## wpiea2019076

## Source details

**Canonical URL:** [wpiea2019076](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019076.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019076.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019076.pdf.json)

---

### Section V — A primer on household debt and vulnerabilities
- Empirical literature: not household debt per se but metrics such as rising leverage, excessive credit, or debt distribution that matter for macro analysis.
- Mechanisms and prior findings:
  - High leverage combined with asset price shocks can lead to demand-driven recessions (Mian, Roa and Sufi, 2013).
  - Australia-specific findings:
    - Atalay, Whelan, and Yates (2017): positive relationship between changes in housing wealth and consumption; wealth effect strongest for middle-aged homeowners; wealth effect became smaller post-GFC as households with higher loan-to-value ratios became more conservative.
    - Bilston, Johnson, and Read (2015): stress testing with HILDA data found households resilient in 2000s because debt distribution concentrated among households well placed to service it.
    - Price, Beckers, and La Cava (2018): households typically cut back spending when they have higher levels of outstanding mortgage debt (debt overhang effect).
    - La Cava, Hughson, and Kaplan (2016): central estimates indicate a 100 basis points cut in interest rates resulted in a 0.1 to 0.2 percent increase in household spending in aggregate.
    - La Cava and Price (2017): high debt relative to income and assets, and low debt-servicing capacity can reduce household spending growth relative to income.
- This study: follows Cloyne, Ferreira, and Surico (2018) methodology but uses longitudinal HILDA data, enabling direct use of wealth data and grouping by indebtedness.

### Evolution and characteristics of household debt (Australia)
- Long-run aggregates and drivers:
  - Household debt rising faster than household disposable income for past three decades; among highest household debt ratios in advanced economies.
  - Drivers: financial deregulation in the 1990s; historically low interest rates post-GFC; housing boom driven by income and population growth, relatively inelastic supply, and expectations of future capital gains.
  - Meng, Hoang, and Siriwardana (2013): housing prices, GDP, and population positively affect household borrowing; interest rates, unemployment, number of new dwellings, and inflation negatively affect household debt.
- Key aggregate statistics and trends (preserved exactly):
  - Housing debt at 140 percent of household disposable income accounted for about three-quarters of household debt outstanding as of September 2018.
  - Owner-occupied housing debt accounted for a relatively stable share of about one half of housing debt.
  - Other personal debt remained broadly stable at about 46 percent of household disposable income since 2000 (one quarter of household debt outstanding).
  - Household debt rose steadily to almost 190 percent of gross disposable income by September 2018.
  - Housing debt has remained below 20 percent of total housing assets since 2010, helped by a 42 percent rise in housing prices.
  - Household leverage (liabilities-to-net-worth ratio) trended down to close to 22 percent in 2017, but could increase rapidly if house prices corrected downward.

### Debt measurement issues
- Common scaling metrics:
  - Debt-to-income (DTI): total debt outstanding scaled by annualized gross household disposable income.
  - Debt-service-to-income (DSTI): focuses on size of debt service payments as influenced by interest rates and loan terms.
- Critique and alternatives:
  - Proportional metrics like DTI or DSTI may understate capacity of higher-income households to carry debt because living expenses do not increase one-for-one with disposable income.
  - Alternative scalings: assets (liquid, financial, non-financial) and net worth.
- Notable numeric observations (preserved exactly):
  - Household debt approached 190 percent of gross disposable income by September 2018.
  - Housing debt remained below 20 percent of total housing assets since 2010.
  - Housing prices rose by 42 percent (period referenced in source).
  - Liabilities-to-net-worth ratio declined to close to 22 percent in 2017.

### Distribution of debt — micro evidence (SIH and HILDA)
- SIH 2015-16 (ABS) highlights:
  - Three quarters of households held debt.
  - Share of households with a debt-to-income ratio of three and above has risen across income quintiles.
  - Rise in median DTI most pronounced in the top 40 percent of higher-income households.
  - DTI rose across all age groups; reference person 35-44 years had highest median DTI of 2.3 in 2015-16.
  - SIH 2015-16 sampled 17,768 households over July 2015 to June 2016.
- HILDA comparisons (2002 vs 2014):
  - About 30 percent of respondent households indicated they had no debt in 2014.
  - About one-third of households had a DTI ratio of less than or equal to one in 2014.
  - Close to 20 percent of households had DTI above 3 in 2014.
  - About 27 percent of total households had a debt-to-net worth ratio of less than zero (net liability position).
  - Of the 73 percent with positive net worth, about three quarters had debt-to-net worth ratio between zero and one inclusive, and close to 15 percent had debt-to-net worth ratio between one and two inclusive.
  - Share of total household debt held by the top two disposable income quintiles increased from 70 percent in 2010 to 78 percent in 2014.
  - Household debt held by head of household 45 years or older rose from 38 to 47 percent during 2002-14.
  - By tenure in 2016: mortgagors held 77 percent of total household debt, outright owners 13 percent, renters 10 percent.

### Debt-level and age-group statistics (2002-14)
- Average debt level by quartile (preserved exactly):
  - High debt group: A$460 thousand.
  - Medium-high debt group: A$274 thousand.
  - Medium-low debt group: A$58 thousand.
  - Low debt group: A$36 thousand.
- Household net worth largest for medium-low debt group and smallest for high debt category.
- Age patterns: peak household debt skewed toward higher age groups and higher income in 2002; 30-39 medium-income borrowers had higher debt in 2006; 40-49 high-income households had highest debt in 2014.

### Debt-servicing dynamics and mortgage buffers
- Debt-servicing and repayment behavior:
  - Debt-servicing ratio (scheduled principal and interest mortgage repayments to household disposable income) remained broadly unchanged at 9 to 11 percent in the post-GFC period despite rising household debt.
  - Mortgage interest to scheduled principal repayment ratio declined from 5.8 in 2008 to 1.4 in 2017.
  - Households made higher-than-scheduled total payments by about 2 percent of household disposable income in 2017.
- Prudential measures (APRA, since end-2014):
  - Interest rate buffer of at least 2 percentage points above the effective variable rate applied for the term of the loan.
  - Minimum floor assessment interest rate of at least 7 percent.
- Aggregate mortgage buffers and distribution (RBA, 2018):
  - Total household mortgage buffers were 18 percent of outstanding loan balances or around 2½ years of scheduled repayments at current interest rates in 2017.
  - Distribution: one-third of outstanding owner-occupier mortgages had at least two years equivalent of buffer; about one-quarter had a mortgage buffer of less than one month.
  - Mortgage buffers (offset accounts and redraw facilities) have been rising as households used falling interest rates to pay down debt faster than required.

