Two-asset HANK
US HANK
Shows how a stylised US shock moves through a household wealth distribution, and which households drive the response.
Good for
- How a shock transmits: monetary, fiscal spending, productivity
- MPCs by liquid wealth and the hand-to-mouth share
- Wage and labour paths for the incidence bridge (experimental)
Not for
- Scoring statute:
score_reformrefuses it; PolicyEngine microsimulation costs it - Magnitudes to quote: one published response is matched, the rest are unchecked (limits)
- US spending behaviour: its MPCs sit below US estimates (why)
A worked example
An easing shock to the Taylor-rule intercept: −0.0025 per quarter, ρ 0.61. Changes from steady state.
Monetary easing: Taylor-rule intercept shock of −0.0025
- The quarterly policy rate falls about 13bp on impact, not 25bp: the rule sets it, and its response depends on persistence (how)
- Quarterly inflation rises 0.08pp; the labour tax rate falls 0.63pp on impact as debt service drops
- Output, consumption, investment and inflation peak on impact, with no hump (limit)
- The by-household split is imposed and nearly flat across total-wealth quartiles (why)
- Regression values of this implementation, not a published result (pinned test)
How far to trust it
Three checks, weakest included; each opens its evidence.
- Solved steady-state parametersTable B.III, largest gap 4.2e−4;
vphiafter a wage-markup conversion6 of 7 direct, 1 after conversionpass - Impulse-response magnitudesThe one the paper states: impact consumption after a −5pp monetary shock (Fig. 7a), units inferred; all others regression-pinned only1.485% against 1.46%partial
- Quarterly MPC out of a rebateMPCs for $500 / $1,000 / $1,500 rebates per household (refined grid), against 0.15–0.25 in US rebate studies0.104 / 0.091 / 0.083fail
Read signs and mechanisms, not magnitudes. The hand-to-mouth share, counted by liquid wealth, is also above US estimates, while few households spend much of a rebate (both numbers).
Run it
The worked example above, three ways. State ρ 0.61: the hosted default is 0.9.
In the US HANK model, apply a monetary easing shock of −0.0025 to the Taylor-rule intercept with persistence 0.61, and show output, consumption and investment with the split by wealth quartile.The hosted server calls hank_shock; no install. Connect
pe-macro hank-shock --kind monetary --size -0.0025 --persistence 0.61 --distributionfrom us_hank import model
ss = model.solve_steady_state(); G = model.solve_jacobian(ss, T=300)
irf = model.shock("monetary", size=-0.0025, persistence=0.61, ss=ss, G_jac=G)
print(irf["Y"][:8])With the microsimulation
Two distributional cuts of the same shock: the model’s and the microsimulation’s.
hank_shock_incidence scales US employment income with the model’s wage and labour responses; the microsimulation reports the change after taxes, SNAP and the EITC, decile by decile (every join in the suite).
The distribution is calibrated, not matched to US data: its hand-to-mouth share and its MPCs both miss US estimates (limits).
Model cardinputs, outputs, horizon, data, access
- Answers
- How a stylised US shock transmits through a household wealth distribution, and which households drive the response
- Inputs
- A stylised shock: kind (monetary, fiscal spending, productivity), size, persistence
- Outputs
- GDP, consumption, investment, inflation, the ex-post real return; MPCs by liquid-wealth quartile (out of a windfall of 25% of mean quarterly post-tax labour income) and the hand-to-mouth share
- Horizon
- Quarterly impulse responses, typically 5 years, around a calibrated steady state
- Data
- Auclert, Bardóczy, Rognlie and Straub (2021); no data vintage
- Uncertainty
- None quantified: deterministic, first-order responses
- Access
- Hosted stylised shocks, CLI, Python package (Code)
- Runtime
- About 10 seconds cold on a laptop (steady state plus Jacobian, then cached); instant warm
- Joins the microsimulation
- Wage and labour responses scale US employment income (the join above, experimental)
- Status
- Partial replication of Auclert et al. (2021): six of seven solved parameters match directly and the seventh (vphi) after a wage-markup convention conversion; steady-state targets are inputs; shock signs are tested; one published magnitude (impact consumption after a monetary shock, units inferred) is matched within about 2%, the others are not checked; not a forecaster; distributional outputs are steady-state statistics and an MPC-proportional split of the aggregate
- Cannot answer
- Forecasts; detailed tax reforms (the labour tax is endogenous); PolicyEngine reforms; nonlinear or state-dependent dynamics