## 1. Non-Durable and Durable Consumption

## Source details

**Canonical URL:** [1. Non-Durable and Durable Consumption](https://www.imf.org/-/media/files/publications/wp/2019/wp1911.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wp1911.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wp1911.pdf.json)

---

### Introduction
- Common perception: monetary policy in advanced economies has been less effective since the crisis because of higher household debt and associated credit constraints (Sufi, 2015).
- Prior literature typically finds more indebted and less liquid households react more to monetary policy (Aldangady, 2014; Cloyne, Ferreira, and Surico, 2018; Di Maggio and others, 2017; Flodén and others, 2017; Luo, 2017).
- Open question: at very high debt levels, monetary easing may do little to alleviate credit constraints; responsiveness could display an inverted U-shaped pattern (Alpanda and Zubairy, 2017; Sufi, 2015).

### Hypotheses and Research Questions
- Main questions explored:
  - (1) Has the response of household consumption to monetary policy shocks declined since the global financial crisis?
  - (2) Do households with greater indebtedness respond more strongly to monetary policy shocks? Is there evidence of nonlinearities—particularly a decline in responsiveness after a threshold?
  - (3) Do households with low levels of liquid assets react more to monetary policy shocks? Are non-linear effects discernable?
  - (4) Can shifts in the distribution of household indebtedness and liquidity between the pre- and post-crisis periods explain changes in the average response of household consumption to monetary policy?

### Data, Variables, and Measurement
- Data source and period:
  - Consumer Expenditure Survey (CEX), 1996Q1–2014Q4; about 7,500 households interviewed per quarter.
- Consumption measures:
  - Durable and non-durable consumption expenditures constructed; variables in constant dollars (2000Q1 = 100) and winsorized at 1 percent of each tail.
- Balance-sheet measures:
  - Indebtedness: ratio of total mortgage balance (summed over all properties owned) to the value of the houses owned (LTV); excludes other liabilities.
  - Liquidity: ratio of liquid assets to monthly income (liquid assets include checking and savings; income is after-tax).
- Stylized sample statistics:
  - On average, households spend four times more on non-durable consumption relative to durable consumption in any given quarter.
  - Standard deviation of durable consumption is notably larger than that of non-durable consumption.
  - Homeownership: 74 percent of households own a house; almost two-thirds of owners have mortgage debt.
  - Average expenditures: Non-durable goods: $4,320; Durable goods: $1,048.

### Methodology (overview)
- Identification:
  - Monetary policy shocks identified using high-frequency data at the time of policy announcements, in the Bernanke and Kuttner (2005) tradition; for comparability use changes in 2-year bond yields.
  - Quarterly monetary policy surprise constructed by summing announcement-day surprises within a quarter.
- Empirical approaches:
  - Synthetic cohort analysis to obtain longer time series (42 cohorts).
  - Standard panel data methods exploiting full micro data.
  - Local projections (Jordà, 2005) for horizons h = 1,...,12 and household-level GMM with 2yyt instrumented by high-frequency shocks and lags.
- Key regressions:
  - Cohort-level: cumulative log change in real consumption by cohort j between t and t+h regressed on 2yyt and postGFC (2009Q1 onwards) interactions; controls include age, age squared, inflation, GDP growth, quarterly dummies.
  - Household-level: cumulative log change in real consumption for household i between t-1 and t+1; household controls include race, education, age, family size, marital status; clustered standard errors.

### Key Findings — Aggregate and Time Variation
- Average response over time:
  - The response of household consumption to monetary policy shocks has diminished since the global financial crisis.
- Pre-GFC vs post-GFC:
  - Pre-GFC effect (β1(h)) is negative on both durable and non-durable consumption growth.
  - The additional post-GFC effect (β3(h)) is positive at most horizons, implying a weaker (smaller absolute) response of consumption to monetary policy post-crisis; evidence is suggestive but imprecise and sensitive to specification.
- Selected empirical estimates:
  - Full-sample 2-yr yield effects (Table 4): Non-durables: -24.24*** (std. error 4.46); Durables: -4.37* (std. error 2.27).
  - Micro-level pre-crisis GMM: a 10-basis-point reduction in the 2-year yield increases non-durable consumption by about 3 percent and durable consumption by about 2 percent.
  - Post-crisis responses are clearly weaker; for durables the post-crisis effect is only marginally significant in some specifications.
  - Example: a 10 basis-point increase in the 2-year yield reduces non-durable consumption by 2.5 percent in full-sample estimates.

