How to Build a Macro Scenario Analysis: Growth, Inflation, Rates and Liquidity

Answer first

What macro scenario analysis is (the answer first)

Macro scenario analysis is the practice of describing assumptions about growth, inflation, policy-rate direction and liquidity as several conditional paths (scenarios), and attaching observable indicators, triggers, falsification criteria, transmission channels and an update routine to each. The goal is not to predict one future but to write down in advance, in an observable form, which conditions would make each scenario more likely.

Once that organization is in place, every time new data is released you can mechanically check which scenario’s trigger it fired and which scenario’s falsification line it crossed. As a result, the report is not something you throw away as “right” or “wrong”; it becomes an updateable document that you grow by revising only the parts where the assumption changed. This is work for anyone who wants to move beyond merely collecting macro information and shape it into conditional branches, monitored items and an update cadence.

This is the final article in the Macro Research Workbench cluster. How to read each individual dataset is left to the other cluster articles; here we focus on the step of binding them into a single scenario. If you want the big picture first, starting from the Macro Analysis Guide (connecting COT, rates, real yields and EIA data) makes this article’s place in the series easier to see. Every number and figure shown here is fictional educational data, not a real market value, forecast or trade recommendation.

Terms

How a scenario differs from a forecast

If you confuse the two, scenario analysis collapses into “several guesses.” The difference is whether you assert or describe with conditions.

• Forecast: asserts one future—”this will happen.” It is judged by hit rate and loses its value when it is wrong.
• Scenario: a conditional statement—”if this condition is met, this path becomes more likely.” It bundles assumptions, confirming indicators and falsification criteria together, and is judged not by being right or wrong but by “how quickly you notice when an assumption breaks.”

The labels base, upside and downside are also common, but they are differences in how assumptions are set, not value judgements (good or bad). A strong-growth path is not “good,” a weak one is not “bad”; they are labels that distinguish which assumption was met. Likewise, avoid automatically assigning numeric probabilities to each scenario. Assign a probability only when you can also state its source (market-implied pricing, a survey, an explicitly labelled subjective view) and how it will be updated. When you want to step into comparisons with past episodes, split that into Macro Regime Analysis (historical comparisons, vintage data and look-ahead bias), which handles causal testing. The key is dividing the roles: scenarios are conditional branches about the future, regimes are the validation of past episodes.

Base design

The four base axes: growth, inflation, policy rate and liquidity

Build too many scenarios and they become unmanageable. Start by narrowing to four axes and classifying each into a few levels.

• Growth: improving / flat / worsening. Example proxies are manufacturing PMI (an index with 50 as the expansion-contraction boundary) and the year-over-year real GDP growth rate.
• Inflation: rising / flat / falling. An example proxy is the year-over-year change in core CPI (seasonally adjusted).
• Policy-rate direction: hike / hold / cut. Classify by the central bank’s direction at each meeting.
• Liquidity: expanding / neutral / tightening. Approximate it with the year-over-year change in the central-bank balance sheet or reserves. Because definitions differ across markets, always state which series you use to approximate it.

Placing growth and inflation on the vertical and horizontal axes gives four quadrants, over which you layer policy-rate direction and liquidity. If you want to add the relationship between rate differentials and FX as an axis, see Interest Rate Differentials and FX (comparing nominal, real and expected rates); if you want to check the level of rates themselves and the curve shape, see How to Read Treasury Yields and the Yield Curve (2s10s, real yields and breakevens).

Visual 1

Growth-by-inflation four-quadrant scenario map

The figure below places inflation on the horizontal axis (left = falling / right = rising) and growth on the vertical axis (up = improving / down = worsening), with a typical situation name in each of the four quadrants. Over it, the three fictional scenarios used consistently in this article (Base / Upside / Downside) are plotted as points (all fictional educational data).

What matters is not to color the quadrants “good” or “bad.” The figure above adds labels (U / B / D) and quadrant names on top of color, so meaning is never conveyed by color alone. Which quadrant a scenario lands in reflects a difference in assumptions and does not translate directly into a buy/sell direction.

