper end of the expenditure distribution relative to income changes suggests significant wealth transfers via unexpected changes in interest rates and inflation.
  - Household balance sheets may play an important link in the monetary transmission mechanism.

### Implications for modeling and policy
- Results point to a need for models with heterogeneity across households suitable for monetary policy analysis.
  - Heterogeneous agent models with incomplete insurance markets have become increasingly common, but their implications for monetary policy remain underdeveloped.
- Sensitivity of inequality measures to monetary policy actions implies larger costs of the zero-bound on interest rates than commonly identified in representative agent models.
  - Nominal interest rates hitting the zero-bound when the central bank’s systematic response calls for negative rates is conceptually similar to a prolonged period of contractionary monetary policy shocks.
  - Given such shocks appear to increase income and consumption inequality, standard representative agent models may significantly understate the welfare costs of zero-bound episodes.

### Recommendations and avenues for future research
- Identify the causal link(s) driving differential labor-income responses across the income distribution (capital–labor complementarity, age profiles, labor supply elasticities, or consumption–labor complementarities).
- Develop heterogeneous-agent monetary models that explicitly incorporate incomplete insurance markets and household balance-sheet heterogeneity to evaluate distributional consequences of monetary policy.
- Further investigate the quantitative importance of portfolio and financial market segmentation redistribution channels using data that better identify household net-worth.

*Italicized source: conclusions section of IMF working paper _wp12199 (conclusions about whether higher inflation necessarily increases or decreases economic inequality).*

### 4. See section 3.2 for

### _wp12199 - 4. See section 3.2 for

### Distributional effects of a 1 percentage point (100 b.p.) contractionary monetary policy shock (Figure 5; section 3.3)
- Data: CEX, 1980Q1-2008Q4.
- Percentiles analyzed: P10, P25, P50, P75, P90.
- Outcomes plotted: income, salary (earnings), expenditure, consumption distributions.
- Visual responses (impulse responses by percentile) indicate heterogeneous percentile-level effects across:
  - Income Inequality (Standard Deviation)
  - Gini Coefficient
  - 90th-10th Percentiles
  - Earnings Inequality
  - Expenditure Inequality
  - Consumption Inequality
- See section 3.3 for details.

### Time-varying probabilities of transitioning between consumption quintiles (Figure 6; section 3.4)
- Left column: frequency of households staying in same consumption quintile (non-durables and services) from one quarter to the next.
- Right column: response of stay-in-quintile probabilities to a squared monetary policy shock; one standard deviation confidence intervals shown as dotted lines.
- Grey shaded areas denote U.S. recessions.
- Quintiles shown: Q1 => Q1, Q2 => Q2, Q3 => Q3, Q4 => Q4, Q5 => Q5.
- Time series axis shown from 1980 to 2000.
- See section 3.4 for details.

### Contribution of monetary policy shocks to forecast error variance of inequality (Figure 7; section 3.5)
- Plots contribution of monetary policy shocks to forecast error variance of economic inequality at different time horizons (quarterly, x-axis).
- Visual details:
  - Dotted blue lines: 68 percent confidence intervals.
  - Solid line: median estimate.
- Inequality measures included:
  - Income Inequality (Standard Deviation)
  - Gini Coefficient
  - 90th-10th
  - Earnings Inequality
  - Expenditure Inequality
  - Consumption Inequality
- See section 3.5 for details.

### Predicted vs actual historical variation in U.S. inequality attributable to monetary policy shocks (Figure 8; section 3.5)
- Plots predicted path of inequality due only to monetary policy shocks (solid lines) versus actual path of inequality measures (dotted lines).
- Inequality measured using cross-sectional standard deviation; both actual and predicted series normalized by trend growth and averaged over previous and subsequent quarter values.
- Outcomes displayed over 1985–2005 for:
  - Income Inequality
  - Earnings Inequality
  - Expenditure Inequality
  - Consumption Inequality
- Grey shaded areas denote recessions.
- See section 3.5 for details.

### Income and consumption responses by net-worth groups (Figure 9; section 4)
- Shock: 1 percentage point (100 b.p.) contractionary monetary policy shock.
- Groups:
  - Low net-worth households (solid line; grey shaded area = one standard deviation confidence interval).
  - High net-worth households (solid line with circles; dashed lines = confidence interval).
- Outcomes plotted (impulse responses):
  - Income (horizon labels include 5, 10, 15)
  - Earnings
  - Expenditures
  - Consumption
- See section 4 for definitions of low and high net-worth households and details.