### Over-indebtedness and vulnerability (definitions and prevalence)
- Over-indebtedness definition used: debt is three or more times income or/and 75 percent or more of the value of assets.
- Under this definition, 29 percent of households are over-indebted based on the ratio of debt to either disposable income or assets.
- Distribution by income quintile:
  - Top three income quintiles: one quarter of households in each quintile are over-indebted.
  - Lowest and second income quintiles: about one-sixth are over-indebted.
- SIH results: about three quarters of over-indebted households did not have sufficient liquid assets to cover a quarter of their debt values.
- Staff estimate (2015-16 data): a one percentage point increase in interest rates would increase the debt service repayment (principal and interest) of over-indebted households in the highest (lowest) income quintile by 22 percent (18 percent).

### Data sources and empirical sample
- Micro data: HILDA survey annual frequency from 2001 to 2016 (2017 wave not included).
- Macroeconomic variables: IMF’s International Financial Statistics (IFS) and World Economic Outlook (WEO) databases.
- Empirical sample: full HILDA survey for 2001-16 for about 16,000 participating households; up to 26,400 households over 16 years used in Arellano-Bond dynamic panel regressions (unbalanced panel).

### Household grouping, variable construction, and outlier treatment
- Debt quartiles constructed using total debt, total property-apportioned debt, and home-apportioned debt.
- Three household groups:
  - high debt — the highest quartile;
  - medium debt — the two middle quartiles;
  - low debt — the lowest quartile.
- Consumption aggregation:
  - detailed consumption aggregated into current consumption and durable expenditure panel series on annual basis; both divided by total disposable income to normalize and detrend.
- Outlier treatment:
  - debt ratios: all outliers removed and negative reported debt indicators removed;
  - current consumption ratios: negative observations removed and 99th percentile for current consumption-to-disposable-income ratio dropped;
  - durable expenditures ratios: first and 99th percentiles dropped.
- Wealth questionnaire frequency: every fourth year (starting 2002); balance sheet assumed unchanged for the four years prior to next wealth questionnaire.

### Monetary policy shock identification and empirical framework
- Shock identification:
  - Romer and Romer (2004) narrative approach; estimates for Australia from Bishop and Tulip (2017) used.
  - Monetary policy shock variable almost perfectly correlated with RBA’s cash rate (correlation coefficient 0.9987).
  - Cash rate excluded from regressions due to collinearity with the monetary policy shock variable.
- Estimation:
  - Specification: regress consumption ratios on lagged monetary policy shocks, controlling for lagged endogenous variable and other controls (household disposable income, average annual inflation, average mortgage rate, world GDP growth for current consumption; year dummies for durable expenditures in some specifications).
  - Estimator: Arellano-Bond dynamic panel regressions for unbalanced panel; robust two-step estimators; one lag for dependent variables.
  - Monetary policy shocks at t and (t-1) considered; final specification uses lagged monetary policy shock due to collinearity.

### Empirical results — differential effects by debt group (preserved coefficients and findings)
- Hypothesis: a positive monetary policy shock (tighter conditions) negatively impacts current consumption and durable expenditure, especially for high-debt households.
- Current consumption to disposable income:
  - Households react negatively and significantly to lagged monetary policy shocks for most debt-ratio groupings; contemporaneous shock often positive (smaller coefficient).
  - Reaction differs by debt ratio type:
    - For total debt ratios, strongest combined reaction for high- and medium-debt households.
    - Low-debt households react positively to the monetary policy shock.
  - Quantitative example (ratio of total debt to total non-financial assets grouping):
    - A monetary policy shock of one basis point would cause the total decrease in the current consumption ratio to total disposable income by 0.28 percentage points for high-debt households.
    - The same shock would decrease the ratio by 0.13 percentage points for medium-debt households.
    - The same shock would cause an overall increase by 0.19 percent for low-debt households.
- Durable expenditures to disposable income:
  - Lagged monetary policy shocks: negative and significant for high- and medium-debt households.
  - High-debt household responses tend to be larger than medium-debt households.
  - Specific preserved coefficients (Table excerpts):
    - Table 1 (Current Consumption; Total Debt to Total Net Worth, Full Sample):
      - L.ConsTDI 0.130***
      - MPShock 0.0153***
      - L.MPShock -0.0161***
      - Observations 168,028
      - Number of xwaveid 26,431
    - Table 1 (examples across debt groups):
      - High Debt MPShock 0.0134***; L.MPShock -0.0151***
      - Medium Debt MPShock 0.0130***; L.MPShock -0.0153***
      - Low Debt MPShock 0.0196***; L.MPShock -0.0177***
    - Table 2 (Durable Expenditures; Total Debt to Total Net Worth, Full Sample):
      - Durable Exp (t-1) 0.0948***
      - MP Shock (t-1) -0.00166***
      - Observations 43,874
      - Number of xwaveid 16,327
    - Table 2 (by debt groups; MP Shock (t-1) examples):
      - High Debt MP Shock (t-1) -0.00306***
      - Medium Debt MP Shock (t-1) -0.00217***
      - Low Debt MP Shock (t-1) 0.000372
    - Table 3 (Current Consumption by Ownership Groups; Full Sample):
      - Curr Cons (t-1) 0.130***
      - MP Shock 0.0153***
      - MPShock (t-1) -0.0161***
      - Observations 168,028
      - Number of xwaveid 26,431
    - Table 4 (Durable Expenditures by Ownership Groups; Full Sample):
      - Durable Exp (t-1) 0.0948***
      - MP Shock (t-1) -0.00166***
      - Observations 43,874
      - Number of xwaveid 16,327

### Empirical magnitudes and subgroup findings (selected preserved estimates and textual summary)
- Durable expenditures: high-debt households reduce durable expenditures by 0.3 percentage points; medium-debt households reduce durable expenditures by 0.2 percentage points. Low-debt households do not respond significantly for durable expenditures.
- Ownership subgroup responses:
  - Current consumption: most negative and statistically significant for mortgagors, followed by renters and outright owners.
  - Durable expenditures: mortgagors most negative and significant, followed by renters; outright owners’ reaction positive but insignificant.
- Consistency: results consistent with Cloyne, Ferreira, and Surico (2018) and La Cava, Hughson, and Kaplan (2016).