Data dictionary

Fictional dataset, timestamps, units and transforms

Scenario analysis assumes that every scenario shares the same indicator definitions, frequency, units and transforms. Here we define, once, the fictional dataset used throughout this article, and we do not change the values in later prose, figures, tables or the mini-tool. Everything is fictional educational data, not an actual published value.

Alongside this, we fix two auxiliary fictional statistics used in the transmission discussion. The 52-week rolling correlation between the gold price and the 10-year real yield is −0.62 (n = 52 weeks, dimensionless), and the z-score of the core CPI surprise is +1.4 (sample: trailing 24 months, standardized by mean and standard deviation). As a rule, a rolling correlation should state its window, minimum sample, endpoints and missing-data handling, and a z-score should state its sample size and denominator definition. If you add COT positioning as a confirming indicator, see COT Percentile and Z-Score (compare positioning extremes and changes) for how to read percentiles and z-scores, and Gold and Real Yields (test the inverse relationship across regimes and rolling windows) for the gold-and-real-yield relationship.

Visual 2

How to build a scenario card (assumption → observable → trigger → transmission → falsification → update)

A scenario is easier to run when you distill it onto a single card. The card has six fields: the assumption, observable confirming indicators, the trigger, transmission channels, the falsification criterion, and the next-update date with data sources. The figure below shows this flow from left to right.

The trick is always to design the trigger (the condition that makes the scenario more likely) and the falsification criterion (the condition under which the scenario can no longer be maintained) as a pair. Deciding the falsification criterion first helps you avoid the confirmation bias of inventing reasons after the number is out. The transmission field is a place to write “hypotheses to test,” not a place to write buy/sell directions or price targets.

Transmission

Treat cross-asset transmission as a hypothesis

How each scenario might propagate to rates, currencies, equities, gold and oil is one of the highlights of scenario analysis. But it is also the part most prone to misreading. A transmission channel is a monitoring signpost—”under this assumption, go and check this channel”—not a conclusion of “therefore buy/sell.” The mappings below are one example of commonly discussed channels, organized on the basis of fictional educational data.

Correlation does not prove causation. Even if the 52-week correlation between gold and the real yield is −0.62, that only expresses how much they moved together over the same period; it does not establish that either one is the cause, and the sign and strength can change if you change the window. To verify lead and lag relationships, see Lead-Lag Analysis Explained (test time-shifted relationships without confusing causality), which covers the design of time-shifted correlations, and check the supply/demand side of oil in How to Read EIA Crude Oil Inventories (stocks, production, imports and refinery runs). Do not turn a transmission channel into the assertion that “it is a leading indicator, so it can predict”; use it as a tool that points to the next data to check.

Visual 3

Compare across countries with country macro cards

When you extend scenarios to several countries, aligning indicator definitions, frequency, seasonal adjustment and currency units is essential. Lining up countries whose definitions differ is not a comparison. The dashboard-style figure below places three fictional countries (Country A, B and C) side by side under the same indicator definitions (all fictional educational data).

Viewed by country, even the same “slowdown” comes with a different policy-rate level and real yield, so the scenario assumptions change from country to country too. Note also that unless you align the treatment of purchasing-power parity and currency units, comparing levels can be misleading.

Consistent example

Three fictional scenarios: how an update fires a trigger and falsifies another

Using the fictional dataset defined so far (Table 1), let us actually write out three scenario cards. Each card has an assumption, observable indicators, a trigger, a falsification criterion, a next-update date and transmission channels (hypotheses). The values match Table 1.

Now suppose new observations arrive on the update date. For example, manufacturing PMI improves from 50.2 to 53.5 and stays above 53 for two consecutive months. This observation crosses the Base falsification criterion (PMI above 53 for two consecutive months) and at the same time fires the Upside trigger (PMI above 53 for two consecutive months). In other words, a single data update works by falsifying one scenario and making another more likely at the same time.

What matters is that the action at this point is “revising the report,” not “making a trade.” Because the Base assumption has broken, you switch the transmission channels you go and check to the Upside side (upward rate pressure and a firm USD) and update the indicators to watch next. Even if there was a similar episode in the past, only part of the conditions being similar does not guarantee the same outcome. The treatment of past episodes is split into the Macro Regime Analysis article.