### Role of Household Indebtedness (LTV)
- Core patterns:
  - Higher LTV increases responsiveness to monetary policy shocks—particularly for durable consumption.
  - Responsiveness rises with LTV and is non-linear: for durables the overall impact becomes significant only at LTV levels higher than 0.5.
  - No evidence of a debt-overhang where very high indebtedness reduces responsiveness; instead, households with the highest indebtedness respond most.
- Threshold and distributional findings:
  - Non-durables: response to a 10-basis-point rise in 2-year yield by LTV thresholds:
    - LTV < 70: -0.63*** (0.004)
    - LTV > 70: -0.72*** (0.002)
    - LTV < 80: -0.63*** (0.004)
    - LTV > 80: -0.76*** (0.002)
    - LTV < 95: -0.63*** (0.004)
    - LTV > 95: -1.09*** (0.002)
    - LTV < 99: -0.64*** (0.004)
    - LTV > 99: -2.03*** (0.001)
  - Durables: threshold coefficients reported (often not significant) e.g., -2.19 (0.201), -2.35 (0.203), -2.22 (0.198), -2.55 (0.185), -2.24 (0.193), -3.49 (0.204), -2.25 (0.202), -6.55 (0.21).
- Distributional change and implied effect on aggregate responsiveness:
  - Proportion in top 10 percentile of LTV grew from 5 percent pre-crisis to 8 percent post-crisis.
  - This shift increases responsiveness of non-durable consumption to a 10-basis-point rise in the 2-year yield by 2 basis points, and durable consumption by 4 basis points (from equation 3-based calculations).
  - Using equation 4, the 5→8 percent shift implies a 3-basis-point increase (non-durables) and 6-basis-point increase (durables) in responsiveness to a 10-basis-point hike.
- Conclusion on indebtedness:
  - Indebtedness increases responsiveness and exhibits non-linearities; distributional shifts in LTV do not explain the post-crisis decline in aggregate responsiveness (if anything they would have increased responsiveness).

### Role of Household Liquidity
- Core patterns:
  - Households with lower liquid assets respond more strongly to monetary policy shocks for non-durable consumption.
  - Direct aggregate interaction estimates (β2) are often insignificant, but threshold analyses reveal stronger responses at low liquidity levels.
- Selected results:
  - Share of “hand-to-mouth” households is nearly 60 percent; about two thirds of hand-to-mouth households are homeowners (“wealthy hand-to-mouth”).
  - Threshold regressions for non-durables show stronger responses at low liquid-assets-to-monthly-income ratios (up to around one).
  - Example coefficients:
    - I.(liq > 0.10)*2-yr yield = 3.93*** (1.51) indicating stronger response among low-liquidity households.
    - Table 8: Non-durables responses at liquidity thresholds (examples): LIQ < 25 vs LIQ > 25: -2.34*** (0.006) vs -2.11*** (0.007); as liquidity lowers from the 20th to the 10th and 5th percentiles non-durable response rises by about 0.2 percentage points.
- Distributional change and implied effect:
  - Share in lower 25th percentile of liquidity rose from 24 percent pre-crisis to 28 percent post-crisis.
  - This shift should strengthen responsiveness of non-durable consumption by 0.1 basis point for a 10-basis-point increase in 2-year yield (equation 6 implication).
- Conclusion on liquidity:
  - Liquidity constraints matter for non-durables; however, distributional shifts in liquidity do not explain the observed post-crisis decline in average monetary policy effectiveness.

### Role of Economic Policy Uncertainty
- Specification:
  - Interact monetary policy shocks with the Uncertainty index (Baker, Bloom, and Davis, 2016).
- Findings:
  - Transmission to household consumption is stronger when economic policy uncertainty is low, especially for non-durables.
  - Table 9 selected coefficients:
    - 2-yr yield on Non-Durables -13.77*** (2.70); Uncertainty*2-yr yield 0.01*** (0.00); Uncertainty index -0.14*** (0.03).
    - For Durables the interaction coefficients are not significant at conventional levels.
- Quantitative contribution:
  - Using equation (7) and mean uncertainty levels pre- and post-crisis, higher uncertainty explains about one fourth of the reduction in non-durable consumption responsiveness and about one fifth for durable consumption.
- Conclusion:
  - Elevated economic policy uncertainty post-crisis is a plausible partial explanation for diminished monetary transmission; evidence is tentative and warrants further research.

### Robustness, Identification and Estimation Notes
- Identification:
  - 2-year bond yields on announcement days capture surprise component of monetary policy; 2yyt instrumented with contemporaneous high-frequency shocks and lags.
- Estimation:
  - GMM used with overidentification exploiting contemporaneous shocks and lags; Jordà (2005) local projections for IRFs.
  - Household-level regressions use cumulative 2-quarter consumption growth with 2yyt instrumented; clustered standard errors by household.
- Robustness:
  - Results robust to alternative decompositions of monetary policy shocks (signal shock and risk shock).
  - Size and significance of post-GFC interaction vary with specification (e.g., inclusion of additional lags), so weakening evidence is suggestive rather than conclusive.

