From Macro Factors to Investable Portfolios
Why Factor-Mimicking Portfolios Matter for Asset Allocation
In my previous article, I argued that asset allocators should look beyond asset-class labels and think more directly about the economic risks that matter to the institution.
For a long-horizon asset owner, I find it useful to begin with a small number of macro factors rather than an ever-expanding list of style factors. Economic growth, inflation and real interest rates provide a useful starting point because they describe much of the economic environment through which long-term portfolios have to navigate.
But identifying the factors is only the beginning.
- Growth is not a security.
- Inflation is not a portfolio.
- A change in real interest rates is not an investable return in itself.
If factors are going to play a practical role in asset allocation, portfolio construction and risk management, they need to be translated into something observable in financial markets.
That is where factor-mimicking portfolios become useful.
Start by Defining the Factor
Before constructing a factor-mimicking portfolio, the factor itself needs to be defined carefully.
“Inflation,” for example, can mean realised inflation, a change in expected inflation or an inflation surprise relative to what markets had already priced. There is also the question of which inflation. Headline inflation combines components with very different economic and asset-pricing properties, and recent evidence from Fang, Liu and Roussanov (2026) suggests that assets commonly described as inflation hedges can have very different sensitivities to core and energy inflation.
These distinctions are related, but they are not interchangeable. An inflation rate of 3 per cent may have very different implications for asset prices depending on whether investors expected 2 per cent or 4 per cent.
The same applies to growth and real rates. For investment purposes, what often matters is not simply the level of an economic variable, but how it changes relative to what markets had priced.
For an asset owner, however, the definition should ultimately come back to the mandate. If pension benefits are linked to a broad cost-of-living measure, the relevant inflation factor should reflect the inflation risk embedded in those liabilities. An institution without the same linkage may reasonably place less emphasis on it.
The factor framework should follow the institutional problem, not the other way around.
Moving the Factor into Return Space
Once the factor has been defined, the next task is to translate it into a portfolio return.
The basic idea is straightforward.
First, select a broad universe of liquid, transparent market instruments. Depending on the factor, this might include equity indices, government bonds, inflation-linked bonds, commodities, currencies and other instruments with different sensitivities to the economic environment.
Second, estimate how those returns respond to the macro factor.
Some assets will respond positively, some negatively, and some may have relatively little sensitivity.
Third, combine those instruments into a portfolio designed to have meaningful exposure to the target factor while limiting unintended exposure to the others. The weights therefore depend not only on each instrument’s sensitivity to the target factor, but also on how the instruments move with one another.
Conceptually:
A simple representation is:
\[
R^{FMP}_t = w’R_t
\]
R^{FMP}_t = w’R_t
\]
where \(R_t\) represents the returns of the liquid instruments and \(w\) represents the portfolio weights.
An inflation-mimicking portfolio, for example, might place greater weight on instruments that historically respond positively to inflation surprises and offset those that respond negatively, while trying not to introduce a large unintended growth exposure.
The exact construction can vary.
For an allocator, the key point is that a non-traded economic factor has now been translated into a market return series.
Defining the factor tells us what economic risk we care about. The factor-mimicking portfolio gives us a way to observe that risk in return space.
A Mimicking Portfolio Is a Tool, Not a Truth
A factor-mimicking portfolio is a proxy.
It is not a claim that a particular combination of securities is inflation, growth or real rates.
Its behaviour depends on the assets used to construct it, the estimation period, the factor definition, and the constraints placed on the portfolio.
The construction therefore needs to remain economically intuitive.
A portfolio may have an excellent statistical fit and still be of little use if its exposures are unstable, difficult to implement or hard to explain.
For an asset owner, robustness matters more than mathematical elegance.
The objective is not to find a perfect factor. It is to develop a sufficiently stable representation that improves portfolio decisions.
From Factor Construction to Asset-Class Mapping
Once the factor-mimicking portfolios have been constructed, the direction of analysis changes.
The first exercise moved from the macro factor into market returns.
The next step is to use those return series to examine the economic exposures embedded in different asset classes.
Conceptually:
\[R_{asset,t} = \alpha + \beta_G F_{G,t} + \beta_I F_{I,t} + \beta_R F_{R,t} + \varepsilon_t \]
The estimated betas tell us how sensitive an asset has historically been to growth, inflation and real-rate factors.
Calling equities “growth assets” is a useful starting point, but it says little about the magnitude or stability of that exposure.
Similarly, describing real estate as an inflation hedge is only useful if the relationship is actually present in the data and sufficiently persistent to matter.
