Karan Bhasin

This paper introduces a hierarchical Large Language Model (LLM) framework for the automated identification of narrative fiscal shocks. We develop a multi-stage architecture to extract austerity episodes from IMF Article IV reports (2004–2020) for 17 OECD countries. Relative to manual coding, our approach improves replicability and auditability by generating a documented sequence of classification steps. Benchmarking against Adler et al. (2024), we find that the LLM-based classification aligns closely with the narrative benchmark, while differing on a small subset of episodes. Local projection estimates indicate that LLM-identified shocks are associated with smaller estimated multipliers than the narrative benchmark, with the difference linked in large part to differences in shock persistence and endogeneity.

We examine the impact of formal adoption of inflation targeting (IT) on inflation, growth and anchoring of inflation expectations in advanced economies and emerging markets and developing economies. We find that while the early adopters of IT (pre-2000) saw declines in inflation rates following adoption, IT adopters since then have enjoyed such success in only about half the cases.

Our country-level analysis—using the synthetic control method—shows that IT adoption delivers significant inflation gains in about a third of the cases. We find limited support for the concern that adoption of IT systematically leads to poorer growth outcomes.

This paper estimates macroeconomic uncertainty using density forecasts from the Reserve Bank of India’s Survey of Professional Forecasters (2008–2023). These forecasts enable direct measurement of unobservable shocks to uncertainty in real-time as the first difference in the second moment of the densities. In addition, we implement a bin-by-bin calibration test based on the prequential principle. We report four key findings: (i) macroeconomic uncertainty in India has been on a decline since 2008; (ii) shocks to uncertainty derived from density forecasts compare favorably with other popular measures, viz. Economic Policy Uncertainty and VIX; (iii) uncertainty is affected primarily by negative news and is variance rational, and (iv) it captures demand shocks even after controlling for global uncertainty shocks, in contrast to EPU and VIX, which are primarily driven by supply shocks. Distinguishing these shocks is crucial for optimal monetary policy.

The Purchasing Managers' Index (PMI), a monthly economic indicator published by the Institute for Supply Management (ISM), is closely monitored for its timeliness and nowcasting value for the U.S. economy. Regarded as helpful by supply management professionals, business executives and policy makers, ISM has provided a breakeven point for PMI of around 42.5 for real GDP ‘growth-at-risk’. Effective January 2026, the latter threshold has been abruptly increased to 47.5. To prevent potential misinterpretation of the PMI-suggested breakeven points, we investigate their temporal stability and their implications for forecasting. This paper estimates the time-varying PMI breakeven points during 1960 - 2024 for the manufacturing sector and as well as for the overall economy. The results indicate that the empirical PMI thresholds deviate significantly from ISM's proposed thresholds and are highly cyclical depending on financial conditions. Moreover, we show that alternative thresholds use the same underlying information in PMI but imply different combinations of hit rate and false alarms. They are different operating points on the same Receiver Operating Characteristic curve. By publishing one breakeven point without making explicit the implied tradeoff, ISM risks misleading users into treating it as appropriate for every decision.

How much do the output costs of fiscal consolidation differ across economic conditions? We study narrative consolidations in 17 OECD economies using selection-adjusted local projections. This issue has been examined in the literature using coefficients on smooth transition functions. We show that they describe fitted responses at extreme economic states rather than the average responses over observed conditions. As a result, they do not directly answer how the average output costs of consolidation differ across economic states. We address this gap by directly estimating treatment effects by incorporating a smooth transition function within the Augmented-Inverse-Propensity-Weight Local Projections framework. Averaging the conventional fitted responses over observed conditions makes their state differences roughly one-third smaller. For interest rates, these fitted-response averages are close to the selection-adjusted estimates. Recession–expansion differences are small. The baseline output loss is 1.32 and 1.40 percentage points larger in low-rate conditions after two and three years, respectively, although the rate comparison is sensitive to the sample period.

Leading economic indicators are often interpreted through breakeven points, that is, the value of an indicator at which growth turns negative. Yet there is little guidance on how such breakeven points should be estimated when the relationship between an indicator and economic activity is time-varying and affected by omitted factors. We study this problem using the Purchasing Managers' Index (PMI), for which the Institute for Supply Management (ISM) reports an explicit breakeven-point rule. Using U.S. real-time data and partially linear time-varying quantile regression, we estimate PMI breakeven points for Industrial Production and GDP growth. For Industrial Production, the median breakeven point remains close to the expected value of 50 and is relatively stable over time. For GDP, the median breakeven point is materially below 50 and varies substantially over macroeconomic regimes. Estimating breakeven points across the conditional distribution of growth shows that the same PMI reading can be associated with very different growth outcomes depending on the underlying state of the economy. We further show that part of this drift reflects observable state variables such as financial conditions and macroeconomic uncertainty.

This paper explores the behavioral foundations of household inflation expectations. Using data from the Inflation Expectations Survey of Households (IESH), we examine the commonly observed bias in reported expectations in an environment characterized by recurring inflation surges and shifts in monetary policy regimes. The bias is partly driven by dispersion and skewness in households' price experiences, as they tend to overweight recent changes in salient items like food while paying limited attention to official inflation data. In such circumstances, achieving price stability is a necessary precondition for effective central bank communication, especially in low-income contexts.

The paper presents estimates of poverty [extreme poverty PPP$1.9 and PPP$3.2] and consumption inequality in India for each of the years 2004-5 through the pandemic year 2020-21. These estimates include, for the first time, the effect of in-kind food subsides on poverty and inequality. Extreme poverty was as low as 0.8 percent in the pre-pandemic year 2019, and food transfers were instrumental in ensuring that it remained at that low level in pandemic year 2020. Post-food subsidy inequality at .294 is now very close to its lowest level 0.284 observed in 1993/94.