### Aggregate impact calculations and scenarios (contractual short-term 25bps monetary policy shock)
- Methodology: use regression coefficients to compute sensitivity of current consumption and durable expenditures ratios to aggregate household total disposable income under a 25 basis points monetary policy shock in period (t-1).
- Reported numerical impacts (preserved exactly):
  - Using Short-Term Impact = (γ(t-1)+γ(t))*0.25 and Long-Term Impact = ((γ(t-1)+γ(t))/(1-β(t-1))):
    - Full Sample: Short-Term Impact = -0.0002; Long-Term Impact = -0.0009
    - High Debt: Short-Term Impact = -0.0004; Long-Term Impact = -0.0020
    - Medium Debt: Short-Term Impact = -0.0006; Long-Term Impact = -0.0026
    - Low Debt: Short-Term Impact = 0.0005; Long-Term Impact = 0.0022
  - Using Short-Term Impact = γ(t-1)*0.25 and Long-Term Impact = γ(t-1)/(1-β(t-1)):
    - Full Sample: Short-Term Impact = -0.0004; Long-Term Impact = -0.0018
    - High Debt: Short-Term Impact = -0.0008; Long-Term Impact = -0.0032
    - Medium Debt: Short-Term Impact = -0.0005; Long-Term Impact = -0.0024
    - Low Debt: Short-Term Impact = 0.0001; Long-Term Impact = 0.0004
- Qualitative implications from scenarios (preserved conclusions):
  - One-period impact on aggregate ratios of current consumption and durable expenditures to total disposable income would be negative for the full sample and all subsamples except durable expenditures of low-debt households.
  - Impact on ratio of current consumption is lower than on durable expenditures for full sample and high-debt households—those households adjust durable expenditures immediately but maintain certain categories of current consumption.
  - Some low-debt households would increase current consumption and durable expenditures after a contractionary monetary policy shock; low-debt households increase current consumption more than durable expenditures.
  - Long-run effects accumulate: positive for low-debt households; negative for full sample and households with high and medium debt.

### Textual analysis of RBA communications
- Data and coverage:
  - Word phases: household debt, housing debt, household leverage, household debt-to-income, housing prices, housing market, credit to housing sector, mortgage rate.
  - Documents: Interest Rate Decisions (11 releases per year), RBA Board Minutes (11 per year), Statement on Monetary Policy (four times per year in February, May, August, and November), speeches by RBA Governor and senior management.
  - Period: 2008 to cut-off data as of December 12, 2018.
  - Method: Python text mining and word-counting selected word phases.
- Findings (preserved summary):
  - RBA increased focus on household debt issues in communications with increased word counts for household debt word phases in 2015-17, with a decrease somewhat in 2018.
  - Word counts for housing prices and housing market picked up around 2012, coinciding with the start of the housing boom.
  - Word counts on credit to the housing sector increased sharply since (text cut off in source).

### Policy implications and conclusions (Section VII)
- Evolution and distribution:
  - High debt exposure more prevalent among higher-income and higher-wealth households; debt exposure of lower-income and more vulnerable households has increased over time.
  - Presence of over-indebted households at both low- and higher-income quintiles increases macro-financial risks and requires close monitoring.
- Debt service capacity and policy risk:
  - Despite high debt levels, debt service burden manageable due to historically low mortgage interest rates.
  - APRA mortgage serviceability buffers for new lending reduce immediate vulnerability but downside risks remain if global financial conditions tighten.
- Monetary policy transmission and implications:
  - High-debt households reduce current consumption and durable expenditures relatively more in response to contractionary monetary shocks; low-debt households may not respond because income channel dominates intertemporal substitution.
  - Given larger share of high-debt households and higher responsiveness, it may take a smaller increase in the cash rate for the RBA to achieve policy objectives ("more bang for the buck").
  - Recommendation: continue transparent and strengthened communication on household debt and consumption to improve predictability and efficiency of monetary policy in Australia.

### Data appendix highlights and descriptive statistics (selected preserved entries)
- HILDA panel waves 1-16 (2001-2016); original combined datasets produced 2,441 variables with 317,738 observations; trimmed panel used 293 variables and 317,738 observations.
- Selected variable entries (preserved exactly as reported):
  - Observations and counts in tables: Observations 168,028; Observations 43,874; Number of xwaveid 26,431; Number of xwaveid 16,327.
  - Variable summaries (examples preserved exactly):
    - Age — Obs 317738; Mean 36; Std. Dev. 23; Min 1; Max 101
    - ConsCurrA — Obs 244586; Mean 292661553372436828
    - ExpDur — Obs 87364; Mean 112672356401523100
    - HDispTIncome — Obs 317738; Mean 8383471523
    - MPShock — Obs 317738
  - Quartile dummies and indebtedness measures: TDNW, TDTA, TDNFA, TPNW, TPNFA, HNW, HNFA reported with Obs 317738 and quartile dummy structures.

*Source: wpiea2019076 — Section V (HILDA and IMF staff analysis as presented in the supplied content).*

### Section V investigates the role of household debt exposure in households’ consumption

### Section V investigates the role of household debt exposure in households’ consumption

### A primer on household debt and vulnerabilities
- Empirical literature emphasizes that it is not household debt per se but metrics such as rising leverage, excessive credit, or debt distribution that matter for macroeconomic analysis.
- High leverage combined with asset price shocks can lead to demand-driven recessions (Mian, Roa and Sufi, 2013); marginal effect of a decline in home value on tighter credit constraints is significantly larger for postal codes with a high housing leverage ratio.
- Evidence from Australia:
  - Atalay, Whelan, and Yates (2017): positive relationship between changes in housing wealth and consumption expenditure; wealth effect strongest for middle-aged homeowners; wealth effect became smaller post-GFC as households with higher loan-to-value ratios became more conservative in response to house-price changes.
  - Bilston, Johnson, and Read (2015): stress testing using HILDA data in the 2000s found households resilient to shocks because debt distribution remained concentrated among households well placed to service it.
  - Price, Beckers, and La Cava (2018): households typically cut back spending when they have higher levels of outstanding mortgage debt (debt overhang effect).
  - La Cava, Hughson, and Kaplan (2016): using HILDA for 2002-14 found a “borrower” cash flow channel; effect of interest-sensitive cash flows on spending particularly strong for liquidity-constrained households; central estimates indicate a 100 basis points cut in interest rates resulted in a 0.1 to 0.2 percent increase in household spending in aggregate.
  - La Cava and Price (2017): high debt relative to income and assets, and low debt-servicing capacity can reduce household spending growth relative to income; high-debt households more sensitive to income and housing equity shocks; effect stronger during adverse shocks.
- The composition of the household balance sheet explains heterogeneity in responses to monetary policy shocks; theoretical motivation includes Iacoviello (2005), Eggertsson and Krugman (2012), and Kaplan and others (2015).
- This paper follows Cloyne, Ferreira, and Surico (2018) methodology but uses longitudinal HILDA data, allowing direct use of wealth data rather than tenure proxies and grouping households by level of indebtedness for a more precise balance-sheet measure.

