Framework-Guided Forecasting Mosaic
Combine theory, evidence, and markets while updating when the pieces stop fitting
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 96%
This forecasting method starts with an explicit theoretical or intellectual model of how relevant forces should interact. The analyst then builds a mosaic from company intelligence, official statistics, historical precedents, and market behavior. Rather than treating every observation equally, the analyst searches for the few facts carrying the most important message and checks whether they fit the expected mechanism. Markets are monitored as an independent source of feedback, not dismissed when they contradict the thesis. The resulting view should be understandable to others because its causal reasoning is visible. Confidence remains provisional: temporary bumps do not automatically invalidate a sound framework, but accumulating contradictions require a new perspective. The discipline is therefore conviction with adaptability rather than either prediction avoidance or rigid certainty.
Origin
Ed Hyman describes the approach he developed while providing practical economic research to institutional investors. Extracted from Masters in Business.
Core principles
- 01Forecasting is difficult but necessary for forward-looking decisions
- 02A coherent theory makes individual observations interpretable
- 03No single data point should determine the view
- 04Markets provide evidence that deserves attention and respect
- 05A useful thesis must change when observations stop fitting
- 06Good analysis explains both the conclusion and its reasoning
How to run it
- 1
State the mechanism
Define the theoretical relationship connecting the major inputs to the expected outcome. Make the assumptions explicit enough that later evidence can confirm or challenge them.
Pro tip Prefer mechanisms with historical or economic grounding over narrative intuition alone.
Watch out A vague thesis can absorb every outcome and therefore cannot be tested.
- 2
Assemble independent evidence
Collect observations from primary company research, government data, historical analogues, and financial markets. Seek sources with different weaknesses rather than many versions of the same input.
Pro tip Include both slow structural evidence and timely indicators.
Watch out More data does not automatically create more insight.
- 3
Find the decisive observations
Identify the small number of facts that carry the strongest message for or against the thesis. Explain why they matter more than the surrounding noise.
Pro tip Ask which three observations would most change a client's decision.
Watch out Selecting only supportive facts turns synthesis into confirmation bias.
- 4
Listen to markets
Compare actual market behavior with what the framework implies should be happening. Treat persistent disagreement as information that requires explanation.
Pro tip Observe markets regularly while separating durable moves from intraday noise.
Watch out Do not dismiss prices merely because they conflict with the forecast.
- 5
Communicate the view
Present both the conclusion and the reasoning chain so others can understand the perspective. Clarify which evidence would strengthen or weaken it.
Pro tip Use annotated charts to direct attention to the decisive evidence.
Watch out A confident conclusion without visible logic is difficult to evaluate or update.
- 6
Update without drifting
Distinguish temporary bumps from evidence that breaks the mechanism. When the pieces no longer fit, revise or replace the perspective rather than quietly moving the goalposts.
Pro tip Record what would make you fold the view before that evidence appears.
Watch out Do not confuse patience with permanent attachment to a forecast.
In the wild
An analyst begins with the mechanism that monetary tightening affects activity after long and variable lags. They combine yield-curve inversion, balance-sheet contraction, commercial real-estate stress, company surveys, and weekly unemployment claims. Strong current claims prevent a premature recession call, while the lag framework keeps the analyst alert to later deterioration.
→ The analyst maintains a conditional view without treating either current strength or historical precedent as conclusive.
Suppose company surveys weaken while equities rally. Instead of dismissing the market or abandoning the surveys immediately, the analyst asks whether falling inflation, anticipated rate cuts, or sector concentration explains the divergence and identifies what future readings would settle the question.
→ Conflicting evidence becomes a structured investigation rather than an impulsive forecast reversal.
Common mistakes
Reporting the latest number
Repeating new data without a mechanism gives it no perspective and encourages reactive decisions. Place each observation inside an explicit causal framework.
Giving every input equal weight
A large mosaic still requires judgment about which observations carry the important message. Identify decisive evidence rather than producing an undifferentiated data dump.
Refusing to change perspective
A framework is a guide, not a permanent commitment. Move to another view when material observations repeatedly stop fitting.
Is it for you?
Best for
It is best for investors and analysts making decisions under uncertainty from many noisy and sometimes conflicting signals.
Not ideal for
It is not ideal for mechanical decisions governed by complete, stable data and fixed rules.
From the transcript
“Maybe forecasting is impossible, it's certainly difficult, but you have to do it because in order to make money, you have to have some sense…”
“But I try and have a framework, so I'm not just reporting the latest data point. Put it into a perspective.”
“I also part of My job is to connect the dots, to look at one hundred different observations and find the three that have an…”
From the episode
Ed Hyman on How Investors Should Use Economic Data
Ed Hyman