Data-First Market Forecasting
Start with observable facts, then build the explanation
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 97%
The method reverses the common theory-first approach to forecasting. Begin with current data, historical events, sector behavior, demographics, and market internals, then develop the explanation that best accounts for them. Indicators should not be treated as magical signals: the analyst must identify the process through which each indicator could affect the economy or markets. Historical comparisons provide hypotheses rather than scripts, because institutional responses and economic structures change. Conflicting evidence is examined at a more granular level, such as industries, consumer cohorts, or construction categories. The output is a provisional scenario that remains open to revision as new facts arrive, allowing the forecaster to distinguish durable mechanisms from correlations that may have stopped working.
Origin
Extracted from Masters in Business, where Ed Yardeni explains how empirical analysis and historical experience shaped his contrarian economic and market calls.
Core principles
- 01Begin with facts rather than a preferred theory
- 02Test explanations against historical episodes
- 03Examine the mechanism behind an indicator
- 04Revise models when the economy behaves differently
- 05Separate meaningful signals from market noise
How to run it
- 1
Define the decision
State the economic or investment question precisely and identify what evidence would change the decision.
Pro tip Frame the question around a mechanism, not merely whether an indicator is rising or falling.
Watch out A vague question encourages selective use of evidence.
- 2
Assemble observable facts
Collect current data across the economy, markets, demographics, policy, and relevant industries before choosing an explanation.
Pro tip Include disaggregated series that can reveal offsets hidden by an aggregate number.
Watch out Do not begin by searching only for data that supports an existing view.
- 3
Map historical analogues
Compare the facts with several historical periods and identify both similarities and structural differences.
Pro tip Use history to generate alternative scenarios rather than to declare that events must repeat.
Watch out A visually similar chart may conceal a different causal process.
- 4
Trace the mechanism
Explain how the observed inputs would produce the proposed outcome through specific intermediate events.
Pro tip Write the causal chain explicitly so that each link can be monitored.
Watch out Correlation alone does not establish that an indicator causes the predicted result.
- 5
Challenge the conclusion
Search for evidence from sectors, cohorts, and markets that conflicts with the emerging view.
Pro tip Give extra attention to facts that the consensus or your own model cannot explain.
Watch out Ignoring contradictory evidence converts analysis into advocacy.
- 6
Update continuously
Revise the scenario as fresh data reveals whether the proposed mechanism is operating.
Pro tip Automate recurring charts and data updates so attention can remain on interpretation.
Watch out Do not preserve a forecast merely to appear consistent.
In the wild
The consensus expected a recession after the federal funds rate rose from zero to above five percent. Yardeni examined where rates had started, the resilience of consumers, baby-boomer retirement wealth, strong service employment, and fiscal construction spending. Those facts supported normalization and resilience rather than an automatic economy-wide contraction.
→ The framework produced a contrarian no-recession scenario while the widely anticipated recession failed to appear.
A strategist sees the equal-weighted S&P 500 lagging the capitalization-weighted index and concludes that only a few stocks are advancing. Applying the framework, the strategist examines more than one hundred constituent industries and discovers many with positive or bull-market-level gains, though they trail the megacaps.
→ The investor avoids mistaking relative underperformance for an absolute absence of breadth.
Common mistakes
Starting with the theory
Choosing the explanation first encourages the analyst to select only the facts that confirm it.
Treating indicators as laws
An indicator can fail when the causal chain it previously represented changes or is interrupted.
Using one historical analogy
No single decade captures every relevant feature of a new economic environment.
Is it for you?
Best for
It is best for investors, strategists, and business leaders making decisions under economic uncertainty.
Not ideal for
It is not ideal for people seeking a fixed formula that produces precise predictions without judgment.
From the transcript
“I think that they first start out with theories and then look for data to support it.”
“again we start with the data and then come to the conclusion rather than the other way around.”
“you really have to be very flexible in looking at this economy and recognize how things change, and you know how models it used to…”
From the episode
Ed Yardeni on Long-Term Bull Market