### Evolution and characteristics of household debt
- Household debt in Australia has been rising faster than household disposable income for the past three decades, resulting in one of the highest household debt ratios among advanced economies.
- Drivers:
  - Financial deregulation in the 1990s and historically low interest rates post-GFC reduced effective debt-service costs.
  - Housing boom driven by income and population growth, relatively inelastic supply, and expectations of future capital gains, encouraging investment demand for housing.
  - Meng, Hoang, and Siriwardana (2013): housing prices, GDP, and population positively affect household borrowing; interest rates, unemployment, number of new dwellings, and inflation negatively affect household debt.
- Key aggregate statistics and trends:
  - Housing debt at 140 percent of household disposable income accounted for about three-quarters of household debt outstanding as of September 2018.
  - Owner-occupied housing debt accounted for a relatively stable share of about one half of housing debt.
  - Other personal debt remained broadly stable at about 46 percent of household disposable income since 2000 (one quarter of household debt outstanding).
  - Household debt rose steadily to almost 190 percent of gross disposable income by September 2018.
  - Housing debt has remained below 20 percent of total housing assets since 2010, helped by a 42 percent rise in housing prices.
  - Household leverage (liabilities-to-net-worth ratio) trended down to close to 22 percent in 2017, but could increase rapidly if house prices corrected downward.

### Debt measurement issues
- Common scaling metrics:
  - Debt-to-income (DTI): total debt outstanding scaled by annualized gross household disposable income; useful for cross-time and cross-country comparison.
  - Debt-service-to-income (DSTI): focuses on size of debt service payments as influenced by interest rates and loan terms.
- Critique: proportional metrics like DTI or DSTI may not capture higher capacity for debt accumulation among higher-income households, because living expenses do not increase one-for-one with disposable income.
- Alternative scalings: assets (liquid, financial, non-financial) and net worth.
- Notable numeric observations:
  - Household debt approached 190 percent of gross disposable income by September 2018.
  - Housing debt remained below 20 percent of total housing assets since 2010.
  - Housing prices rose by 42 percent (period referenced in source).
  - Liabilities-to-net-worth ratio declined to close to 22 percent in 2017.

### Distribution of debt
- Distributional aspects matter for financial stability and policy: speed of accumulation, extent of leverage, and distribution across households.
- SIH 2015-16 findings (ABS):
  - Three quarters of households held debt.
  - Share of households with a debt-to-income ratio of three and above has risen across income quintiles.
  - Rise in median DTI was most pronounced in the top 40 percent of higher-income households.
  - DTI rose across all age groups; households with reference person 35-44 years had the highest median DTI of 2.3 in 2015-16.
  - SIH 2015-16 sampled 17,768 households over July 2015 to June 2016; SIH integrated with HES.
- HILDA longitudinal survey insights (comparisons 2002 vs 2014):
  - About 30 percent of respondent households indicated they had no debt in 2014.
  - About one-third of households had a DTI ratio of less than or equal to one in 2014.
  - Close to 20 percent of households had DTI above 3 in 2014.
  - About 27 percent of total households had a debt-to-net worth ratio of less than zero (net liability position).
  - Of the 73 percent with positive net worth, about three quarters had debt-to-net worth ratio between zero and one inclusive, and close to 15 percent had debt-to-net worth ratio between one and two inclusive.
  - Share of total household debt held by the top two disposable income quintiles increased from 70 percent in 2010 to 78 percent in 2014.
  - Household debt held by head of household 45 years or older rose from 38 to 47 percent during 2002-14.
  - By tenure in 2016: mortgagors held 77 percent of total household debt, outright owners 13 percent, renters 10 percent (broadly unchanged from previous years).
- Debt-level quartile statistics (2002-14):
  - Average debt level for high debt group: A$460 thousand.
  - Average debt level for medium-high debt group: A$274 thousand.
  - Average debt level for medium-low debt group: A$58 thousand.
  - Average debt level for low debt group: A$36 thousand.
  - Household net worth largest for medium-low debt group and smallest for high debt category.
- Age-group trends:
  - DTI ratios rose over 2002-14 across households defined by age of household head, with DTI ratios of the 30-40 and 40-50 age groups rising (specific DTI values for these age groups are provided in source figures).

*Source: wpiea2019076 - Section V investigates the role of household debt exposure in households’ consumption*

### 2.6 since 2010. The three-dimension

### 2.6 since 2010. The three-dimension

### Evolution of household debt by age and income
- 2002: peak household debt skewed toward higher age groups and higher income.
- 2006: younger borrowers at age 30-39 with medium income had higher household debt level largely because of the need to borrow more to afford higher house prices.
- 2014: the 40-49 age high-income households had the highest debt level.
- Source data: HILDA.

### Distributional patterns and stylized facts
- Rise in household debt relative to disposable income to historically high levels in close synchronization with upward trend in house prices and historical low interest rates.
- Micro-survey data indicate:
  - High debt is more prevalent among higher-income households, who also tend to be more exposed to risks from being over-indebted.
  - Debt exposure of lower-income and more vulnerable households (including higher age and retired households) has also increased over time.
- Conclusion: a sizeable share of households is vulnerable to interest rate changes and other shocks, impacting debt-service repayment and consumption.

### Debt-servicing dynamics and interest-rate environment
- Debt-servicing ratio (scheduled principal and interest mortgage repayments to household disposable income) remained broadly unchanged at 9 to 11 percent in the post-GFC period despite the rise in household debt levels.
- Mortgage interest to scheduled principal repayment ratio declined from 5.8 in 2008 to 1.4 in 2017.
- Households made higher-than-scheduled total payments by about 2 percent of household disposable income in 2017.

### Prudential measures and mortgage buffers
- Since end-2014, APRA requires for new mortgage lending:
  - an interest rate buffer of at least 2 percentage points above the effective variable rate applied for the term of the loan, and
  - a minimum floor assessment interest rate of at least 7 percent.
- Total household mortgage buffers were 18 percent of outstanding loan balances or around 2½ years of scheduled repayments at current interest rates in 2017.
- Distribution of buffers (RBA, 2018):
  - one-third of the outstanding owner-occupier mortgages had at least two years equivalent of buffer,
  - about one-quarter had a mortgage buffer of less than one month.
- Mortgage buffers (offset accounts and redraw facilities) have been rising as households used falling interest rates to pay down debt faster than required.