Educational mini-tool

Scenario matrix builder

The mini-tool below is for practicing how to assemble the skeleton of a scenario card. Choose growth, inflation, policy direction, liquidity and evidence quality, and enter monitored indicators, a trigger, a falsification criterion and a next-update date; it organizes and displays the card elements for up to three scenarios. It does not auto-generate probabilities, price targets or buy/sell directions. The transmission field simply points to the data series to check as a “channel to go and confirm (hypothesis).” Input stays inside your browser; nothing is saved or sent.

First, so it is readable even with JavaScript disabled, here is a static output example for the default inputs. This static example matches the initial state of the tool below.

Limits

Five limits that are easy to misread

Scenario analysis is useful, but forget the following limits and you slide back into “forecasting.”

• Correlation does not prove causation: a 52-week correlation of −0.62 is the degree of co-movement over the same period; it does not establish a cause. The sign can change if you change the window.
• Extremes do not guarantee a reversal: an extreme observation such as a z-score of +1.4 does not necessarily revert to the mean.
• The yield curve does not fix the timing: even an inversion such as 2s10s at −15bp cannot determine whether or when a recession occurs.
• Inventories do not dictate the price direction: even an EIA inventory build of +4.0 million barrels does not, on its own, set the direction of the oil price.
• A scenario is not a forecast: base, upside and downside are differences in assumptions and do not promise that any of them “will happen.”

All of these are restatements of the same principle: “do not leap from a single statistic to a price direction.” Whenever you feel the leap coming, return to the falsification criterion and the additional data to check next.

Operational check

An operational checklist for turning scenarios into a report

The minimum items you want in place when distilling scenarios into an updateable report.

1. Did you classify the four axes (growth, inflation, policy-rate direction, liquidity) into levels and state at least one supporting indicator?
2. Did you write a trigger and a falsification criterion as a “pair” for each scenario? Did you decide the falsification criterion first?
3. Did you align the frequency, unit, seasonal adjustment, reference period and source of the observed indicators?
4. Did you separate the observation, release, retrieval and revision dates, and avoid using a value after the analysis date in the past?
5. Did you write the transmission channels as “channels to check (hypotheses)” and not turn them into buy/sell directions or price targets?
6. Did you enter a next-update date matched to each series’ release frequency?
7. When quantifying a probability, did you also state its source and update method (and if you cannot, record evidence quality instead)?
8. For cross-country comparisons, did you align the indicator definitions, currency units and the treatment of purchasing power?

Using the service

Workbench workflow (Free → Pro → Premium)

The work so far can be made progressively more efficient with the Macro Research Workbench. First review public macro data for free; when you need monitoring, saving and exports, move to Pro; and when you go all the way to data integration, scenarios and reporting, move to Premium. That is the maturity flow. The figure below shows this maturity roadmap. Feature names, storage behavior and coverage can change, so confirm the latest on the plans page and in the implementation.

This tool mechanically organizes and visualizes public macro data and data you load locally on your device. It does not provide lot, margin, trading-cost, spread, swap, P/L calculation, trade signals or personalized investment advice. If you need lot or trading-cost calculation, use the FX and CFD lot-size calculation guide or the trading cost calculation guide, and for strategy validation the TradingView backtesting and robustness guide, each according to its purpose. You can find the English learning articles from the article index.

Summary

Summary: the answer to the core question and your next step

Building a macro scenario analysis is not about pinning down a single future. Split the assumptions across four axes—growth, inflation, policy-rate direction and liquidity—attach observable indicators, triggers, falsification criteria, transmission channels and a next-update date to each scenario, and revise the report by checking for triggers and falsification every time data updates. Running that loop is the essence. Design the trigger and falsification as a pair, and treat transmission channels as “channels to check (hypotheses)” rather than buy/sell directions, and your scenarios will not slide back into forecasting.

In practice, hold to five points: (1) classify the four axes with supporting indicators, (2) write the trigger and falsification as a pair, (3) separate the observation, release and revision dates to keep timestamp integrity, (4) write transmission channels as hypotheses, and (5) match the next-update date to the release frequency. Get these right and you have the skeleton of an updateable macro report. When you want to check the big picture, the quickest route is to return to the cluster’s parent article and review how to read each dataset.

Related articles

Related reading