### Synthesis and Policy Implications
- Heterogeneity matters:
  - More indebted households and households with lower liquid assets are the most responsive to monetary policy shocks.
  - Responsiveness increases non-linearly with indebtedness (LTV).
- Decline in aggregate responsiveness post-GFC:
  - Not explained by shifts in household balance-sheet distributions (indebtedness or liquidity); these shifts would, if anything, have increased responsiveness.
  - Elevated economic policy uncertainty post-crisis can account for a meaningful but partial share (roughly 20–25 percent) of the decline.
- Policy relevance:
  - Monetary policymakers should monitor distributional features of household balance sheets (LTV and liquidity) because aggregate transmission depends on distributional characteristics, not only on means.
  - Consideration of economic policy uncertainty is important when assessing expected monetary policy effectiveness; higher uncertainty weakens transmission to consumption, especially non-durables.

### Appendix — Additional Summary Statistics and Construction Details
- Household indebtedness and liquidity:
  - Average indebtedness over the sample is nearly 60 percent.
  - Standard deviation of consumption is 42 percent.
  - Distribution of indebtedness is skewed right and does not change markedly pre- to post-crisis.
  - More than 80 percent of mortgage contracts in the sample are fixed rate mortgages.
  - Share of “hand-to-mouth” households is nearly 60 percent; about two thirds of these are homeowners.
  - Median liquidity lowest for renters, highest for homeowners without mortgages; liquidity distribution skewed toward lower levels and stable over time.
- Correlations (select exact coefficients):
  - Non durable consumption with Durable consumption: 0.0451
  - Non durable consumption with LTV: -0.0068
  - Non durable consumption with Liquidity: 0.0098
  - LTV with Liquidity: -0.1774
  - Reference age with LTV: -0.4468; with Liquidity: 0.2449
- Synthetic cohorts:
  - 14 birth-year groups (5-year intervals) × 3 housing-tenure groups → 42 representative consumer units; unbalanced panel with a minimum of 20 consumer units per cohort.
- Data construction notes:
  - Consumption categories follow Aladangady (2014); durable components include cars and furniture; non-durable components include food, utilities, healthcare, education, etc.
  - Leverage proxied by mortgage balance to reported house value (LTV); liquid assets include checking and savings (from 2013 include money market and CDs).
  - Interview timing adjusted to align CEX recall period with calendar quarters.

*Source: wp1911 - 1. Non-Durable and Durable Consumption*

### 1. Non-Durable and Durable Consumption ...............................................................................7

### 1. Non-Durable and Durable Consumption

### Introduction
- Common perception: monetary policy in advanced economies has been less effective since the crisis because of higher household debt and associated credit constraints (Sufi, 2015).
- Prior literature: a few studies focused on pre-crisis period and typically find more indebted and less liquid households react more to monetary policy (Aldangady, 2014; Cloyne, Ferreira, and Surico, 2018; Di Maggio and others, 2017; Flodén and others, 2017; Luo, 2017).
- Open question: at very high debt levels, monetary easing may do little to alleviate credit constraints (Alpanda and Zubairy, 2017; Sufi, 2015); responsiveness could display an inverted U-shaped pattern.

### Hypotheses and Research Questions
- Main questions explored:
  - (1) Has the response of household consumption to monetary policy shocks declined since the global financial crisis?
  - (2) Do households with greater indebtedness respond more strongly to monetary policy shocks? Is there evidence of nonlinearities—particularly a decline in responsiveness after a threshold?
  - (3) Do households with low levels of liquid assets react more to monetary policy shocks? Are non-linear effects discernable?
  - (4) Can shifts in the distribution of household indebtedness and liquidity between the pre- and post-crisis periods explain changes in the average response of household consumption to monetary policy?

### Data, Variables, and Measurement
- Data source and period:
  - Consumer Expenditure Survey (CEX) for household-level consumption, income, and balance-sheet data between 1996Q1 and 2014Q4.
  - The CEX provides about 7,500 households interviewed per quarter.
- Consumption measures:
  - Constructed measures of durable and non-durable consumption expenditures to allow different impacts of monetary policy across categories.
  - Consumption variables are in constant dollars (2000Q1 = 100) and winsorized at 1 percent of each tail (note in Table 1).
- Balance-sheet measures:
  - Indebtedness: ratio of each household’s total mortgage balance (summed over all properties owned) to the value of the houses it owns, as reported by households. Excludes other liabilities like credit card balances.
  - Liquidity: ratio of liquid assets to monthly income (liquid assets include total balances on checking and savings; income is after-tax).
- Sample features and stylized statistics:
  - On average, households spend four times more on non-durable consumption relative to durable consumption in any given quarter.
  - The standard deviation of durable consumption is notably larger than that of non-durable consumption.
  - The distribution of consumption quarter-on-quarter growth changes little after the crisis for both durable and non-durable categories; the distribution of consumption levels shifts slightly to the left after the crisis.
  - Homeownership: 74 percent of households own a house, of which almost two-thirds have mortgage debt.

