Tina Koziol

Comparable monetary expansions produced radically different inflation outcomes after the Global Financial Crisis and after 2020. The dominant explanations — supply shocks, fiscal support, expectations de-anchoring — treat the divergence as contingent on the shocks themselves. This paper offers a structural alternative: the two episodes are reconciled once housing is placed at the centre of the transmission mechanism and CPI shelter measurement is recognised as a symmetric lag operator that understates inflation on the way up and overstates disinflation on the way down.

The core contribution is a 5×2 transmission taxonomy: five macro-financial channels (credit supply, collateral valuations, rental-market pass-through, wealth effects, and mortgage cash-flow effects) mapped across two inflation regimes (post-GFC and post-2020). The taxonomy is organised around four institutional primitives that vary across advanced economies and jointly determine which channels fire in which regime: (i) CPI shelter construction — rental-equivalence vs. actual rents vs. house prices, (ii) mortgage-contract structure — fixed-rate vs. floating, recourse vs. non-recourse, prepayment penalties, (iii) housing supply elasticity, and (iv) outright homeownership rate.

Using a harmonised panel of twenty advanced economies, I show that cross-country heterogeneity in inflation outcomes is driven primarily by these four institutional features rather than by shock composition. The framework recasts cross-country divergence as structural rather than contingent, identifies which channels can be reactivated by policy, and produces testable predictions for the next expansion. A companion SUERF Policy Brief summarises the implications for inflation-targeting regimes.

This paper develops an agent-based model of a three-population economy — workers, capitalists, and a central bank — in which capitalists hold heterogeneous reference returns: subjective anchors for what they believe they "should" be earning on financial wealth. Reference returns adapt sluggishly to realised performance, combining habit formation with psychological anchoring. The gap between reference and realised returns drives portfolio behaviour: when realised falls below reference, capitalists reach for yield, compressing risk premia and shifting demand toward riskier assets.

The novelty lies in two coupled mechanisms. First, reference returns are heterogeneous across the capitalist population, reflecting differences in experience and exposure. Second, beliefs about achievable returns spread through a social-network contagion process: as neighbours reach for yield and realise higher returns, peer reference returns drift upward, amplifying the aggregate response and generating endogenous feedback loops. The methodology builds on the Brock–Hommes tradition of heterogeneous-expectation models and the LeBaron line of agent-based financial modelling, extended with explicit network-contagion dynamics.

The central theoretical contribution is a political-economy channel from wealth-weighted capitalist preferences to the central bank's policy stance. Because capitalists hold the bulk of financial wealth and their reference returns drift with realised performance, prolonged periods of low realised returns generate persistent political pressure for accommodative policy — institutionalised through asset-price stability concerns, financial-stability mandates, and lobbying. The central bank's endogenous response further validates elevated reference returns, closing the loop.

Preliminary simulations reproduce several empirical regularities of low-rate environments: persistent demand for higher-yielding assets, narrowing risk premia across asset classes, endogenous regime shifts between low- and high-volatility states, and asymmetric responses to tightening vs. easing. The framework offers a mechanism for the observed secular drift in monetary policy stance across advanced economies that does not rely on exogenous shifts in natural rates or policy preferences.

We develop an agent-based epidemiological model in which agents are nodes in a network, their physical interactions are time-varying edges, and each agent carries an epidemiological state drawn from the Susceptible–Exposed–Infected–Recovered (SEIR) compartments. Into this setting we introduce naive social learning in the sense of DeGroot (1974): agents form subjective beliefs about the prevalence and severity of the disease by repeatedly averaging signals received from their neighbours, and adjust their protective behaviour accordingly.

The model is calibrated to detailed data for Cape Town, South Africa. Incorporating social learning materially improves fit: the best-fit squared difference between modelled and observed excess fatalities drops from 19.34 to 11.40. More importantly, the qualitative behaviour differs from standard compartmental models — social learning both flattens and shortens the curves for infections, hospitalisations, and excess fatalities, whereas contact-rate or transmission-probability reductions flatten the curve at the cost of lengthening the epidemic.

The paper contributes a tractable framework for embedding heterogeneous belief formation into epidemic models and shows that the behavioural response to information is a first-order determinant of epidemic dynamics — separate from, and sometimes substituting for, formal non-pharmaceutical interventions.