### Historical estimates of the Federal Reserve’s target rate of inflation (Figure 10; section 5)
- Plots estimated inflation target rate of the U.S. Federal Reserve:
  - Ireland (2006, solid line)
  - Coibion and Gorodnichenko (2011, dashed line)
- Grey shaded areas are U.S. recessions.
- Annual inflation rate axis span visible from 1970 to 2005 with tick marks showing 0 through 10.

### Response of inequality to permanent changes in the inflation target (Figure 11; section 5)
- Shock: permanent 1 percent point decrease in the inflation target.
- Rows:
  - First row: impulse responses and one standard deviation confidence intervals of effects on inequality (cross-sectional standard deviations).
  - Second row: contribution of changes in inflation target to forecast error variance of inequality measures.
  - Bottom row: historical contribution of target inflation changes to deviations of inequality from trend.
- Inflation-target measures:
  - Coibion and Gorodnichenko (2011): grey shaded areas denote one standard deviation confidence intervals; dotted lines and circles used in some panels.
  - Ireland (2006): solid line and dotted lines for point estimates and one standard deviation intervals in some panels; dashed lines used in historical contribution panels.
- Inequality outcomes shown: Income, Earnings, Expenditure, Consumption (IRFs, Contribution to Variance, Historical Contributions).
- See section 5 for details.

### Correlations and volatilities of inequality measures (Table 1; data 1980Q1 to 2008Q4)
- Panel A: Correlations across inequality measures (Corr(SD,Gini), Corr(SD,90th-10th), Corr(Gini,90th-10th))
  - Income Inequality: 0.91 0.93 0.89
  - Earnings Inequality: 0.52 0.85 0.41
  - Expenditures Inequality: 0.89 0.89 0.75
  - Consumption Inequality: 0.63 0.80 0.45
- Panel B: Correlations of inequality in Income, Earnings, Expenditures, Consumption (by SD, Gini, 90th-10th)
  - Corr(Income, Earnings): 0.22 0.82 0.12
  - Corr(Income, Expenditures): 0.32 0.50 0.23
  - Corr(Income, Consumption): 0.14 0.67 0.03
  - Corr(Earnings, Expenditures): 0.26 0.55 0.23
  - Corr(Earnings, Consumption): -0.05 0.61 0.00
  - Corr(Expenditures, Consumption): 0.59 0.80 0.42
- Panel C: Volatility (standard deviations) of inequality measures (SD, Gini, 90th-10th)
  - Income Inequality: 0.05 0.02 0.10
  - Earning Inequality: 0.04 0.02 0.13
  - Expenditures Inequality: 0.02 0.01 0.05
  - Consumption Inequality: 0.01 0.01 0.03

### Correlations of inequality measures with macroeconomic variables (Table 2; data 1980Q1 to 2008Q4)
- Panel A: Correlation with quarterly inflation rate (π)
  - Income Inequality: Corr(π,SD) -0.12  Corr(π,Gini) 0.02  Corr(π,90th-10th) -0.05
  - Earnings Inequality: -0.07 -0.16 -0.11
  - Expenditures Inequality: -0.05 -0.09 -0.04
  - Consumption Inequality: 0.06 0.03 0.03
- Panel B: Correlation with unemployment rate (UE)
  - Income Inequality: 0.01 -0.10 -0.02
  - Earnings Inequality: 0.00 0.07 0.08
  - Expenditures Inequality: -0.32 -0.25 -0.22
  - Consumption Inequality: -0.25 -0.25 -0.26
- Panel C: Correlation with Effective Federal Funds Rate (FFR)
  - Income Inequality: Corr(FFR,SD) -0.12  Corr(FFR,Gini) -0.09  Corr(FFR,90th-10th) -0.08
  - Earnings Inequality: -0.08 -0.19 -0.19
  - Expenditures Inequality: 0.00 -0.02 -0.05
  - Consumption Inequality: 0.11 0.09 0.11
- Notes: All series HP-filtered prior to measuring correlations; unemployment and FFR are quarterly averages.

