Horizon Actuarial Survey of Capital Market Assumptions 2026

Horizon Actuarial Services asks investment advisors what they expect from every asset class a pension plan holds, then publishes the averages. The 2026 edition covers 43 advisors. Over ten years they average 6.39% a year from US large cap equities and 2.44% inflation.

Source: Survey of Capital Market Assumptions, 2026 Edition, August 2026. Returns are nominal and annualized (geometric), in USD. Transcribed from the report on 3 September 2026.

What the survey is

Horizon Actuarial Services is an independent actuarial firm, actuary to more than 100 multiemployer defined benefit pension plans in the United States. Setting a plan’s assumed investment return is part of that job, and the firm describes itself as retirement and healthcare actuaries rather than investment professionals. So it asks the investment advisory community instead. The survey has run since 2010, when 8 advisors took part, and Horizon has published a report on it since 2012, when there were 17.

None of the figures are Horizon’s own view, and the report attributes none of them to a named firm, which is a condition of taking part. What it gives an actuary is a reference point for one question: is this plan’s assumed return defensible against what 43 advisors currently expect?

Of the 43 advisors, 14 submitted a single set of assumptions at a 10-year horizon and 29 submitted both a shorter-term and a longer-term set. Every 10-year average below rests on all 43; every 20-year average rests on the 29. Most advisors dated their assumptions on or around 1 January 2026, a few as early as 1 October 2025 and a few as late as 31 March 2026.

Horizon publishes the report through a form on its download page, so reading the source means filling one in.

Average expected returns, 2026 edition

Exhibit 4 of the report, page 5. Annualized (geometric) nominal returns, averaged across advisors with each unique assumption set given equal weight.

Horizon Actuarial 2026 Survey of Capital Market Assumptions, Exhibit 4, page 5. The 10-year column averages all 43 advisors and the 20-year column the 29 who supplied long-term assumptions, so the difference column compares two different groups. Inflation is the same on both horizons.

Returns are geometric throughout, which the glossary defines as the annualized return over a multi-year period. Where an advisor supplied only arithmetic returns, Horizon converted them using E[Rg] = ((1 + E[Ra])² - Var[R])^1/2 - 1. Figures are market returns before any value added by an active manager, and net of fees in the classes where no passive vehicle exists, which the report names as hedge funds and private equity.

Restricting both columns to the 29 advisors who gave both horizons changes little, with one exception. On that narrower basis commodities is the only class where the 10-year average sits above the 20-year one, at 5.03% against 4.91%. Horizon reports that comparison separately as Exhibit 3, page 5.

How far apart the advisors are

Exhibit 20 gives the spread across all 43 advisors over the 10-year horizon, and Exhibit 17 the average volatility each class was assigned. Both sit far enough apart to change what a portfolio built on them looks like.

Exhibits 20 and 18, pages 19 and 17, 10-year horizon, all 43 advisors. Percentiles are of advisors, not of outcomes. Horizon notes that the median advisor’s figure is not the average shown in Exhibit 4, though for most classes the two sit close together. Inflation carries no volatility row in Exhibit 18.

The widest disagreements are in the alternatives. Commodities runs from 1.1% to 10.2% a year, a span of 9.1 points, and private equity from 6.1% to 16.0%, a span of 9.9. Horizon attributes that partly to the different underlying strategies advisors have in mind when they use the same label. Core US bonds, where the answer is mostly a matter of observable yields, spans 3.8% to 5.7%.

What changed since the 2025 edition

Horizon states the year’s change through a hypothetical plan. Run on the 2026 averages, that plan’s expected 10-year return is 6.59% against 6.67% last year, and its 20-year return 7.01% against 7.03%: 8 basis points lower and 2 basis points lower. Underneath that near-standstill, several classes moved by a third of a point.

Exhibit 4 of each edition, page 5 in both. The 2025 column averages 41 advisors and the 2026 column 43, so part of every change comes from the panel and part from the outlook.

Emerging market equity fell furthest, from 7.38% to 7.01%, and emerging market debt with it. Private debt and infrastructure each gave back more than three tenths of a point. US large cap sits at exactly the figure it carried a year ago. The expected inflation line rose from 2.38% to 2.44%, so the real return implied by the equity figures fell slightly further than the nominal ones did.

The hypothetical plan

The report ends by running one portfolio through the survey. The weights are arbitrary except for a rule Horizon states plainly: every asset class in the survey gets at least a small allocation. The plan carries a benchmark return of 7.00% a year, which is what the probabilities are measured against.

Exhibit 8, page 8.

Exhibit 13, page 12, summarized on page 8 as Exhibits 9 and 10. The cautious and optimistic columns are the single most and least cautious advisors in the survey, and Horizon notes they are not necessarily the same firm at each horizon or from one year to the next.

On the survey average this plan has a 50.1% chance of clearing 7.00% a year over twenty years. Take the most cautious advisor’s assumptions and the same portfolio has a 27.1% chance; take the most optimistic and it has 69.8%. Horizon writes that it may be difficult for an actuary to justify an assumption outside the middle 50 percent. On the survey average that band runs from 5.37% to 8.64% over twenty years, and wider still over ten, from 4.28% to 8.90%.

Against our own house comparison

Portfolio Lab keeps a table of 17 firms that publish their assumptions directly, and takes a median across the ones with a comparable 10-year figure. Horizon averages 43 advisors who submit theirs privately. Three of the four classes below land within 0.39 of a point of each other. The exception is emerging markets, where Horizon’s advisors average 7.01% against our 7.65%, a gap of 0.64 points.

Our medians from the 17 houses at /capital-market-assumptions, Horizon’s from Exhibit 4, page 5. A positive difference means Horizon’s advisors expect more.

The two samples overlap, which is why the agreement counts for less than it looks. 11 of the 17 houses we track are named on Horizon’s participant list: Research Affiliates, Vanguard, Invesco, BlackRock, Verus, Amundi, Meketa, J.P. Morgan, Callan, BNY Mellon and PIMCO. Ours is a median of 17 documents anyone can download; Horizon’s is a mean of 43 private submissions, which pulls harder toward the outliers. Horizon also standardizes what it receives, so a firm that publishes a five-year horizon or a probability-weighted range still contributes a 10-year point estimate to the average, where our table leaves those cells blank instead.

Two of the four rows line up on the label alone. Horizon’s US Corporate Bonds - Core is a standardized bucket assembled from whatever each advisor submitted as core fixed income, and the report names no index behind any of its classes, its Non-US Equity - Developed included. Those two lines are close without being the same measurement.

Figures transcribed from Survey of Capital Market Assumptions, 2026 Edition, August 2026, with the exhibit and page number given beside each table, and re-checked when Horizon publishes a new edition. Portfolio Lab is not affiliated with Horizon Actuarial Services, and the survey is reproduced here under fair use with attribution. This is analysis, not investment advice.