Factor mapping turns those broad descriptions into measurable exposures.
Factor Mapping Should Be Measured, Not Assumed
Asset classes do not have permanent economic identities. Their behaviour changes.
Real estate may provide meaningful inflation sensitivity in one environment and much less in another, depending on leverage, lease structures, financing conditions and valuations.
Infrastructure may behave differently depending on whether revenues are regulated, contracted or economically sensitive.
Equity sensitivity to inflation may depend on whether inflation reflects strong nominal demand or a deterioration in real growth.
A static factor matrix therefore tells us only so much.
Once factor-mimicking portfolios provide return histories, the mapping itself can become a time series.
That allows the allocator to see not only the current exposure, but how the role of an asset has changed.
Rolling or time-varying estimates can show when an asset that once provided useful diversification has gradually become more exposed to the same economic forces already dominating the portfolio.
That is much more informative than assigning each asset class a permanent position in a factor table.
A Simple Factor-Mapping Example
Consider a deliberately simplified framework with three factor-mimicking portfolios: growth, inflation and real rates.
Suppose the estimated sensitivities of several asset classes looked like this:
![]() |
| * For simplicity, the real-rate factor is defined so that a positive factor move represents an increase in real rates. |
These numbers are illustrative, not empirical estimates.
Now suppose an institution holds:
- 50 per cent global equities
- 30 per cent government bonds
- 10 per cent real estate
- 10 per cent infrastructure
The approximate portfolio sensitivities would be:
- Growth: about +0.51
- Inflation: about -0.10
- Real rates: about -0.49
Viewed through asset classes, the portfolio appears diversified.
Viewed through factors, the picture is more revealing.
The portfolio retains meaningful growth exposure, offers little protection against inflation, and remains materially exposed to higher real rates.
If the institution’s mandate indicates that greater inflation sensitivity would be desirable, simply adding more “real assets” is not necessarily the answer. Recent evidence suggests that some assets commonly viewed as inflation hedges provide much stronger protection against energy inflation than against core inflation. The asset-class label alone does not tell us whether the exposure matches the inflation risk that matters to the institution.
The allocation problem is therefore to identify an implementation that increases the desired inflation exposure without adding more growth, illiquidity or real-rate sensitivity than the institution wants.
The value of factor mapping is not that it tells us what an asset class is. It tells us what economic risks the asset class is actually delivering to the portfolio.
One More Practical Caveat
Before taking the mapping too far, one practical issue is worth keeping in mind.
The liquid instruments used to construct the factor portfolios need to provide enough variation to identify the macro exposures meaningfully.
It is also useful to avoid making the exercise mechanically circular.
If an asset is used heavily to construct a factor and then immediately “discovered” to have a large exposure to that same factor, the result may tell us less than it appears to.
Using a broad proxy universe, testing the relationships out of sample and checking whether the resulting sensitivities remain economically sensible can help.
The objective is not statistical purity.
It is confidence that the mapping contains useful information rather than simply reproducing the assumptions used to build it.
The Reference Portfolio Provides the Anchor
Once factor exposures can be measured, the reference portfolio can be expressed in the same language.
If the reference portfolio represents the long-term market risk the institution has deliberately chosen to own, it provides the strategic risk anchor. Expressing it in factor terms helps describe the economic character of that risk.
The implementation portfolio can then be compared against it.
\[ \text{Active Factor Exposure} = \text{Total Portfolio Factor Exposure} - \text{Reference Portfolio Factor Exposure}\]
Suppose the total portfolio has 2 per cent tracking error relative to the reference portfolio.
That number tells us that the portfolio is different.
It does not tell us what the difference means economically.
Factor mapping may show that the total portfolio carries more growth exposure, less inflation sensitivity and roughly the same real-rate exposure.
That allows the discussion to move from statistical deviation to investment intent.
- Is the additional growth exposure deliberate?
- Is the reduction in inflation sensitivity consistent with the mandate?
- Is the portfolio being adequately compensated for those deviations?
Tracking error tells us how different the portfolio is. Factor exposure helps tell us what that difference actually means.
Exposure Is Not the Same as Risk Premium
Factor sensitivity tells us what economic risk the portfolio owns. It does not tell us whether the institution is being adequately compensated for owning it. A high growth beta is not inherently good or bad.
If the institution has the capacity to bear growth risk and expected compensation is attractive, additional exposure may be entirely appropriate.
If the portfolio is already dominated by growth exposure and expected compensation is weak, the same position may be difficult to justify.