### Decomposition of income by consumption quintile (Table 3; section 3.3)
- Quintiles by consumption of nondurables and services; income source shares and ratio of mean consumption to mean in 3rd quintile.
- Panel A: 1980s (columns: Labor Earnings (1), Business (2), Financial (3), Other (4), Ratio (5))
  - Quintile 1: 0.352 0.022 0.112 0.515 0.42
  - Quintile 2: 0.588 0.040 0.112 0.260 0.73
  - Quintile 3: 0.694 0.057 0.096 0.153 1.00
  - Quintile 4: 0.762 0.059 0.081 0.098 1.34
  - Quintile 5: 0.767 0.088 0.078 0.067 2.18
- Panel B: 1990s
  - Quintile 1: 0.380 0.020 0.106 0.494 0.43
  - Quintile 2: 0.597 0.040 0.097 0.267 0.73
  - Quintile 3: 0.704 0.050 0.086 0.160 1.00
  - Quintile 4: 0.770 0.056 0.071 0.103 1.35
  - Quintile 5: 0.773 0.082 0.076 0.069 2.27
- Panel C: 2000s
  - Quintile 1: 0.435 0.019 0.086 0.460 0.43
  - Quintile 2: 0.653 0.029 0.085 0.234 0.73
  - Quintile 3: 0.740 0.037 0.072 0.151 1.00
  - Quintile 4: 0.801 0.042 0.065 0.092 1.36
  - Quintile 5: 0.812 0.051 0.071 0.065 2.32

### Decomposition of expenditures and consumption by quintile (Table 4; section 3.3)
- Shares in consumption (Nondurables, Durables, Services), selected shares in total spending (Interest-sensitive expenditures, Mortgage payments, Purchases of new vehicles), and Ratio of total spending to consumption of nondurables and services.
- Panel A: 1980s (examples)
  - Quintile 1: 0.697 0.054 0.249 | 0.098 0.015 0.014 | Ratio 1.84
  - Quintile 5: 0.611 0.114 0.275 | 0.260 0.085 0.058 | Ratio 1.83
- Panel B: 1990s (examples)
  - Quintile 1: 0.655 0.059 0.285 | 0.113 0.021 0.015 | Ratio 2.13
  - Quintile 5: 0.567 0.116 0.317 | 0.267 0.100 0.051 | Ratio 1.91
- Panel C: 2000s (examples)
  - Quintile 1: 0.630 0.057 0.313 | 0.120 0.033 0.014 | Ratio 2.23
  - Quintile 5: 0.541 0.109 0.351 | 0.278 0.110 0.051 | Ratio 1.99
- Note: Consumption includes nondurables, services and durables. Total spending is sum of consumption and other expenditures. See section 3.3 for details.

### Robustness and auxiliary figures (Appendix Figures; section references)
- Appendix Figure 1: Robustness of baseline inequality results to sample and lags (Gini coefficients for income, salary, expenditures, consumption). Scenarios include longer lag lengths, dropping Volcker disinflation, dropping recessions; baseline vs dotted alternative intervals. See section 3.2.
- Appendix Figure 2: Robustness to econometric approach (MA estimation, controlling for other shocks, VAR estimation). See section 3.2.
- Appendix Figure 3: Robustness to controlling for household size, observables, and hours (cross-sectional standard deviations). See section 3.2.
- Appendix Figure 4: Robustness of earnings responses by percentiles (P10, P25, P50, P75, P90) under controls for household observables and labor supply. See section 3.4.
- Appendix Figure 5: Contribution of monetary policy shocks to variance of macroeconomic variables (horizons quarterly, x-axis) including Real GDP, Real Consumption, Real Investment, Unemployment rate, GDP Deflator inflation, Stock Prices, Federal Funds Rate, 10-Year Treasury, House Price Index, Wages and salary, Business income, Financial income, Transfers/Other, Durable goods, Nondurable goods, Services. Solid line = median estimate; dotted lines = one standard-deviation confidence intervals. See section 3.5.
- Appendix Figure 6: Distributional effects by percentile of inflation target increases (responses to a 1 percent point increase in target inflation measured by Coibion and Gorodnichenko (2011) and Ireland (2006)). Percentile differentials plotted include P10-P50 and P90-P50 for Income, Earnings, Expenditures, Consumption. See section 5.

*Content derived from _wp12199 - 4. See section 3.2 for (PDF: _wp12199 - 4. See section 3.2 for).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12199.pdf_
