HANK - heterogeneous-agent New Keynesian - joins two traditions. Bewley (1986), Huggett (1993) and Aiyagari (1994) built incomplete-markets models in which households face uninsurable income risk and a borrowing limit, so wealth inequality and a whole distribution of marginal propensities to consume (MPCs) arise endogenously. The New Keynesian tradition (Woodford 2003, Gali 2015) supplied the sticky-price framework central banks use for policy analysis. Papers such as McKay, Nakamura and Steinsson (2016) and Kaplan, Moll and Violante (2018) put the two together; Kaplan, Moll and Violante's 'Monetary Policy According to HANK' showed that once wealth heterogeneity is realistic, the representative-agent Euler equation is a poor guide to how monetary policy reaches consumption. HANK-lite, as used on this page, is the analytically tractable branch of that literature - Bilbiie (2008, 2020), Challe, Matheron, Ragot and Rubio-Ramirez (2017), Ravn and Sterk (2017, 2021), Debortoli and Gali (2018), Acharya and Dogra (2020) - which shrinks the distribution to a few states so that aggregate demand is again one Euler equation, but with coefficients that carry the heterogeneity.
The core mechanism is MPC heterogeneity interacting with general equilibrium. Households with little liquid wealth spend a much larger share of a windfall within a quarter than households with ample liquidity (Johnson, Parker and Souleles 2006; Parker, Souleles, Johnson and McClelland 2013; Fagereng, Holm and Natvik 2021). A representative-agent model imposes one small MPC on everyone. Auclert (2019) decomposed the consumption response to a monetary shock into an aggregate income channel, three redistribution channels - earnings heterogeneity, Fisher, and unhedged interest rate exposure - each governed by a covariance between MPCs and exposure, and the intertemporal substitution channel. Kaplan, Moll and Violante (2018) found that in HANK the indirect, general-equilibrium effects through labor income account for most of the consumption response to a rate change, while the direct effect of the interest rate - nearly all of it in a representative-agent model - is small.
Auclert, Bardoczy, Rognlie and Straub (2021) developed the sequence-space Jacobian (SSJ) method that makes full HANK models fast to solve and estimate. The tractable branch answers a different question: when does heterogeneity matter, and in which direction? In Bilbiie's (2020) 'New Keynesian cross', a monetary shock is amplified relative to the representative agent exactly when the income of hand-to-mouth households is more cyclical than average (countercyclical inequality), and with idiosyncratic risk the same condition makes the aggregate Euler equation compound rather than discount news about future policy. Whether HANK dampens the forward-guidance puzzle, as McKay, Nakamura and Steinsson (2016) found in their calibration, or aggravates it therefore turns on how income risk moves with the cycle (Werning 2015; Bilbiie 2020; Acharya and Dogra 2020). Ravn and Sterk (2017, 2021) showed that countercyclical unemployment risk triggers precautionary saving in downturns and so amplifies demand-driven fluctuations.
The Graph tab solves the tractable two-state model. A share λ of households is hand-to-mouth, holding no liquid wealth and consuming its income; the rest are savers, who each quarter face a probability 1−s of becoming hand-to-mouth; and χ measures how strongly hand-to-mouth income moves with aggregate income. Three numbers carry all the heterogeneity, and every response has a closed form. The Proof tab derives that model first, then shows how the same objects - MPCs, income incidence, the power of forward guidance - appear in full two-asset HANK, where the distribution is solved on a grid and the Jacobians are computed numerically.