### Methodology (overview from text)
- Identification of monetary policy shocks using exogenous instruments from high-frequency data in the tradition of Bernanke and Kuttner (2005).
- Empirical approaches:
  - Synthetic cohort analysis to obtain longer time series and derive local projections.
  - Standard panel data methods exploiting the full micro data set.
- Focus on cross-sectional heterogeneity by indebtedness (mortgage-to-value) and liquidity (liquid assets to monthly income).

### Key Findings and Results (summary)
- Average response over time:
  - The response of household consumption to monetary policy shocks has diminished since the global financial crisis.
- Role of indebtedness:
  - Higher-indebted households tend to respond more to monetary policy shocks—particularly for durable consumption—in both pre- and post-crisis periods.
  - Effects appear non-linear, but not U-shaped: households with the highest indebtedness respond most to monetary policy shocks.
  - Because the distribution of debt did not change markedly with the crisis, and its average even increased somewhat, household debt did not contribute to lessening the effects of monetary policy over time.
- Role of liquidity:
  - Households with lower levels of liquid assets react more strongly to monetary policy shocks in both pre- and post-crisis periods.
  - Distribution of liquidity across households remained stable over time; therefore, liquidity constraints cannot explain the decline in monetary policy effectiveness.
- Alternative explanation:
  - The decline in monetary policy effectiveness likely lies elsewhere, such as a higher degree of economic uncertainty brought about by the crisis. The paper finds tentative evidence in favor of this hypothesis.

### Implications and Interpretation
- Heterogeneity matters: indebtedness and liquidity credibly shape heterogeneous household responses to monetary policy, with stronger reactions from highly indebted and low-liquidity households.
- Aggregate decline in responsiveness is not explained by shifts in the distributions of indebtedness or liquidity between pre- and post-crisis periods.
- Policy relevance: researchers and policymakers should consider mechanisms beyond household balance-sheet distributions—such as uncertainty—to explain post-crisis attenuation in monetary transmission to consumption.

*Source: wp1911 - 1. Non-Durable and Durable Consumption*

### Appendix  1)  .

### Appendix  1)

### Household indebtedness and liquidity
- Average indebtedness among households over the entire sample is high, at nearly 60 percent.
- Standard deviation of consumption is 42 percent.
- The distribution of indebtedness is skewed to the right and does not change particularly from pre- to post-crisis.
- More than 80 percent of mortgage contracts in our sample are fixed rate mortgages.
- Liquidity:
  - The share of “hand-to-mouth” households is nearly 60 percent.
  - About two thirds of hand-to-mouth households are homeowners (i.e., “wealthy hand-to-mouth”).
  - Median liquidity is lowest for renters, and highest for homeowners without mortgages.
  - The distribution of liquidity is especially skewed towards lower liquidity levels due to hand-to-mouth households.
  - The distribution of liquidity does not change noticeably from pre- to post-crisis.

### Identification of monetary policy shocks (methods)
- Monetary policy shocks are identified using high-frequency data at the time of monetary policy announcements, in the tradition of Bernanke and Kuttner (2005).
- For comparability across pre- and post-crisis periods, changes in 2-year bond yields are used instead of futures on Federal Fund Rates.
- Identifying assumption: 2-year bond yields on the day prior to a scheduled monetary policy announcement capture market expectations of future policy rates and perceptions of policy uncertainty (term premia); changes on announcement days reflect the surprise component of monetary policy.
- Monetary policy surprises from all announcements in a given quarter are summed to construct quarterly measures, consistent with consumption data aggregation.
- Instrumentation and estimation:
  - The 2-year yield (2yyt) is instrumented using exogenous monetary policy shocks from high-frequency data (the contemporaneous shock and its lags) to address endogeneity and weak instrument bias.
  - Generalized Method of Moments (GMM) is used for estimation.
  - Overidentification is exploited by using contemporaneous shocks and lags as instruments.
  - Jordà (2005) local projection method is used to estimate impulse response functions for horizons h = 1,...,12.

### Construction of synthetic cohorts and panel
- Synthetic cohorts are constructed to obtain longer time series given CEX households are observed for only four consecutive quarters.
- Cohort construction:
  - 14 birth-year groups using 5-year intervals.
  - 3 housing-tenure groups: owners with mortgage, owners without mortgage, renters.
  - Result: 42 representative consumer units with data for the whole sample period.
- Underlying assumption: households in the same bucket respond similarly to monetary policy shocks.