### Over-indebtedness and vulnerability
- Definition used: households are over-indebted if their debt is three or more times their income or/and 75 percent or more of the value of their assets.
- Under this definition, 29 percent of households are over-indebted based on the ratio of debt to either disposable income or assets.
- Distribution of over-indebtedness by income quintile:
  - top three income quintiles: one quarter of households in each quintile are over-indebted,
  - lowest and second income quintiles: about one-sixth are over-indebted.
- SIH results: about three quarters of over-indebted households did not have sufficient liquid assets to cover a quarter of their debt values.
- Staff estimates (2015-16 data): a one percentage point increase in interest rates would increase the debt service repayment (principal and interest) of over-indebted households in the highest (lowest) income quintile by 22 percent (18 percent).

### Data sources and sample
- Micro-level household data: HILDA survey, annual frequency from 2001 to 2016 (2017 wave released after analysis and not included).
- Macroeconomic variables: IMF’s International Financial Statistics (IFS) and World Economic Outlook (WEO) databases.
- HILDA contains micro data on household income, wealth, expenditure, housing tenure status, age, gender, and other characteristics; detailed information on weekly and annual expenditures, mortgage payments, household total debt, household total net worth, financial and non-financial assets, property-apportioned debt and equity, and house-apportioned debt and equity.
- Empirical sample: full HILDA survey for 2001-16 for about 16,000 participating households.
- Empirical panel: unbalanced panel, households as units, years for time dimension; up to 26,400 households over 16 years used in Arrelano-Bond dynamic panel regressions.

### Household grouping and variable construction
- Debt quartiles constructed using series of total debt, total property-apportioned debt, and home-apportioned debt.
- Three household groups:
  - high debt — the highest quartile;
  - medium debt — the two middle quartiles;
  - low debt — the lowest quartile.
- Consumption aggregation:
  - detailed consumption and expenditures aggregated into current consumption and durable expenditure panel series on annual basis;
  - both aggregated variables divided by total disposable income to normalize and detrend.
- Outlier treatment:
  - debt ratios: all outliers removed and all negative reported debt indicators removed;
  - current consumption ratios: negative observations removed and 99th percentile for current consumption-to-disposable-income ratio dropped;
  - durable expenditures ratios: first and 99th percentiles dropped (first percentile including zero and negative durable expenditures).
- Wealth survey frequency and assumption: wealth questionnaire every fourth year (starting 2002); study assumes household balance sheet position stays the same for the four years prior to the next wealth questionnaire.

### Monetary policy shock identification
- Approach: Romer and Romer (2004) narrative approach; estimates for Australia from Bishop and Tulip (2017) used.
- Notable shocks: one of the largest policy shocks occurred in 2008-09 (RBA cut the cash rate by more than suggested by historical relationships); contractionary shocks in 1994 were unusually large.
- By construction, the monetary policy shock variable is almost perfectly correlated with the RBA’s cash rate (correlation coefficient is 0.9987).
- Cash rate excluded from regressions because of collinearity with the monetary policy shock variable.

### Empirical framework and estimation
- Specification: regressions of current consumption ratio or durable expenditures ratio on lag of monetary policy shocks, controlling for lagged endogenous variable and other controls.
- Controls for current consumption ratio: household’s total disposable income, average annual inflation, average mortgage rate, and world GDP growth.
- Controls for durable expenditures: year-dummies (other specifications faced collinearity issues due to short time series per household).
- Estimator: Arellano-Bond dynamic panel regressions for unbalanced panel, robust two-step estimators of the variance-covariance matrix using one lag for the dependent variables.
- Note: Monetary policy shocks at t and (t-1) considered; final specification uses lagged monetary policy shock due to collinearity.

### Empirical results — effects of monetary policy shocks by debt group
- Hypothesis: a positive monetary policy shock (tighter financial conditions) has a negative impact on current consumption and durable expenditure, especially for high-debt households.
- Current consumption to disposable income:
  - Households react negatively and significantly to monetary policy shocks in the previous year for most debt-ratio groupings.
  - Households react positively (smaller coefficient) to current monetary policy shocks.
  - Current consumption depends significantly on previous-period consumption decisions.
  - Reaction differs by debt ratio type:
    - For total debt ratios, strongest combined reaction for high- and medium-debt households.
    - Low-debt households react positively to the monetary policy shock.
  - Quantitative example (grouping by ratio of total debt to total non-financial assets):
    - A monetary policy shock of one basis point would cause the total decrease in the current consumption ratio to total disposable income by 0.28 percentage points for high-debt households.
    - The same shock would decrease the ratio by 0.13 percentage points for medium-debt households.
    - The same shock would cause an overall increase by 0.19 percent for low-debt households.
- Durable expenditures to disposable income:
  - All coefficients of the lagged monetary policy shocks are negative and significant for high- and medium-debt households.
  - For most specifications, the response of high-debt households tends to be larger than that of medium-debt households.
- Robustness:
  - Results robust to inclusion of other macro variables (e.g., ratio of general government balance over GDP).
  - Inclusion of year dummies can reverse sign of monetary policy shock coefficients (possible collinearity and short time series effects).
  - Excluding years 2008 and 2009 as robustness check: previous results continue to hold.

*Source: HILDA and IMF staff calculations (content unit: wpiea2019076 - 2.6 since 2010. The three-dimension).*

### 0.3 percentage points, while medium-debt households would react by only 0.2 percentage

### wpiea2019076 - 0.3 percentage points, while medium-debt households would react by only 0.2 percentage

### Empirical findings on household responses to monetary policy shocks
- High-debt households reduce durable expenditures by 0.3 percentage points; medium-debt households reduce durable expenditures by 0.2 percentage points.
- Low-debt households do not respond significantly to monetary policy shocks for their durable expenditures.
- Interpretation: the income channel of monetary policy transmission may dominate the intertemporal substitution channel for low-debt households; being savers, they likely hold a higher quantity of interest-earning assets and can smooth consumption.
- For current consumption, behavior of all household quartiles drives the full-sample results.
- Younger households have higher debt, including property-related debt, and tend to have higher consumption and durable expenditures; these households would react faster to changes in policy rates.
- Robustness checks:
  - Introducing a fiscal policy variable: results hold.
  - Excluding years 2008 and 2009: results hold.

### Subgroup analysis (mortgagors, renters, outright owners)
- For current consumption:
  - Reaction to a monetary policy shock is most negative and statistically significant for mortgagors, followed by renters and outright owners.
- For durable expenditures:
  - Mortgagors’ reaction to monetary shocks is the most negative and statistically significant, followed by renters.
  - Outright owners’ reaction is positive but insignificant—income channel dominates intertemporal substitution for this group; they have enough interest-earning assets to avoid adjusting durable expenditures.

### Consistency with other studies
- Results are consistent with Cloyne, Ferreira, and Surico (2018) for the United States and the United Kingdom.
- Results align with La Cava, Hughson, and Kaplan (2016) for Australia: cash flow channel of monetary policy transmission is significant when net borrowers are used as a proxy for high-debt households.
- Aggregate-level finding: household spending increases with a reduction in interest rates, matching negatively-signed coefficients for monetary policy shocks in this study.

