I write about applied economics, causal inference, pricing strategy, industrial organization, and marketplace economics. Most posts connect economic theory with real business decisions, drawing from my research and experience advising organizations on pricing, experimentation, and demand analysis.
July 2026 | Applied Economics | Causal Inference
A standard two-way fixed effects Difference-in-Differences estimator can produce biased estimates under staggered treatment timing, even when parallel trends holds. This post explains why, using the Goodman-Bacon decomposition, and how cohort-based estimators avoid the problem.
Read articleJuly 2026 | Applied Economics | Causal Inference
A Difference-in-Differences estimate can look statistically clean and still be wrong. This post explains why the parallel trends assumption is the foundation of every DiD analysis, what happens when it fails, and how to test it using event study plots.
Read articleJuly 2026 | Applied Economics | Causal Inference | Industrial Organization
The hardest part of causal inference is not choosing a method. It is constructing a credible counterfactual. When randomization is not possible, the challenge becomes selecting the identification strategy whose assumptions are most credible for the setting.
Read articleJuly 2026 | Marketplace Economics | Industrial Organization
Most pricing frameworks are built for one group of customers. Platform businesses require something different. A pricing decision on one side propagates through the platform, changes behavior on the other side, and feeds back to affect the first side. This is the indirect network effect, and why pricing structure often matters more than price itself in platform markets.
Read articleJune 2026 | Marketplace Economics | Customer Lifetime Value
The customers generating the most lifetime revenue were not the best retention targets. In this marketplace, the highest-spending segment returned least often, while lower-spending customers generated the strongest retention opportunities. The result challenges a common assumption behind many customer lifetime value programs.
Read articleJune 2026 | Marketplace Economics | Industrial Organization
96.3% of products in the marketplace I analyzed had exactly one seller. Among products with multiple sellers, duopolies exhibit lower price dispersion than markets with three or more sellers — a pattern consistent with tacit coordination at low seller counts and greater pricing heterogeneity as competition increases. Seller count alone may not be a reliable proxy for competitive intensity at the product level.
Read articleJune 2026 | Marketplace Economics | Industrial Organization
In the marketplace I analyzed, freight cost has a comparable effect on repeat purchase probability as product price. The coefficients are -0.076 and -0.079 respectively. Freight costs also vary by 50 to 80 percent across regions, driven by geographic distance and distribution structure. Together, these results suggest that shipping policy is not just a logistics decision — it shapes retention in a way that systematically disadvantages high-freight regions.
Read articleMay 2026 | Marketplace Economics | Industrial Organization
Most marketplace platforms earn commission on gross merchandise value, not seller profit. This creates a structural bias toward volume over margin. In the Electronics category I analyzed, a 20% price increase is seller-optimal, improving profit by 5.6% while reducing platform commission revenue by approximately 19%. The result highlights a structural divergence between seller-optimal pricing and platform incentives — not only a pricing problem, but an incentive design problem.
Read articleMay 2026 | Applied Economics | Pricing Strategy
The Lerner Index says elastic demand means lower markups. A revenue-maximizing algorithm applied to an electronics category with elasticity of -2.18 recommends a 40% price cut, projecting an 82.7% revenue increase. The Lerner rule recommends a 20% price increase. Both are internally correct. They are optimizing different objective functions under the same underlying demand structure. The key is what happens when you include the cost structure.
Read articleMay 2026 | Applied Economics | Pricing Strategy
In one electronics category, the revenue-optimal recommendation produced an 82.7% increase in revenue and a 144% reduction in profit at the same time. The elasticity estimate was correct. The optimization system was working as designed. The problem was the objective function.
Read articleMay 2026 | Applied Economics | Econometrics
I estimated price elasticity at the bucket level and obtained a statistically insignificant result. The model was not the problem. Aggregation masked the underlying demand signal through composition effects within product buckets. Disaggregating back to the category level recovered economically meaningful elasticities, illustrating a classic unit-of-analysis problem first documented by Tellis (1988) that still appears in modern e-commerce marketplaces.
Read articleMay 2026 | Marketplace Economics
In the marketplace I analyzed, 97% of customers made exactly one purchase. That looks like a loyalty problem. It is not. The repeat rate reflects the aggregator model, not dissatisfaction. Among returning customers, 53% purchased from a different category within 29 days — a sequential pattern that looks like completing a room, not random browsing. The implication for recommendation systems is direct.
Read articleApril 2026 | Applied Economics | Pricing Strategy
Most pricing teams treat cost assumptions as background inputs and invest heavily in demand modeling. My research suggests this sequence is backwards. A 5 percentage point error in cost assumptions can reverse an optimal pricing recommendation entirely, while a 20 percent elasticity error typically affects magnitude but not direction. The key distinction is whether uncertainty crosses the breakeven cost threshold.
Read articleApril 2026 | Applied Economics | Pricing Strategy
A revenue-maximizing algorithm recommends a 40% price decrease for an electronics category with elastic demand, projecting an 82.7% revenue increase. At a 65% COGS structure, that recommendation reduces profit by 144%. The algorithm is technically correct on its own terms and commercially destructive in practice. Here is why the revenue-profit divergence happens and what to do about it.
Read article
Dinesh R Poddaturi, PhD
Applied Economist
Causal Inference • Pricing
Please use either LinkedIn or the following email to get in touch.