How the CME Tracker works.
Methodology for the cross-engine view, with honest framing of what comes later.
Most published capital-market expectations come from one house with one process. A single number from a single methodology is a starting point; what helps an allocator more is the spread between methodologies on the same asset. The CME Tracker publishes that cross-engine view today, and will add the cross-house panel as the external publications are folded in. Here is how the methodology works.
A CME, or capital-market expectation, is a long-horizon expected-return estimate for an asset class. Major institutional houses publish them annually or quarterly, and serious allocators read them in groups, not individually, because the most useful number is rarely the consensus and almost never any single house’s view. It is the spread.
The CME Tracker is built around that observation. This piece explains the methodology in three layers: what gets published today, what is coming, and what the methodology cannot say.
What gets published today: the cross-engine view.
For each asset class in the taxonomy, TheBuySide computes the expected long-horizon return under three engine families. Each engine starts from textbook decompositions; the differences between them are not noise, they are methodological assumptions about which components dominate.
Building-block. The expected return is decomposed into a real risk-free rate, an inflation expectation, an equity risk premium plus a country premium where relevant, and a volatility-drag term. The estimates are anchored on observable rates and on standard term-structure assumptions. Strong on tractability; weak when the dominant return driver is something that does not appear in the decomposition, for instance a major repricing of multiples or a structural shift in payout policy.
Grinold-Kroner. The expected equity return is decomposed into dividend yield, real earnings growth, net share issuance drag, inflation, and multiple repricing. This is the workhorse model in long-horizon equity expectations; it puts payout structure and share-count behaviour at the centre, which is where they actually belong over multi-decade windows. Strong on equities. Less natural for bonds and for assets where the decomposition does not map cleanly.
Damodaran-ERP (implied). The expected return is anchored on the implied equity risk premium derived from current index levels and forward-looking payouts, plus the risk-free rate and any country-risk premium. Strong on equity, with the methodological honesty of being market-implied rather than analyst-forecast.
For each asset class and horizon, the Tracker shows each engine’s standalone expected return and the cross-engine aggregate: mean, min, max, and spread in basis points. The spread is the thing to look at first. A 50-basis-point spread on a 10-year equity expectation is normal disagreement. A 250-basis-point spread is a methodological structural break and should be read as a warning that one or more engines is operating outside its zone of competence on that asset.
What is coming: the cross-house panel.
Major institutional houses publish CMEs. The list is well-known: BlackRock, JPM, Morgan Stanley, Vanguard, GS, Robeco, Invesco, Mawer, RBC Global Asset Management, BMO Global Asset Management, Franklin Templeton, AQR, and a longer tail of credible publishers. Each shop publishes one or two views per asset class per cycle, with the methodology explained somewhere in an appendix.
The cross-house panel is the second layer of the Tracker. It folds those external CMEs into the same taxonomy and publishes the aggregate, with the spread, and quarter-over-quarter deltas. House views are summarised, never republished verbatim. The aggregate is analysis, not redistribution.
That layer is in flight. The infrastructure exists; the panel-loading is the editorial work that turns it on. When it is on, the two views (cross-engine, cross-house) sit side by side. They tell different stories.
- The cross-engine spread is methodological disagreement: which model dominates on this asset.
- The cross-house spread is institutional consensus disagreement: which house’s house view is which way.
A reader who treats them as the same thing is reading the Tracker wrong.
The taxonomy.
The Tracker classifies every asset under a two-tier taxonomy. Tier 1 is the headline group: developed-market equity, developed-market sovereign fixed income, developed-market credit, emerging-market equity, emerging-market debt, private equity, private credit, real estate, infrastructure, commodities, hedge-fund-class strategies, and cash and equivalents. Tier 2 is the granular slice underneath: US large-cap equity, Canada large-cap equity, EAFE, EM equity, US Treasuries by tenor, IG credit, high yield, and so on through the full tree.
The taxonomy is a forced choice. A house that publishes “developed-market equity excluding North America” maps under EAFE; a house that publishes “US small-cap value” maps under US small-cap, with a footnote on the value tilt. The lineage table records each mapping decision so a reader can audit how an external category arrived at a Tracker slot. Lossy in places; the alternative is one taxonomy per house, which is no taxonomy at all.
What the methodology cannot say.
Three honest caveats.
Horizon-mismatch across the panel. Some houses publish ten-year CMEs; others publish twenty, fifteen, or one-decade-plus-near-term. The Tracker harmonises to a single horizon per published table, defaulting to ten years. CMEs at off-horizons are shown separately, never adjusted to match.
Currency-base inconsistency. External CMEs come in different base currencies and different real-versus-nominal conventions. The taxonomy records the published base; the aggregate is computed only on like-for-like slices. A CAD-base allocator reading a USD-base aggregate must layer her own FX view; the Tracker does not assume she wants USD CMEs translated into CAD without her own hedge.
Selection bias in the panel. The panel will be the set of houses that publish CMEs at all. Houses that decline to publish remain absent. That is not a Tracker choice; it is a fact of the data. The published panel composition will be disclosed at every release so the bias is visible.
How the editorial cycle runs.
Quarterly. A release piece names the date, the panel composition for that release, and the headline changes from the prior quarter. Inline tables for each asset class. Engine-breakdown details available on each row. Source registry of every external publication folded into the panel, with stable links to the underlying methodology.
No house’s CME is republished verbatim. No specific allocator’s portfolio is reverse-engineered from the aggregate. The numbers describe the panel; they do not constitute investment advice or a solicitation.
A note on what this column is not.
The CME Tracker is not a forecast competition. The Tracker does not score houses against realised returns at the end of the horizon, because the realised return of a ten-year CME is forty quarters away and the conditions then will not match the conditions now. There is a separate, narrower piece in the methodology engine that does score engine accuracy on the slices where realised data exists, but that is a different exercise. The Tracker, as published, is a snapshot of where the panel is, what they agree on, and where they disagree.
For the first issue, the cross-engine view is on the column landing page. The cross-house panel follows. Comments and methodology critiques welcome at [email protected]; we will publish substantive ones with attribution.