HANK-lite Model
A lighter heterogeneous-agent model that keeps liquid and illiquid balance-sheet channels visible.
Graph workspace
Distributional impulse response panel
HANK's defining claim is that households with different wealth respond differently to the same shock, and that this changes the aggregate. The panel shows both: consumption of the poorest and richest wealth quintiles side by side, and every variable drawn against the representative-agent response to the same shock.
Current results
Amplification vs. representative agent
Impact output response divided by the representative-agent response to the same shock (the dashed line). For the monetary shock it exceeds 1 exactly when . Undefined (--) for the transfer, which leaves the representative agent unmoved.
1.08
ratio
Hand-to-mouth share of the response
Share of the impact change in aggregate consumption made by hand-to-mouth households, who respond only to income - the indirect, general-equilibrium channel. It equals for the monetary and forward-guidance shocks and exceeds 100 for the transfer, because savers cut spending. Zero with a representative agent.
37.50%
percent
Q1/Q5 response ratio
Impact consumption response of the poorest quintile over the richest. For the monetary and forward-guidance shocks it is once , above 1 under countercyclical inequality. Negative when the two move in opposite directions, as under the transfer. Exactly 1 with a representative agent.
1.40
ratio
Forward-guidance power (8 quarters ahead)
Date-0 output response to a real-rate cut announced for quarter 8, relative to the representative agent, computed whichever shock is traced. It is below 1 exactly when : heterogeneity then weakens forward guidance, and with risk () the more so the further ahead the news. Above 1 it strengthens it and makes the forward-guidance puzzle worse.
1.27
ratio
Controls
Hand-to-mouth share: 0.3
The share of households with no savings to fall back on, who spend whatever income they get.
With , a higher share raises the cross multiplier and , so output and Q1 consumption fall further after a tightening and the amplification readout rises, while each saver bears a smaller share of the loss. With heterogeneity dampens instead, and more so the higher the share.
Cyclicality of hand-to-mouth income: 1.25
How strongly the incomes of cash-strapped households move with the economy; above 1 they swing more than average.
Above 1, heterogeneity amplifies a tightening relative to the dashed representative-agent path and Q1 moves more than Q5; at exactly 1 the aggregate paths coincide with the benchmark; below 1 heterogeneity dampens and Q1 moves less.
Idiosyncratic risk (saver to hand-to-mouth probability): 0.04
The quarterly chance that a household with savings ends up with none next quarter, the risk savers insure against.
With , more risk raises above 1, so the aggregate Euler equation compounds the future: the forward-guidance panels and readout move most, while a short-lived surprise barely changes, because risk acts only through expected future income. With it has no effect on the aggregate paths.
Taylor rule inflation response: 1.5
How strongly the central bank pushes back on inflation; below about 1 the model has no unique outcome.
Higher shrinks the output and inflation responses to the monetary shock and the transfer. Below the determinacy threshold nothing is drawn. It does not change the forward-guidance experiment, which fixes the real rate until quarter 8.
Shock persistence: 0.5
How slowly the rate hike or the transfer fades away, quarter by quarter.
Higher persistence makes every response last longer. A persistent tightening also lowers inflation by more, so the rule offsets more of it: at the default calibration the amplification readout falls from 1.09 at to 1.01 at .
Shock traced: 0
Pick the disturbance: a rate rise, a cash transfer to struggling households, or a promised future rate cut.
The tightening contracts output and both quintiles; the transfer raises output and Q1 consumption while Q5 savers, who pay for it, cut theirs and the benchmark stays at zero; forward guidance raises output at once, by more than the benchmark when .