Ep. 378: Apurv Jain on Alternative Data, Real-Time Macro…

PodcastMacro Hive ConversationsApurv JainOct 2

Apurv Jain describes how alternative data and real-time nowcasting can complement official economic statistics, while noting that measures such as GDPNow can shift substantially during a quarter. He discusses wage pressure, labor bargaining, consumer anxiety, and the limits of forecasting through structural breaks. He also outlines focused uses of LLMs, the role of trusted data in market liquidity, and the relationship between nowcast changes and asset prices.

Economic Growth Labor Market Artificial Intelligence

Preview · 16 more points in the app

Read the Full Summary

Key Takeaways

Apurv Jain says LLMs are most useful in contained workflows—such as generating specific text features, with repeated checks and other data used to limit reliance on a mistaken output—rather than producing a full nowcast. He also says the tools can let one engineer do the work of five and allow AI researchers to take on some engineering tasks.

Jain says wage pressure drives services inflation, which he estimates makes up about two-thirds of inflation in developed markets, and points to worker bargaining power as one way to assess that pressure. He describes the BLS Employment Cost Index as a carefully calibrated benchmark that arrives two months late, against a labor market where bargaining power has shifted from workers in 2022 toward employers.

Jain cautions that GDPNow’s estimate can move from about 2% early in a quarter to minus 2% by quarter-end, making a single latest reading potentially misleading for quantitative trading models. His team’s nowcast draws on different sources from official GDP statistics, which he says remain mostly survey-based.

Jain says the team generally avoids forecasting individual nonfarm payroll releases because revisions make them difficult to predict, though it may flag a release when a large move is coming and the team has additional information. More broadly, he argues that real-time measurement is easier than forecasting through structural breaks, when the data distribution has changed.

GEO-POLITICS

Trade-War Effects — Apurv Jain says the Liberation Day announcements drove an immediate decline in the nowcast, initially led mostly by animal-spirit signals. Traffic and other harder indicators slowed the following month. He says the nowcast then moderated as trade-war rhetoric became more contained and markets grew less worried.

“Like when Liberation Day happened, the first week our nowcast fell a lot led mostly by animal spirit like things…”—

TECHNOLOGY

LLM Nowcasting — Jain says LLMs handle contained tasks, such as generating specific features from text, rather than producing a full nowcast. The team can run them several times to check consistency, then combine their outputs with other data to reduce reliance on any single mistaken result.

“So what we'll do is like, well, let's take this text and then I want you to generate features in this way…”—

ECONOMY

Wage Inflation — Apurv Jain says wage pressure drives services inflation, which he estimates accounts for about two-thirds of inflation in developed markets. He presents worker bargaining power as one way to assess wage pressure.

“So one of the things that drives inflation, obviously services inflation, a big piece, as you and I know about two thirds of inflation, especially in developed markets…”—

Wage Benchmark — Apurv Jain describes the BLS Employment Cost Index as a carefully calibrated benchmark for wage pressure, but says it arrives two months late.

“Or you can wait two months for the gold standard employment cost index. So that's the index that the BLS very carefully calibrates…”—

Labor Bargaining — Apurv Jain describes a reversal in labor bargaining power. In 2022, employers struggling to find specialist workers had less power, while workers had more. He says the balance has since reversed, with employers holding more power and workers less.

“And then employers, are they complaining about not being able to find specialist workers and having to pay them extra bonuses and letting them work from anywhere…”—

MARKETS

Nowcast And Stocks — Jain says stocks tended to rise the month after the nowcast rose. He presents relationships between changes in the nowcast and asset prices as a test of whether the measures contain useful information.

“So if you know the state and we look at changes to our nowcast and then we say, oh, when the nowcast went up, basically stocks go up the next…”—

Options In Practice — Apurv Jain says practical options trading can differ from academic treatments based on the Black-Scholes formula, citing gaps between the formula and how options trade in practice.

“And it was really cool to get into options also because I felt back then options were not as well understood by academia or I mean, in terms of how…”—

MacroHive Markets Launch

The host announces MacroHive Markets, introduces guest Apurv Jain and his alternative-data company, and opens a discussion of real-time economic measurement.

Alternative Data For Nowcasting

Jain says MacroX combines more than a thousand alternative-data pipelines to measure growth, inflation, consumers, labor, and other economic conditions ahead of delayed official releases.

Measuring Wage Bargaining

Jain describes using worker and employer signals to estimate bargaining power and wage pressure in real time, while cross-checking imperfect sources such as social media with many other data streams.

AI And Data Pipelines

Jain explains that different data types call for different methods, from convolutional networks for images to language models for text, and argues that a broad set of signals is more useful than any single source.

Nowcasts Versus GDPNow

Jain says real-time measurement can provide an early lead over government statistics, while Atlanta Fed GDPNow can swing sharply during a quarter because it updates as new releases arrive.

Validating Economic Estimates

Jain says MacroX validates against a range of official and survey measures, finds information beyond both initial and revised GDP prints, and uses its readings to explain rather than precisely predict every release.

Structural Breaks And Labor

Jain argues that measuring current conditions is less vulnerable to structural breaks than forecasting, citing 2023 labor-market strength that narrow recession signals failed to capture.

Jain's Quant And Trading Career

Jain traces his path through quantitative finance, Deutsche Bank FX, Bridgewater, and Microsoft, where he combined portfolio management with machine-learning research.

Using MacroX In Markets

Jain says fundamental investors use MacroX to position for economic trends one to three months ahead, while quantitative users trade changes in its readings across asset classes.

LLMs And The US Outlook

Jain says LLMs are used in tightly controlled, repeatable tasks rather than to generate whole forecasts, and reports above-trend US growth alongside moderate wage and corporate pricing pressures.

Official Statistics And Trust

Jain warns that falling survey response rates and reduced funding threaten official statistics, while emphasizing that trusted public data supports shared economic understanding and deeper market liquidity.

AI Risks And Career Advice

Jain recognizes real but uncertain AI risks and says LLMs help teams scale engineering work; he advises graduates to travel, pursue deep practical projects, and demonstrate initiative.

Influential Books And Closing

Jain recommends works associated with Fama, Taleb, Soros, and Minsky, shares a favorite lesson from Siddhartha, and closes with ways to connect and learn about MacroX.