Together, CPHS and AIDIS provide income, spending and a detailed balance sheet.
CPHS
Households over time
Observed income Observed consumption
+ Predicted assets and debts
↔Common household characteristics
AIDIS
The balance sheet
Observed assets Observed debts
+ Predicted income
Prediction across complementary surveys, not linkage of the same households. Performance is evaluated on held-out households.
Testing the predictions
Held-out households show how well the models recover income and wealth.
Three balance-sheet groups
Hand-to-mouth means having limited liquidity relative to income.
○
Poor HtM
Limited liquidity
Nonpositive net illiquid wealth
△
Wealthy HtM
Limited liquidity
Positive net illiquid wealth
□
Non-HtM
Sufficient liquidity under the classification rule
HtM = hand-to-mouth. “Poor” and “wealthy” identify balance-sheet groups, not income ranks. Framework: Kaplan, Violante and Weidner (2014).
Who has gold-collateral capacity?
Separate wealthy HtM households by whether gold can bridge their liquid shortfall.
Gold changes the classification
Counting gold as liquid expands the group with a usable spending buffer.
How much gold is pledgeable?
The capacity test depends on the share of gold value available to borrow against.
Housing value and housing quality
Housing adjustments change the poor–wealthy split within hand-to-mouth households.
The preferred combined definition
The proposed grouping combines housing quality with collateralizable gold.
Gold illiquid + ₹100,000 housing screen + Two of three infrastructure markers + 75% gold capacity
New poor HtM
Original poor
+ Low-infrastructure wealthy HtM
+ High-infrastructure wealthy HtM without gold capacity
New wealthy HtM
High-infrastructure wealthy HtM with gold capacity
Non-HtM remains a separate group.
Matching joint estimates have not yet been saved. The previous exhibits show completed, separately defined comparisons.
Who are these households?
Income, portfolios and demographics give the groups economic context.
Income shocks and consumption responses
Temporary income changes and lasting changes can produce different spending responses.
Temporary shock
The income innovation fades.
Permanent shock
The continuing income level changes.
Elasticity % consumption change per 1% income change
MPC ₹ consumption change per ₹1 income change
The paths are conceptual illustrations. The annual estimates preserve each method’s units; no log coefficient is converted in this deck.
Annual consumption responses
Compare responses across household balance sheets.
Three annual estimation approaches
Each approach uses a different feature of the income and consumption panel.
BPP
Fit income and consumption second moments to separate permanent and transitory innovations.
Commault
Use future income growth to identify the response to a temporary income change.
Crawley–Kuchler
Compare consumption and income changes across saved change lengths.
Use the previous slide’s controls to compare saved annual specifications. Group rankings depend on the method and specification.
A concrete income event: job loss
Patterson’s unemployment design connects lost earnings to household spending.
Previously employed wage worker
→
Observed unemployment
First stage · π
Change in the worker’s earnings
Reduced form · ρ
Change in household spending
IV response = ρ / π
Spending change per rupee of earnings change
Adapted from Patterson (2023), Equation (5), pp. 991–992. Her PSID design uses two-year changes; the CPHS exercise uses observed spending and interview-based windows.
The CPHS interview windows
Compare complete financial windows around the observed employment transition.
Pre-period sum−8 to −5
Between interviews−4 to 0
Post-period sum+1 to +4
Earlier interview: employedLater interview: unemployed, willing and looking for work
The event is observed between interviews; the exact loss date is not observed.
Months are relative to the later interview. Outcomes are four-month sums in constant-2014 rupees. Prior wage workers are aged 25–62; eligible comparison observations need not all remain employed.
Spending and wages after job loss
Relate the change in household spending to the change in the worker’s wages.
Job-loss responses across groups
Keep each group’s estimate and saved uncertainty visible.
Income and spending around job loss
Monthly profiles show the timing of the income and spending changes.
Wealth composition and financial constraints
01Gold changes who is classified as constrained. Liquid-gold and collateral-capacity definitions reveal different ways to access wealth.
02Consumption responses depend on the exercise. Annual income innovations and observed job loss provide distinct comparisons.
03Complementary surveys connect balance sheets to household spending. Predictions, classifications and response estimates retain their own samples.
2019-classified households with someone aged 25–64. Positive income and consumption; CPI-deflated values; calendar-month 1st–99th percentile trimming of ratios to mean real monthly income. Annual observations require at least eight months. Estimation is unweighted, with saved household-clustered uncertainty.
Job-loss estimation sample
Prior wage workers aged 25–62; complete eight-month financial coverage; baseline zeros retained. Unweighted IV controls for age, age squared, sex, prior wages and state-by-endpoint-month effects. Uncertainty clusters by household-family spell. Group comparisons start in 2022 and use fixed 2019 classifications.
Classification populations and uncertainty
Shares are survey-weighted percentages of classification populations. Annual chart lines show one saved standard error; job-loss charts use saved 95% confidence intervals. A different sample, definition or unit is always labeled separately.