### Aggregate impact calculations and scenarios (contractual short-term 25bps monetary policy shock)
- Methodology:
  - One-period impact: use regression coefficients to calculate sensitivity of current consumption and durable expenditures ratios to aggregate household total disposable income under a 25 basis points monetary policy shock in period (t-1), holding everything else equal.
  - Current consumption depends on contemporaneous and past monetary policy shocks; durable expenditures depend only on a one-period lagged shock.
  - Contemporaneous shock sign in current consumption regressions is positive; lagged shock sign is negative and larger in absolute value, yielding a cumulative negative effect for full sample and households with high and medium debt, and a cumulative positive effect for low-debt households.
  - Long-run impact: take partial derivative of consumption ratio over cumulative monetary policy shock; long-run effects accumulate and differ by debt group.

- Numerical short-term and long-term impacts (as reported):
  - Using Short-Term Impact = (γ(t-1)+γ(t))*0.25 and Long-Term Impact = ((γ(t-1)+γ(t))/(1-β(t-1))):
    - Full Sample: Short-Term Impact = -0.0002; Long-Term Impact = -0.0009
    - High Debt: Short-Term Impact = -0.0004; Long-Term Impact = -0.0020
    - Medium Debt: Short-Term Impact = -0.0006; Long-Term Impact = -0.0026
    - Low Debt: Short-Term Impact = 0.0005; Long-Term Impact = 0.0022
  - Using Short-Term Impact = γ(t-1)*0.25 and Long-Term Impact = γ(t-1)/(1-β(t-1)):
    - Full Sample: Short-Term Impact = -0.0004; Long-Term Impact = -0.0018
    - High Debt: Short-Term Impact = -0.0008; Long-Term Impact = -0.0032
    - Medium Debt: Short-Term Impact = -0.0005; Long-Term Impact = -0.0024
    - Low Debt: Short-Term Impact = 0.0001; Long-Term Impact = 0.0004

- Qualitative implications from these scenarios:
  - One-period impact on aggregate ratios of current consumption and durable expenditures to total disposable income would be negative for the full sample and all subsamples except durable expenditures of low-debt households.
  - The impact on the ratio of current consumption is lower than that on durable expenditures for the full sample and high-debt households—those households adjust durable expenditures immediately but maintain certain categories of current consumption.
  - Some low-debt households would increase current consumption and durable expenditures after a contractionary monetary policy shock; low-debt households increase current consumption more than durable expenditures.
  - Long-run effects accumulate: positive for low-debt households; negative for full sample and households with high and medium debt.

### Textual analysis of RBA communications
- Goal: examine frequency of keywords related to household debt, housing prices, housing market, credit to housing sector, and mortgage rate in RBA communications on its website.
- Data and coverage:
  - Selected word phases include household debt, housing debt, household leverage, household debt-to-income.
  - RBA documents analyzed: Interest Rate Decisions (11 releases per year), RBA Board Minutes (11 per year), Statement on Monetary Policy (four times per year in February, May, August, and November), and speeches by the RBA Governor and senior management.
  - Period covered: 2008 to cut-off data as of December 12, 2018.
  - Python used for text mining and word-counting selected word phases.
- Findings:
  - RBA increased focus on household debt issues in communications as reflected in increased word counts for household debt word phases in 2015-17, with a decrease somewhat in 2018.
  - Word counts for housing prices and housing market picked up around 2012, coinciding with the start of the housing boom.
  - Word counts on credit to the housing sector increased sharply since (text cut off in source).

### Policy implications and RBA communication
- With recent increases in debt-to-income and debt-to-wealth ratios and a larger share of households with higher debt ratios, aggregate sensitivity of household spending should, in principle, have increased.
- This information may influence RBA’s policy rate decisions and its communication about monetary policy.
- The analysis suggests that over recent years the RBA should have been paying more attention to household debt and consumption in its monetary policy decisions and related communication.

*Sources: HILDA and IMF staff estimates.*

### 2015. On the other hand, the word count for mortgage rate also increased since 2005 but

### VII. CONCLUSIONS AND POLICY IMPLICATIONS

### Evolution and distribution of household debt
- High debt exposure is more prevalent among higher-income and higher-wealth households.
- Debt exposure of lower-income and more vulnerable households has increased over time, increasing exposure to risks from rising debt service.
- Presence of over-indebted households at both low- and higher-income quintiles suggests macro-financial risks have increased and require close monitoring.

### Debt service capacity and monetary policy risks
- Despite high debt levels, households’ debt service burden has remained manageable due to historically low mortgage interest rates.
- Financial institutions assess mortgage serviceability for new mortgage lending with interest rate buffers above the effective variable rate applied for the term of the loans.
- Downside risks remain: a sharp tightening of global financial conditions could spill over to higher domestic interest rates, posing risks to debt service capacity and consumption.

### Empirical findings on monetary policy transmission
- The paper uses the HILDA survey for 2001-16 and investigates transmission of monetary policy shocks to current consumption and durable expenditures across households with different debt-to-wealth ratios.
- Main empirical results:
  - Households with high debt tend to reduce their current consumption and durable expenditures relatively more than other households in response to a contractionary monetary policy shock.
  - Households with low debt may not respond to monetary policy shocks because they hold more interest-earning assets and can smooth consumption via higher interest income; for these households the income effect dominates the intertemporal substitution effect.
  - Given a larger share of high-debt households and their higher responsiveness, it may take a smaller increase in the cash rate for the RBA to achieve its policy objectives compared to past episodes of policy rate adjustments ("more bang for the buck").
  - By responding gradually, the RBA can still meet its mandates.

### Communication and market perceptions
- Textual analysis shows the RBA’s communication increasingly focused on the impact of household debt on monetary conditions and financial stability over the past decade, consistent with the rise in debt-to-income ratios.
- Markets have started to factor household debt into their assessment of monetary policy and market expectation analysis.
- Recommendation: continuing with a transparent and strengthened communication strategy on issues related to household debt and household consumption will further improve predictability and efficiency of monetary policy in Australia.

