Research

We describe micro-level data from the Canadian Survey of Consumer Expectations through an heterogeneous-expectation model to study the state-dependent risk of  inflation expectations unanchoring in low and high-inflation environments. We interpret the degree of mean-reversion in inflation expectations as an measure  of anchoring, which varies over time with the shares of respondents using trend-chasing or mean-reverting autoregressive forecasting models.  We find that during the post-pandemic inflation spike, trend-chasing expectations surged,  resulting in a heightened risk of unanchoring expectations and entrenching above-target inflation. Furthermore, forming trend-chasing inflation expectations is associated with higher expectations  for other key economic variables -- such as  interest rates, wages, and house prices --  and a  restraint in household spending. We provide additional new insights into household expectation formation, documenting that forecasting behaviors, attention, and noise in beliefs vary across socio-demographic groups and correlate with views about monetary policy.

Presentations

• Brown Bag Series, Department of Economics at University of Ottawa, Canada 2023

• Brown Bag Series, Bank of Canada, Canada 2024 (presented by co-authors)

Price-level targeting (PLT) is optimal under the full information rational expectations (FIRE) benchmark but lacks empirical support. We utilize a laboratory experiment to collect data on expectations, inflation and output dynamics under a traditional inflation targeting (IT) framework and a PLT regime with both deflationary and cost-push shocks. We are then able to faithfully reproduce the dynamics observed in the lab economies with a micro-founded heterogeneous expectation New Keynesian (HENK) model that can account for forward-looking communication. We do not observe the benefits of PLT over an IT regime because both human subjects and HENK agents are unable to learn the history-dependent inflation pattern under PLT, which results in excess macroeconomic volatility. However, once augmented with an inflation guidance consistent with closing the price gap, the advantage of PLT materializes, both in the lab and in the model, and the inflation gap is minimized no matter the type of shocks. The model offers a behavioral description of the cross-regime differences observed in the experiment and the lab data provide an empirical validation to the model. The combination of these two methods is a promising addition to the central banks’ toolkit since it can easily provide a first test of their policies.

Presentations

• CIREQ's 1st Interdisciplinary Conference on Big Data and Artificial Intelligence, McGill University, Canada 2023

• (Accepted) World Economic Science Association Meetings in Lyon, France 2023

• 56th Canadian Economics Association Conference  in Ottawa, Canada 2022 

• 1st Ph.D. Conference on Expectations in Macroeconomics, Barcelona School of Economics,  Virtual 2022

• 15th Vietnam Economist Annual Meeting in Ha Noi, Vietnam 2022

• Brown Bag Series, Department of Economics at Simon Fraser University, Canada 2022

• 48th Atlantic Canada Economics Association in Halifax, Canada 2022

We develop a behavioral model of inflation which  contains an indicator that quantifies the expectations unanchoring risk  over the business cycle.  We estimate this model using US and Canadian inflation and output gap data. We find that during the post-pandemic inflation surge, the macroeconomic data are  compatible with non mean-reverting expectations, which reveals an elevated, albeit transitory, risk of  expectations drifting away from the target. This risk has fully subsided in Canada but has not yet entirely dissipated in the United States. We find no evidence of such risk on the downside,  despite prevailing policy narratives in the 2010s decade.

Concerning an environment where initial information plays a key role in shaping economic expectations, we build a Bayesian updating model where agents selectively translate information to the formation of expectations through a Thompson-Sampling exploration-exploitation scheme. In this setup, we vary the initial information set available to agents while keeping their cognitive capacity and learning algorithm constant; which allows us to isolate the impact of initial information on expectation formation. Our model build on the lab experiment environment in Mirdamadi & Petersen (2018) and complement its laboratory evidence with a behavioral narrative of how different macroeconomic literacy communications affect the expectation making process. Our key findings show that precise quantitative information induces model-consistent forecasting by encouraging agents to exploit the readily available information. Conversely, for non-numerical qualitative information treatment, this vaguer information environment insufficiently induce agents to exploit the provided initial information; agents consequently over-explore the forecasting strategy space, which lead to more complex but more prone to misperceived forecasting strategy. Interestingly, minimal or no initial information induces agents to adopt simplified forecasting models, which can be more stabilizing. These findings suggest that central banks' outreach efforts to the general public may lead to unintended costs of rising uncertainty. Importantly, we do not discount the positive effect of high financial and economic literacy. Our results underscore the need for caution when central banks choose to engage the general public in efforts to enhance economic literacy.

Presentations

• Barcelona School of Economics Summer Forum, Computational and Experimental Economics Workshop, Spain 2025.

Is gender identity binary or nonbinary? My analysis shows that while both are possible, the latter is a more attracting equilibrium under an adaptive learning perspective. I frame the gender identity problem as a modified \textit{battle of the sexes} game, where individuals define their gender identity under pairwise matching motives. From a baseline game-theoretical standpoint, I demonstrate that the binary-only world and the nonbinary-only world are both Nash equilibria in the stage game and are locally stable in the infinitely repeated game. Thus, any state of gender identity could theoretically persist. I then adopt a genetic learning algorithm as an equilibrium selection criterion to investigate evolutionary dynamics further and provide a rationale for the transition from binary to nonbinary gender identity. Specifically, in a binary-origin world, divergence occurs as individuals identifying as nonbinary gender evolve to become the majority due to their higher flexibility in matching outcomes. My framework captures how adaptive learning drives identity evolution, offering a parsimonious tool to analyze how diversity and exclusivity emerge in varying economic environments.