### Empirical specification highlights
- Local projection specification (summary):
  - Dependent variable: cumulative log change in real consumption by cohort j between t and t+h.
  - Key regressor: 2-year yield (2yyt), interacted with a postGFC dummy for 2009Q1 onwards to test for changes after the Global Financial Crisis.
  - Controls: cohort-specific controls (age and age squared), macro controls (inflation, GDP growth), quarterly dummies.
- Household-level micro regressions:
  - Dependent variable: cumulative log change in real consumption for household i between t-1 and t+1 (two-quarter growth).
  - Household controls include race, education level, age, family size, marital status; seasonal fixed effects are included.
  - 2-year yields instrumented with high-frequency monetary policy shocks and their lags; GMM estimation applied.

### Main empirical findings
- Pre-GFC vs post-GFC responsiveness:
  - The pre-GFC effect of monetary policy (β1(h)) is negative on both durable and non-durable consumption growth.
  - The additional post-GFC effect (β3(h)) is positive at most projection horizons, implying a weaker (smaller absolute) response of consumption to monetary policy post-crisis.
  - The evidence points to a likely weakening of monetary policy effects on household consumption after the crisis, but measurement is imprecise and sensitive to specification (e.g., inclusion of additional lags).
- Micro-level estimates (household-level GMM results):
  - In the pre-crisis period, an expansionary monetary policy shock (a 10-basis point reduction in the 2-year yield) increases:
    - Non-durable consumption by about 3 percent.
    - Durable consumption by about 2 percent.
  - In the post-crisis period, the response of both durable and non-durable consumption to monetary policy is clearly weaker (statistically captured by positive and significant β2).
  - For durable consumption, the post-crisis effect is only marginally statistically significant in some specifications.
- Full-sample estimates (no post-GFC dummy):
  - Expansionary monetary policy boosts both durable and non-durable consumption.
  - A 10 basis-point increase in the 2-year yield reduces non-durable consumption by 2.5 percent (and reduces durable consumption by a smaller amount as reported in the estimates).

### Additional empirical notes
- Household-level controls: college-educated, white, married, and older households display higher non-durable consumption growth following looser monetary policy; these characteristics are not important determinants of durable consumption.
- Robustness:
  - Results are robust to instrumenting the policy rate with alternative monetary policy shock decompositions (signal shock and risk shock), as described in Appendix III.
  - Size and significance of the post-GFC interaction vary with specification, leading authors to characterize evidence of weakening as suggestive but not conclusive.

*Source: IMF staff estimates and analysis in Appendix 1 of the provided document.*

### 0.5 percent.

### 0.5 percent.

### Key empirical setup and data
- GMM estimation, 1996Q1 - 2014Q4. Dependent variable: 2-quarter ahead consumption growth. In the first stage 2-year yield is instrumented by monetary policy shocks.
- Regressions include a constant and quarter (seasonal) effects. Clustered standard errors (by households) reported in parenthesis.
- Main samples and counts reported in tables: e.g., Observations 166,921 (Non-durables) and 69,781 (Durables); No. of Households 85,246 and 47,356 in some specifications.

### Main findings on consumption response (aggregate)
- Average responsiveness of U.S. household consumption to well-identified monetary policy shocks has declined since the global financial crisis (GFC).
- Consumption responsiveness is measured as the impact of a 10-basis-point increase in the 2-year yield on consumption growth (expressed in percentage points).
- Example coefficient estimates (selected):
  - Full-sample 2-yr yield effects (Table 4, columns): Non-durables: -24.24*** (std. error 4.46); Durables: -4.37* (std. error 2.27).
  - Table 9 (uncertainty interaction): 2-yr yield effect on Non-Durables -13.77*** (2.70); on Durables -17.51 (12.81). Uncertainty*2-yr yield 0.01*** (0.00) for Non-Durables.