### Key empirical coefficients and statistics (selected estimates preserved exactly)
- Table 1 (Current Consumption and Monetary Policy Shocks; Total Debt to Total Net Worth, Full Sample):
  - L.ConsTDI 0.130***
  - MPShock 0.0153***
  - L.MPShock -0.0161***
  - Observations 168,028
  - Number of xwaveid 26,431
- Table 1 (examples across debt groups; MPShock and L.MPShock preserved exactly):
  - High Debt MPShock 0.0134***; L.MPShock -0.0151***
  - Medium Debt MPShock 0.0130***; L.MPShock -0.0153***
  - Low Debt MPShock 0.0196***; L.MPShock -0.0177***
- Table 2 (Durable Expenditures and Monetary Policy Shocks; Total Debt to Total Net Worth, Full Sample):
  - Durable Exp (t-1) 0.0948***
  - MP Shock (t-1) -0.00166***
  - Observations 43,874
  - Number of xwaveid 16,327
- Table 2 (by debt groups; MP Shock (t-1) examples preserved exactly):
  - High Debt MP Shock (t-1) -0.00306***
  - Medium Debt MP Shock (t-1) -0.00217***
  - Low Debt MP Shock (t-1) 0.000372
- Table 3 (Current Consumption and Monetary Policy Shocks by Ownership Groups; Full Sample):
  - Curr Cons (t-1) 0.130***
  - MP Shock 0.0153***
  - MPShock (t-1) -0.0161***
  - Observations 168,028
  - Number of xwaveid 26,431
- Table 4 (Durable Expenditures and Monetary Policy Shocks by Ownership Groups; Full Sample):
  - Durable Exp (t-1) 0.0948***
  - MP Shock (t-1) -0.00166***
  - Observations 43,874
  - Number of xwaveid 16,327

### Data and methodology notes
- Data source: Household, Income and Labour Dynamics in Australia (HILDA) panel covering waves 1-16 (2001-2016).
- Original HILDA combined datasets produced 2,441 variables with 317,738 observations; trimmed panel used 293 variables and 317,738 observations.
- Key data manipulations:
  - Derived seven household indebtedness ratios.
  - Excluded outliers by dropping the first and 99th percentiles and negative values.
  - Ratios split into quartiles and dummy variables assigned for indebtedness risk: 1=Low-risk (1st quartile), 2=Medium-low risk (2nd quartile), 3=Medium-high risk (3rd quartile), and 4=High risk (4th quartile).
  - Additional dummy for respondents whose indebtedness jumped from the 1st to the 4th quartile during any period within the 16 waves.
  - Wealth and debt variables collected every four years (2002–2014) and assumed unchanged until the next wealth survey.
- Macroeconomic indicators (nominal and real GDP, interest rates, inflation, house price index) downloaded from Haver Analytics; world growth from IMF, World Economic Outlook database.
- Monetary policy shock series based on James Bishop and Peter Tulip, “Anticipatory Monetary Policy and the Price Puzzle”, RDP-2017-02.

*Italic: Source: IMF staff analysis and HILDA survey data as presented in the supplied content.*

### Appendix Table 1. Data and Descriptive Statistics

### Appendix Table 1. Data and Descriptive Statistics

### Variables and labels (selected, as reported)
- xwaveid — Row id for the panel data; Obs 317738; Mean 3332033965041000011601187
- Year — Year; Obs 317738; Mean 2009; Std. Dev. 52001; Min 2016
- hhrhid — Household id; Mean 0
- MortTaken — Mortgage taken (1=took out institutional loan, 2 no institutional loan); Obs 216258; Mean 1; Std. Dev. 0; Min 1; Max 2
- PaidOffLoan — paid off (1=yes, 2=no); Obs 171699; Mean 2; Std. Dev. 0; Min 1; Max 2
- PaySchedule — Payment schedule of mortgage (1=ahead of time, 2=on time, 3=behind schedule); Obs 117939; Mean 2; Std. Dev. 1; Min 1; Max 3
- OtherLoan — Other loan (1=yes, 2=no); Obs 316608; Mean 1; Std. Dev. 1; Min 0; Max 2
- SecondMortEq — Second mortgage against equity (1=yes, 2=no); Obs 316632; Mean 1; Std. Dev. 1; Min 0; Max 2
- SecondMortRep — Amount paid for second mortgage (0=no second mortgage); Obs 196820; Mean 0; Std. Dev. 0; Min 0; Max 0
- HouseProvision — House provision (1=yes, e.g. house part of job compensation etc); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- RentW — Weekly rent received if rented; Obs 7214; Mean 2902; Std. Dev. 2403500
- ExpGroceryW — Weekly expences, groceries; Obs 212079; Mean 197; Std. Dev. 1040; Min 1250
- ExpFoodW — Weekly expences, food; Obs 212079; Mean 154; Std. Dev. 880; Min 1000
- MortOutst — Oustanding mortgage; Obs 117154; Mean 20596617234201854628
- Age — Age of persons in household; Obs 317738; Mean 36; Std. Dev. 23; Min 1; Max 101
- Gender — Gender of persons in household; Obs 317738; Mean 2; Std. Dev. 0; Min 1; Max 2
- HDebtTotal — Total household debt; Obs 76713; Mean 15696530940302888969
- HDebtOther — Other household debt; Obs 76713; Mean 6397371330869776
- HNetWorth — Household net worth; Obs 76713; Mean 6430461050509; Std. Dev. -48559808481406
- HNFinAssets — Household non-financial assets; Obs 76713; Mean 554967888434010800000
- HFinAssets — Household financial assets; Obs 76713; Mean 24547551563904948122
- HTotAssets — Household total assets; Obs 76713; Mean 8010871204960010600000
- HTPropAppDebt — Household wealth, total property, apportioned debt; Obs 76713; Mean 12445524967202620156
- HTPropAppEq — Household wealth, total property, apportioned equity; Obs 76713; Mean 344004613223; Std. Dev. -599927210500000
- HHomePropApp — Household wealth, home, apportioned equity; Obs 76713; Mean 253207371850; Std. Dev. -37750004122694
- HIndNFinA — Household indebtedness, HH total debt over HH non-financial assets; Obs 317738; Mean 16008278
- HIndTA — Household indebtedness, HH total debt over HH tota assets; Obs 317738; Mean 2109012500
- HIndProp1 — Household indebtedness, HH total property apportioned debt as share of HH net worth; Obs 317738; Mean 0; Std. Dev. -9601919
- HIndProp2 — Household indebtedness, HH total property apportioned debt as share of HH non-financial assets; Obs 317738; Mean 15908278
- HIndHome1 — Household Indebtedness, HH home apportioned debt as share of HH net worth; Obs 317738; Mean 0; Std. Dev. -6601469
- HIndHome2 — Household indebtedness, HH home apportioned debt as share of HH non-financial assets; Obs 317738; Mean 15908278
- HGrTotIncome — Household gross total income; Obs 317738; Mean 10243898763; Std. Dev. -20100001452536
- HDispTIncome — Household disposable total income; Obs 317738; Mean 8383471523; Std. Dev. -2010000937853
- ConsCurrA — Current consumption expenditure including alcohol; Obs 244586; Mean 292661553372436828
- ConsCurrNA — Current consumption expenditure excluding alcohol; Obs 244586; Mean 277791484972436828
- ExpDur — Household expenditure on durable goods; Obs 87364; Mean 112672356401523100
- GrIncBand — Gross income band (takes value 1-9 ...; 10-11 ...; 12 ...; and 13 ...); Obs 169874; Mean 8; Std. Dev. 3; Min 1; Max 13

