Bottom-Up Economic Forecasting
Ask your mum before you read the textbook — real financial lives beat models
Bottom-Up Economic Forecasting is Gary Stevenson's alternative to textbook-driven macroeconomic analysis. The core insight: economists trained in universities come exclusively from wealthy backgrounds, know only wealthy people, and have no first-hand exposure to how ordinary households actually experience the economy. Their models are, therefore, systematically blind to the signals that matter most for consumption, debt behaviour, and recovery timing.
The method replaces model-first analysis with people-first observation. Walk the high street. Look at what shops are closing. Ask your friends what their financial situation is like. Ask their parents. Look at the adverts on the tube — they reveal who advertisers think has money to spend. These observations create a ground-truth signal that orthodox economists cannot access because of who they are and where they come from.
For Gary, this approach was his primary competitive edge on the trading floor. While colleagues ran the same models as every other bank, he had real-world data from ordinary financial lives that contradicted those models. The gaps between what the models predicted and what people actually did — and why — were exactly where mispriced assets lived.
- The gap between what economists predict and what people do is found in the gap between elite social networks and ordinary financial lives.
- Real household financial data — rent burden, debt levels, savings rates — is more predictive than aggregate GDP or inflation figures.
- Ordinary people spend when they have money, not when borrowing is cheap — stimulus that bypasses them produces no multiplier.
- Direct observation of ordinary spending environments (high streets, adverts, friend conversations) provides signal unavailable in Bloomberg terminals.
- The best economists are mostly hidden inside banks, being paid not to share their analysis publicly.
- Map your social network's financial realitySystematically ask the people in your real-life network — family, friends, their parents — what their financial situation actually looks like. Not how they feel about the economy in the abstract, but: can they save? Is their mortgage under control? Are they cutting spending?Pro tipWorking-class and ordinary-income contacts are the signal; high-income contacts will reproduce the consensus blindness you're trying to escape.WarningConfirmation bias risk: if your entire network is in similar circumstances, you're not sampling the distribution — deliberately seek out contacts across income levels.
- Walk the commercial environment with analytical eyesWalk your local high street or shopping area not as a consumer but as a researcher. Note which shops are closing, which are opening, what the advertising copy says and who it's targeting, what prices have changed. This is the economy's ground-level display.Pro tipAdverts on public transport are calibrated to who advertisers think has money — a shift in ad content (payday loans, discount retailers) signals a spending-power shift faster than GDP data.
- Identify the specific constraint blocking the model's predicted behaviourWhen a model predicts increased spending but you see none, find the specific reason in ordinary people's lives. Is it rent crowding out disposable income? Debt service from expired fixed-rate mortgages? Job insecurity preventing commitment spending? Name the exact constraint.WarningDon't stop at 'people aren't confident' — that's the output, not the cause. Find the balance-sheet reason.
- Cross-check with the model to find the gapCompare your ground-truth findings against what official forecasts or market consensus implies. The gap between lived reality and model output is your research edge. Positions with consensus on one side and human observation on the other are high-conviction entries.Pro tipGary's sustained edge came from being the only person on his trading floor who had access to both elite financial data and working-class financial reality. Own that duality.
- Update iteratively — the signal evolves with the economic cycleGround-level observation is not a one-off exercise. The constraint that blocked spending in 2011 (post-crisis debt) differed from 2021 (COVID savings exhaustion). Re-run the observation process at each new economic inflection point.WarningDo not extrapolate past ground-truth signals into new cycles without fresh observation — the mechanism can change even if the structural inequality trend doesn't.
Gary asked his flatmate why he wasn't spending more in the post-2008 period when the consensus predicted a spending recovery. His flatmate showed him shoes with holes in the soles.
Gary challenged Oxford economics professors who lectured on interest rates, noting their predictions had been wrong for a decade. One professor said 'you always knew rates would stay at zero' — unaware that he had predicted the opposite every year.
Standard models predicted that cheap borrowing post-2008 would stimulate household spending. Ground-truth observation showed households couldn't borrow more — they already had too much debt, rising housing costs, and no wage growth.
After a bad Swiss interest-rate trade lost the bank $8 million in 2010, Gary instinctively returned to economics textbooks. A senior colleague — from Liverpool, never university-educated, self-made over 30 years — watched for two days, then knocked the books from Gary's hands. His instruction: "If you want to know about the economy, go home and ask your mum what her financial situation is like. Walk down the high street. Ask your friends what is going on in their financial lives."
Gary describes this as the most important thing he ever heard. When he applied it, he found that the entire post-2008 consensus on recovery was wrong — consumers weren't spending because they had no money, not because of interest rates. That insight, grounded in direct personal observation, drove his most profitable trading period.