### Role of household indebtedness (LTV)
- Higher indebtedness increases responsiveness to monetary policy shocks; estimated β2 has a negative sign (consistent with higher responsiveness of more indebted households) and is significant for durable consumption in some specifications.
- Non-linearities: responsiveness rises with LTV and becomes significant for durables only beyond certain LTV thresholds.
- Figure/estimates:
  - For durable consumption, overall impact of monetary policy is significant only at LTV levels higher than 0.5.
  - Table 5 selected coefficients: e.g., 2-yr yield coefficients vary widely across specifications: -22.41*** (7.19), 5.66 (6.02), -8.73*** (2.81), -27.33 (22.05), -18.93*** (4.30), -41.46* (23.41).
  - Interaction LTV*2-yr yield: example coefficient -10.67 (10.11) and -21.43** (10.04) in columns reporting LTV interactions.
- Threshold analysis (Table 6 and discussion):
  - Non-durables: at various LTV thresholds the response to a 10-basis-point rise in 2-year yield: LTV < 70: -0.63*** (0.004); LTV > 70: -0.72*** (0.002); LTV < 80: -0.63*** (0.004); LTV > 80: -0.76*** (0.002); LTV < 95: -0.63*** (0.004); LTV > 95: -1.09*** (0.002); LTV < 99: -0.64*** (0.004); LTV > 99: -2.03*** (0.001).
  - Durables: coefficients reported (not always significant) e.g., -2.19 (0.201), -2.35 (0.203), -2.22 (0.198), -2.55 (0.185), -2.24 (0.193), -3.49 (0.204), -2.25 (0.202), -6.55 (0.21).
- Distributional change and implications:
  - The proportion of households in the top 10 percentile of LTV distribution grew from 5 percent before crisis to 8 percent in the post-crisis period.
  - Using equation 3 estimates (Figure 3), the shift in the LTV distribution post-crisis increases responsiveness of:
    - Non-durable consumption to a 10-basis-point rise in the 2-year yield by 2 basis points.
    - Durable consumption to a 10-basis-point rise in the 2-year yield by 4 basis points.
  - According to equation 4, the 5→8 percent shift in top 10 percentile implies a 3-basis-point increase (non-durables) and 6-basis-point increase (durables) in responsiveness to a 10-basis-point hike in the 2-year yield.
- Conclusion on indebtedness: indebtedness increases responsiveness and exhibits non-linear effects; no evidence that very high indebtedness produces a debt-overhang that weakens responsiveness in general.

### Role of household liquidity
- Hypothesis: households with low liquidity respond more to monetary policy shocks (positive β2 in liquidity interaction).
- Direct estimates of β2 are insignificant for aggregate tests (Table 7, columns 1 and 2).
- Non-linear analysis and threshold regressions reveal:
  - Responsiveness of non-durable consumption is significant mainly at relatively low liquidity values (liquid-assets-to-monthly income ratios up to around one).
  - Threshold specification (I.(liq > 0.25) etc.) shows that I.(liq > 0.10)*2-yr yield = 3.93*** (1.51) for non-durables, indicating stronger response among low-liquidity households.
- Table 7 selected coefficients:
  - 2-yr yield: e.g., Non-durables -25.58** (10.39); -23.41*** (8.58); -24.92*** (9.27). Durables coefficients in corresponding columns weaker or insignificant.
  - Liquid assets/Income coefficient examples: -23.82 (35.52) and -1.85 (14.13) (not significant).
  - I.(liq > 0.25) coefficient -4.00* (2.25); I.(liq > 0.25)*2-yr yield 2.27*** (0.75).
- Table 8 (response at liquid-asset-to-income thresholds) shows:
  - Non-durables: LIQ < 25 versus LIQ > 25: -2.34*** (0.006) vs -2.11*** (0.007); LIQ < 20: -2.39*** (0.006) etc. As liquidity is lowered from the 20th to the 10th and 5th percentiles, non-durable consumption response rises (e.g., increases by about 0.2 percentage points).
  - Durables responses are generally not statistically significant in these thresholds.
- Distributional change:
  - Share of households in the lower 25th percentile of liquidity rose from 24 percent pre-crisis to 28 percent post-crisis.
  - Based on equation 6 estimates, this shift should strengthen responsiveness of non-durable consumption by 0.1 basis point (for a 10-basis-point increase in 2-year yield).
- Conclusion on liquidity: liquidity constraints matter; non-durable consumption of low-liquid households responds more strongly to interest rate changes, but overall distributional shifts in liquidity do not explain the observed post-crisis decline in average monetary policy effectiveness on consumption.

### Role of economic policy uncertainty
- Specification interacts monetary policy shocks with the Uncertainty index (Baker, Bloom, and Davis 2016).
- Results indicate transmission to household consumption is stronger when economic policy uncertainty is low, especially for non-durables.
- Table 9 selected coefficients:
  - 2-yr yield on Non-Durables -13.77*** (2.70); Uncertainty*2-yr yield 0.01*** (0.00). Uncertainty index -0.14*** (0.03).
  - For Durables coefficients are not significant at conventional levels for the interaction.
- Quantitative contribution:
  - Using equation (7) and mean uncertainty levels pre- and post-crisis, higher uncertainty explains about one fourth of the reduction in non-durable consumption responsiveness and about one fifth for durable consumption.
- Conclusion: elevated economic policy uncertainty post-crisis is a plausible partial explanation for diminished transmission; evidence is preliminary and warrants further research.