### Quartile-based indebtedness measures and dummies
- TDNW — Household total debt over household net worth, quartiles; Obs 317738; Mean 2114
  - TDNW_DL — dummy for 1st quartile (low risk); Obs 317738; Mean 0; Std. Dev. 0; Min 1
  - TDNW_DML — dummy for 2nd quartile (medium-low risk); Obs 317738; Mean 0; Std. Dev. 0; Min 1
  - TDNW_DMH — dummy for 3rd quartile (medium-high risk); Obs 317738; Mean 0; Std. Dev. 0; Min 1
  - TDNW_DH — dummy for 4th quartile (high risk); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- TDTA — Household total debt as share of household total assets, quartiles; Obs 317738; Mean 2114
  - TDTA_DL; TDTA_DML; TDTA_DMH; TDTA_DH — each reported with Obs 317738 and Mean 0 or 1 per table
- TDNFA — Household total debt as share of household non-financial assets, quartiles; Obs 317738; Mean 2114
  - TDNFA_DL; TDNFA_DML; TDNFA_DMH; TDNFA_DH — each reported with Obs 317738 and Mean 0 or 1 per table
- TPNW, TPNFA — Household total property-apportioned debt as share of household net worth / non-financial assets, quartiles; Obs 317738; Mean 2114
  - TPNW_DL — Obs 317738; Mean 1; Std. Dev. 0; Min 1
  - TPNW_DML — Obs 317738; Mean 0; Std. Dev. 0; Min 0
  - TPNW_DMH; TPNW_DH — Obs 317738; Means 0 or 1 as reported
- HNW, HNFA — Household home-apportioned debt as share of household net worth / non-financial assets, quartiles; Obs 317738; Mean 2114
  - HNW_DL — Obs 317738; Mean 1; Std. Dev. 0; Min 1
  - HNW_DML — Obs 317738; Mean 0; Std. Dev. 0; Min 0
  - HNW_DMH; HNW_DH — Obs 317738; Means 0 or 1 as reported
  - HNFA_DL; HNFA_DML; HNFA_DMH; HNFA_DH — same reporting format

### Macro, trends, and control variables
- trend — Defined as Year minus 2000; Obs 317738; Mean 8850; Std. Dev. 15
- ptrend — Polynomial trend; Obs 317738; Mean 111416403615
- CPI — Consumer price index (2011Q3-2012Q2=100, Haver); Obs 317738; Mean 93; Std. Dev. 1175109
- InflEOP — Inflation, end-of-period; Obs 317738; Mean 3; Std. Dev. 114
- InflPA — Inflation, period average; Obs 317738; Mean 3; Std. Dev. 114
- HPI — House price index, existing homes (2011Q3-2012Q2=100, Haver); Obs 317738; Mean 932547137
- GDPNom — Nominal GDP (Haver); Obs 317738; Mean 12647803262297274941700118
- GDPRGDPcap — Real GDP per capita (Haver); Obs 317738; Mean 139772318483910825741678018
- GDPGrw — Real GDP growth (Haver); Obs 317738; Mean 57043109723752669726
- MortR — Standard variable mortgage rate, owner-occupier (end of period, %, Haver); Obs 317738; Mean 3124
- GGBalGDP — General government fiscal balance (%GDP, The Treasury); Obs 317738; Mean -22; Std. Dev. -52
- OISEOP — Overnight indexed swap rates, 3-months, end of period (Haver); Obs 317738; Mean 7159
- OISPA — Overnight indexed swap rates, 3-months, period average (Haver); Obs 317738; Mean 4227
- BBREOP — Bank accepted bill rates, 3-month (end of period, %, Haver); Obs 317738; Mean 4127
- BBRPA — Bank accepted bill rates, 3-month (period average, %, Haver); Obs 317738; Mean 4127
- MPShock — Monetary policy shocks 1/; Obs 317738; Mean 4127

### Consumption and tenure indicators
- ConsTDIa — Consumption expenditure as share of total disposable income (including alcohol); Obs 244272; Mean 1; Std. Dev. -5250924
- ConsTDI — Consumption expenditure as share of total disposable income (excluding alcohol); Obs 244272; Mean 0; Std. Dev. -4235873
- ExpDurTDI — Household expenditure on durable goods as share of total disposable income; Obs 87243; Mean 0; Std. Dev. -20457
- Ouright — Outright owner (1=outright owner, 0=otherwise); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- Mortgagor — Mortgagor (1=has mortgage, 0=otherwise); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- Renter — Renter (1=has rent, 0=otherwise); Obs 317738; Mean 0; Std. Dev. 0; Min 1

### International and additional variables
- US3mTB — US 3-month T-bill rate; Obs 317738; Mean 1205
- US1yTB — US 1-year T-bill rate; Obs 317738; Mean 1205
- WorldGr — World growth (World Economic Outlook database); Obs 317738; Mean 3; Std. Dev. -124
- LowIndebt1-6 — Low indebtedness (1 if TDNW (TDTA, TDNFA, TPNW, TPNFA, HNW, HNFA)=1, 0 otherwise); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- Change1-6 — HH moving from low to high indebtedness (1 if TDNW (TDTA, TDNFA, TPNW, TPNFA, HNW, HNFA) becomes 4 from 1 within waves); Obs 317738; Mean 0; Std. Dev. 0; Min 1
- HIndNWorth — Household indebtedness (HH total debt over HH net worth); Obs 317738; Mean 1; Std. Dev. -286337200
- IncQuintile — HH Gross total income, quintiles; Obs 317738; Mean 3; Std. Dev. 115
- AssQuintile — HH Gross total assets, quintiles; Obs 76713; Mean 3; Std. Dev. 115
- IncQuart — HH Disposable Income, quintiles; Obs 317738; Mean 3; Std. Dev. 115
- EDispTIncome — Enumerated person total disposable income; Obs 317738; Mean 2891241458; Std. Dev. -2010000886310
- HeadHH — Age of head of household; Obs 317738; Mean 45; Std. Dev. 153101

*Source: HILDA unless otherwise noted. 1/ “Anticipatory Monetary Policy and the Price Puzzle”, James Bishop and Peter Tulip (2017).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019076.pdf_