GSMIF macro allocation

We group assets by how they respond to growth, inflation and financing conditions. The examples above describe typical exposures. We start with four fixed ETF baskets, shown below, and adjust their weights each month using the portfolio rules.

Our Winter basket includes growth stocks and gold miners. The trend rule can move these holdings to BIL before we apply GSMIF's government allocation and ETF limits.

We measure mostly US economic conditions, alongside US and international investment prices. We do not calculate separate growth and inflation scores for every country. A positive growth score means the combined readings are stronger than their own historical averages. It can occur during a weak economy that is improving. Cooling inflation means easing price pressure. Consumer prices may still be rising.

Background reading: Bridgewater on growth, inflation and asset exposures; Fidelity on sectors through the business cycle; US Treasury on inflation-linked bonds. These sources explain the general investment principles. Our settings and results still need their own testing.

The decision date is 30 September 2026. A reading may refer to an earlier month because economic statistics are published with a delay. The source tables give the observation and availability dates used by the calculation.

ALFRED supplies the historical versions of industrial production, payrolls, jobless claims, CPI, five-year breakeven inflation and WTI oil available by each growth/inflation decision. Yahoo Finance supplies adjusted ETF prices. M2 also uses ALFRED.

For the remaining liquidity data, we apply publication delays to revised FRED histories. Earlier observations may therefore include revisions published later. This leaves data-revision risk in the liquidity backtest.

For each reading, subtract the average of its last 120 monthly observations, including the decision month. Divide by the typical size of departures from that average, measured as the standard deviation. At least 36 observations are required. Cap the score at ±3.

Quantities and price ratios use logarithmic changes, which measure proportional changes on a consistent scale. Yield changes use percentage points. The source details show the exact values and arithmetic.

The 0.75 setting controls how strongly a score changes the split. A score of +0.75 gives roughly 73% to stronger growth or heating inflation. −0.75 gives roughly 27%. We use the same setting each month.

The exact conversion is 1 / (1 + exp(−score / 0.75)). The other direction receives the remaining share. Calculations use unrounded values, so the four weights total 100% before display rounding.

Both portfolios rebalance at the first trading close each month. Existing holdings earn that day's return before the trade. We charge 0.10% times the share of the portfolio traded, measured as half the sum of absolute weight changes. Adjusted prices include distributions. Returns already include these costs.

The first decision is 31 March 2017, when all five growth readings and all four inflation readings have enough history. The first trade is at the 3 April close. The new holdings earn returns from 4 April. Each chart starts at 100 before its first included return. Shorter periods keep the same historical holdings and restart the displayed value at 100. We measure daily drawdown from the running peak, including that starting value.

We use the same corrected growth/inflation calculation throughout the comparison. Liquidity, portfolio rules and the Northern Trust tilt remain unchanged. Before eligible Northern Trust reports begin in 2019, its line follows Macro Seasons without an analyst adjustment.

We selected the Northern Trust rule using June 2019 through December 2022. We adopted the revised baskets on 6 October 2026 after reviewing historical comparisons. All displayed periods are therefore backtests of rules chosen after reviewing historical data. They cannot be treated as untouched out-of-sample results or live performance.

The October allocation is a recalculated research target. We adopted it after October's scheduled trading date. November is the first prospective monthly target. The earlier simulations and separate paper portfolio remain archived and unchanged.

We check growth/inflation publication dates, but have not verified the precise release times. M2 uses ALFRED. The other liquidity series retain data-revision risk. Today's ETF list also cannot establish which funds were approved in earlier years.

These holdings express our macro preferences as portfolio weights. They show which assets receive more or less capital and how the allocation differs from the benchmark.

An extra 5 percentage points in a category describes an allocation preference. It does not imply a 5% expected return. For Black–Litterman, the PM team needs to translate these preferences into return views over an agreed period and assign uncertainty to each view. The weights shown here provide no calibrated return forecasts.

ACWI, BIL, CPER, DBC, EEM, FXF, FXY, GDX, GLD, IEF, IWF, IWM, QQQ, SHY, SMH, SPY, TIP, TLT, USMV, UUP, VLUE, XHB, XLB, XLC, XLE, XLI, XLK, XLP, XLU, XLV, XLY.

Corporate bond funds and leveraged products are excluded. Bitcoin and silver are outside the currently adopted set. A fund needs its own price history for 13 consecutive months, with measurable return variation, before it can be used. All funds in the adopted baskets and additional defensive holdings are refreshed monthly.