### Synthesis and policy implications
- Household balance sheets matter for monetary policy transmission:
  - More indebted households and households with lower liquid assets are the most responsive to monetary policy shocks.
  - Responsiveness increases non-linearly with indebtedness (LTV).
- The decline in average consumption responsiveness after the GFC is not explained by worsening household balance sheets (indebtedness or liquidity). In fact:
  - Distributional shifts in LTV and liquidity would, if anything, have increased responsiveness post-crisis.
- Elevated economic policy uncertainty post-crisis can account for a meaningful but partial share (roughly 20–25 percent) of the decline in responsiveness.
- Policy relevance:
  - Monetary policy makers should monitor household balance sheet distributions (LTV and liquidity) because non-linearities imply aggregate transmission depends on distributional features, not only means.
  - Consideration of economic policy uncertainty is important when assessing expected monetary policy effectiveness; higher uncertainty appears to weaken transmission to consumption, especially non-durables.

*Source: IMF staff estimates (excerpts from wp1911 - 0.5 percent).*

### REFERENCES

### wp1911 - REFERENCES

### Key bibliographic coverage
- The references list encompasses empirical and theoretical literature on household debt, monetary policy transmission, consumption responses, and methods for identification of macroeconomic shocks.
- Notable recurring topics and authors:
  - Household debt and consumption dynamics: Aladangady (2014, 2017), Dynan (2012), Mian and Sufi (2013, 2014, 2017), Melzer (2017), Justiniano, Primiceri, and Tambalotti (2015).
  - Monetary policy transmission and asset-price/interest-rate channels: Bernanke and Kuttner (2005), Gürkaynak, Sack and Swanson (2005, 2007), Gilchrist, López-Salido, and Zakrajšek (2015), Hanson and Stein (2015).
  - Debt overhang, deleveraging, and macro-financial interactions: Eggertsson and Krugman (2012), Eggertsson, Mehrotra, and Robbins (2017), Drehmann, Juselius, and Korinek (2017).
  - Heterogeneity and micro-founded macro frameworks: Kaplan and Violante (2014, 2018), Kaplan, Moll, and Violante (2018), Guerrieri and Lorenzoni (2011).
- Methodology and measurement sources include:
  - Measurement of economic policy uncertainty: Baker, Bloom, and Davis (2016).
  - Local projections and impulse response estimation: Jordà (2005).
  - Use of survey data and synthetic cohort methods: Attanasio and Davis (1996); Narita and Narita (2011); Cloyne and others (2017).

### Data sources and survey design (Appendix I)
- Consumer Expenditure Survey (CEX):
  - Conducted by the Census Bureau; primarily used by the Bureau of Labor Statistics for CPI weights.
  - Rotating panel: each household interviewed once per every three months for, at most, 15 consecutive months.
  - Designed to be representative of the U.S. civilian non-institutional population.

- Data cleanup steps and resulting sample:
  - Drop observations with negative consumption.
  - Drop observations for households with more than one consumption unit.
  - Drop households with less than four interview observations.
  - Resulting sample: roughly 5,000 quarterly household observations.
    - 74 percent are homeowners.
    - 45 percent are homeowners with outstanding mortgage balance.
  - Average expenditures:
    - On non-durable goods: $4,320.
    - On durable goods: $1,048.
  - Consumption variables are in constant dollars (2000Q1 = 100) and winsorized at one percent of each tail.

- Interview timing and alignment:
  - CEX asks about consumption over the three months prior to interview month.
  - Interview month may not align with calendar quarter; adjustments are made to align consumption data with calendar quarters.

### Definitions and construction of key variables
- Consumption categories (following Aladangady (2014)):
  - Non-durable consumption components: food, alcohol, tobacco, housing operations, utilities, gasoline, public transportation, personal care, reading, entertainment, apparel, healthcare, and education expenses.
  - Durable consumption components: cars (new and used), furniture, and equipment.
- Housing tenure and summary (Appendix I Table 1 highlights):
  - Share (percent) and average levels/growth presented by tenure:
    - Homeowners: Share 74; Non-Durable 4,803; Durable 1,204; Average Growth 0.14, -3.44
    - w/ mortgage: Share 45; Non-Durable 5,286; Durable 1,397; Average Growth 0.02, -4.02
    - w/o mortgage: Share 29; Non-Durable 4,044; Durable 900; Average Growth 0.32, -2.25
    - Renters: Share 26; Non-Durable 2,935; Durable 601; Average Growth -0.23, 4.26
    - All: Share 100; Non-Durable 4,320; Durable 1,048; Average Growth 0.04, -3.60

