You can also browse my Google Scholar profile.
2024
Approximating Auction Equilibria with Reinforcement Learning
Traditional methods for computing equilibria in auctions become computationally intractable as auction complexity increases, particularly in multi-item and dynamic auctions. This paper introduces a self-play based reinforcement learning approach that employs advanced algorithms such as Proximal Policy Optimization and Neural Fictitious Self-Play to approximate Bayes-Nash equilibria. This framework allows for continuous action spaces, high-dimensional information states, and delayed payoffs. Through self-play, these algorithms can learn robust and near-optimal bidding strategies in auctions with known equilibria, including those with symmetric and asymmetric valuations, private and interdependent values, and multi-round auctions.
A Deep Learning Approach to Heterogeneous Consumer Aesthetics in Retail Fashion
In some markets, the visual appearance of a product matters a lot. This paper investigates consumer transactions from a major fashion retailer, focusing on consumer aesthetics. Pretrained multimodal models convert images and text descriptions into high-dimensional embeddings. The value of these embeddings is verified both empirically and by their ability to segment the product space. A discrete choice model is used to decompose the distinct drivers of consumer choice: price, visual aesthetics, descriptive details, and seasonal variations. Consumers are allowed to differ in their preferences over these factors, both through observed variation in demographics and allowing for unobserved types. Estimation and inference employ automatic differentiation and GPUs, making it scalable and portable. The model reveals significant differences in price sensitivity and aesthetic preferences across consumers. The model is validated by its ability to predict the relative success of new designs and purchase patterns.
2023
Designing Auctions when Algorithms Learn to Bid: The Critical Role of Payment Rules
This paper examines the impact of different payment rules on efficiency when algorithms learn to bid. We use a fully randomized experiment of 427 trials, where Q-learning bidders participate in up to 250,000 auctions for a commonly valued item. The findings reveal that the first price auction, where winners pay the winning bid, is susceptible to coordinated bid suppression, with winning bids averaging roughly 20% below the true values. In contrast, the second price auction, where winners pay the second highest bid, aligns winning bids with actual values, reduces the volatility during learning and speeds up convergence. Regression analysis, incorporating design elements such as payment rules, number of participants, algorithmic factors including the discount and learning rate, asynchronous/synchronous updating, feedback, and exploration strategies, discovers the critical role of payment rules on efficiency. Furthermore, machine learning estimators find that payment rules matter even more with few bidders, high discount factors, asynchronous learning, and coarse bid spaces. This paper underscores the importance of auction design in algorithmic bidding. It suggests that computerized auctions like Google AdSense, which rely on the first price auction, can mitigate the risk of algorithmic collusion by adopting the second price auction.
2020
Inflation Targeting in the United Kingdom: Is there Evidence for Asymmetric Preferences? (with Dr. Naveen Srinivasan)
Master Thesis, Madras School of Economics
In recent times, inflation targeting has been one of the most successful monetary frameworks in advanced economics. However, critics claim that policy rates have been kept higher than necessary. They claim that central banks did not pursue a symmetric inflation target. If a central bank pursues symmetric inflation and output targets, the optimal monetary policy response is a linear forward-looking Taylor rule (Clarida et al., 1999). We use the Linex Loss function as outlined in Nobay and Peel (2003) to relax the assumption of symmetric preferences. The presence of asymmetric preferences implies that monetary policy reacts not only to the conditional expectation of inflation and output gap but also to their conditional variances. Non-linear Taylor rules are estimated on UK data from 1995: Q2 and 2003: Q3. The results support the critics. Inflation targeting was indeed pursued with asymmetric preferences. The findings are robust to the Bank of England's ex-ante forecasts, 'real-time' estimates of the output gap, non-linearities in the supply curve, and alternative forecast horizons. Policy rates have been about 30 basis points higher than necessary due to asymmetric preferences.
2017
Volatility, Persistence and Synchronisation in Indian State Business Cycles (1960-2014)
Research Internship, Reserve Bank of India
This paper studies Indian state business cycles in the period 1960-2014. The Hodrick-Prescott filter is applied on log-linearised Annual Net State Domestic Product (at Constant Factor Prices) to obtain estimates of state cycles. These were consequently analysed. After liberalisation in 1991, state business cycles were less volatile and more serially correlated. Across time, average volatility has fallen and first order auto-correlation has risen. In the post reform period, some states were less synchronised, with the national cycle, but average synchronisation of all states has been increasing over time. The largest Indian states were even more synchronised. Robustness checks show that these results always hold at larger values of the smoothening parameter and at different sizes of the rolling window. However the finding that volatility has fallen, holds even at smaller values of the smoothening parameter.
Findings of a Socio-economic Survey of Pavement Dwelling families in Central Kolkata
Sociology Seminar, Presidency University
This paper is based on one of two socioeconomic surveys that took place in August and October of 2013 respectively. The first survey covered 30 pavement dwellers living in College Street, Central Kolkata. The first survey was preceded by a few field tests, and used a quantitative household questionnaire that pertained to basic socio-economic indicators. This survey was preceded by field tests in College Street itself. The second survey covered 196 households living in various parts of central Kolkata. It also was designed to cover the individuals within households, and along with data on 196 households it has data on the 524 individuals within it. This survey was also preceded by extensive field testing to expose faults in the questionnaire. This paper presents the findings of the second survey.