- Leverage and liquidity measures:
  - Leverage proxied by ratio of mortgage balance to reported house value (LTV).
    - Mortgage balances aggregated across all properties owned by the household.
    - CEX mortgage variables: QBLNCM1X or QBLNCM2X (mortgage balance at beginning of month, three months prior or two months prior to interview).
    - Property value: PROPVALX; used to construct a house price index that closely matches Case-Shiller Home Price Index in boom-bust dynamics.
    - Refinancing adjustment: mortgage balances before and after refinancing are adjusted to avoid double-counting.
  - Liquid assets:
    - Include checking and savings balances.
    - From 2013 onwards: include money market accounts and certificates of deposit.
    - CEX variables: LIQUIDX for 2013–14; CKBKACTX + SAVACCTX for 1994–2012.
  - Income:
    - Reported in the second and fifth interviews.
    - Use imputed after-tax income, FINCATXM from 2004 onwards.
    - Prior years: use reported after-tax income, FINCATAX, with invalid missing entries replaced by imputed income.

- Cohorts and control variables:
  - Synthetic cohorts constructed using housing tenure (CUTENURE) and household head’s birth year (derived from interview date and AGE_REF).
  - Control variables in panel analysis: race (REF_RACE), education (EDUC_REF), age (AGE_REF), family size (FAM_SIZE), marital status (MARITAL1).

- Appendix I Table 2: mappings of CEX variable names to consumption components (examples):
  - Total Expenditure: TOTEXP
  - Food: FOOD
  - Alcohol: ALCBEV
  - Tobacco: TOBACC
  - Housing operations: HOUSOP
  - Utilities: UTIL
  - Gasoline: GASMO
  - Public transportation: PUBTRA
  - Personal care: PERSCA
  - Reading: READ
  - Entertainment: ENTERT
  - Apparel: APPAR
  - Healthcare: HEALTH
  - Educational expenses: EDUCA
  - Cars & trucks, new: CARTKN
  - Cars & trucks, used: CARTKU
  - Other vehicles: OTHVEH
  - Furnishing & equipment: HOUSEQ

### Correlation patterns (Annex Table 3)
- Select correlation coefficients (exact values from the correlation matrix):
  - Non durable consumption with Durable consumption: 0.0451
  - Non durable consumption with LTV: -0.0068
  - Non durable consumption with Liquidity: 0.0098
  - Durable consumption with LTV: -0.00681
  - Durable consumption with Liquidity: 0.0039
  - LTV with Liquidity: -0.1774
  - Family size correlations: with LTV 0.2209; with Liquidity -0.1502
  - College education correlations: with LTV 0.0839; with Liquidity 0.0879
  - Ethnicity (white = 1) correlations: with LTV -0.0481; with Liquidity 0.0802
  - Marital status correlations: with Liquidity -0.0056; with Family size 0.4539
  - Reference age correlations: with LTV -0.4468; with Liquidity 0.2449; with Family size -0.3801

### Synthetic cohort panel construction and estimation (Appendix II)
- Synthetic cohort construction:
  - Purpose: create panel data from repeated cross-sections to measure consumption responsiveness to monetary policy over time.
  - Cohort dimensions:
    - Birth cohorts defined by 5-year bands; oldest cohort: January 1910–December 1914.
    - Focus on household heads aged 25 to 75.
    - Housing status: owners without mortgage, owners with positive mortgage balance, renters.
  - Resulting panel: unbalanced panel of 42 synthetic cohorts with a minimum of 20 consumer units (CUs) in each cohort.
  - Trade-offs in cohort design:
    - Small set of characteristics chosen to avoid cohorts with few CUs and to avoid short time series.
    - Variability in CU counts across cohorts can increase volatility and standard errors for cohorts with few CUs; short time-series can induce small sample bias.

- Estimation of cohort-level variables:
  - Reduced-form relationship between cohort-level consumption and individual household consumption:
    - Model (as given): log(U_{j,i,t}) = log(U_{i,t}) + ε_{j,i,t}, with ε_{j,i,t} ~ i.i.d. (0, σ_{i,t}^2).
    - Interpretation: individual log consumption is a random draw with mean equal to cohort-level log consumption and variance σ_{i,t}^2.
  - Cohort-level logged consumption estimation:
    - Use weighted average of logged consumption expenditures over households in cohort, employing CEX sample weights ω_{j,i,t}.
    - Define ω_{i,t} := Σ_{j∈I_{i,t}} ω_{j,i,t}.
    - Cohort-level log consumption estimator: lplog(U_{i,t}) := (1/ω_{i,t}) Σ_{j∈I_{i,t}} ω_{j,i,t} lplog(U_{j,i,t}).
    - CEX sample weights interpreted as number of off-sample households represented by each consumer unit.

*Source: wp1911 - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wp1911.pdf_