Portfolio construction therefore requires both Factor exposure and Expected factor premium.
Both must be considered alongside the institution’s mandate, risk capacity, and liquidity needs.
This becomes particularly important when different investments provide similar economic exposure but come with very different liquidity, governance and implementation requirements.
The factor lens tells us what risk is being accessed.
The expected premium helps determine whether the exposure is worth owning.
Risk Budgeting Looks Different Through a Factor Lens
The same framework changes how I think about risk budgets.
Tracking error remains useful, but its relevance depends on the level at which it is applied.
For an asset manager operating within a delegated objective, tracking error against an asset-class benchmark is an important measure. It shows how much active risk the manager is taking relative to the portfolio they were asked to manage.
For the asset owner, that is only one part of the picture. A public-equity team can remain within its tracking-error budget. A private-equity programme can remain within its approved range. Infrastructure and real estate can remain within their respective limits. Yet all of them may be adding exposure to the same underlying growth factor. Each team can therefore operate responsibly within its own framework while the total portfolio moves away from the economic risk profile the institution intended to own.
Asset-class tracking error is useful for managing managers. It is not sufficient for managing an asset owner.
The reference portfolio provides the natural conceptual anchor for total-fund active risk.
Factor sensitivities then show where those deviations are coming from.
- More growth risk.
- Less inflation sensitivity.
- Greater real-rate exposure.
The relevant issue is not deviation alone, but the economic nature of that deviation.
Beta Sensitivity Matters
This suggests a useful hierarchy.
- Strategic risk is the market risk deliberately embedded in the reference portfolio.
- Active risk is the departure of the total portfolio from that strategic anchor.
- Factor exposure describes the economic nature of both.
The factor framework does not replace every other risk measure.
Liquidity, leverage, concentration, collateral requirements and counterparty exposure cannot be reduced to growth, inflation and real-rate betas. They remain important structural risks in their own right.
The purpose is not to explain every investment outcome with three factors.
It is to describe the portfolio’s major economic sensitivities in a way that is useful to the institution.
From Factor Definition to Portfolio Construction
The sequence matters.
- First, determine which economic risks matter to the mandate.
- Then define those factors carefully.
- Translate them into factor-mimicking portfolios using liquid market instruments.
- Use those return series to map asset classes and observe how their exposures change through time.
- Express the reference portfolio in the same factor language.
- Then compare the implementation portfolio with that strategic anchor and determine whether the deviations are intentional and adequately compensated.
The progression is:
At that point, factor thinking becomes more than a conceptual framework. It becomes a practical investment tool.
And Then Asset-Class Labels Begin to Matter Less
Once different investments can be expressed through the same economic-factor framework, some traditional asset-class boundaries become less fundamental.
- Private equity is still equity.
- Private credit is still credit.
Their implementation structures, valuation processes, governance requirements and liquidity profiles may differ considerably from their public-market counterparts.
But much of the underlying economic risk may be familiar.
Once that common economic exposure is visible, “public or private” becomes a second-order question.
The first-order decision is which economic risks the institution wants to own, what compensation it expects for bearing them, and how much exposure is appropriate.
Only then should it decide which implementation offers the best trade-off among return, liquidity, cost and governance.
That is where I will turn next.
References
Lamont, O. A. (2001). “Economic Tracking Portfolios.” Journal of Econometrics, 105(1), 161–184. https://doi.org/10.1016/S0304-4076(01)00074-4.
Pukthuanthong, K., Roll, R. W., Wang, J. L., & Zhang, T. (2019). “A New Method for Factor-Mimicking Portfolio Construction.” SSRN Working Paper No. 3341604. https://doi.org/10.2139/ssrn.3341604.
Pukthuanthong, K., Roll, R. W., Wang, J. L., & Zhang, T. (2019). “A New Method for Factor-Mimicking Portfolio Construction.” SSRN Working Paper No. 3341604. https://doi.org/10.2139/ssrn.3341604.
Pukthuanthong, K., Roll, R., & Subrahmanyam, A. (2019). “A Protocol for Factor Identification.” The Review of Financial Studies, 32(4), 1573–1607. https://doi.org/10.1093/rfs/hhy093.
Fang, X., Liu, Y., & Roussanov, N. (2026). “Getting to the Core: Inflation Risks Within and Across Asset Classes.” The Review of Financial Studies, 39(3), 702–743. https://doi.org/10.1093/rfs/hhaf050.
Fang, X., Liu, Y., & Roussanov, N. (2026). “Getting to the Core: Inflation Risks Within and Across Asset Classes.” The Review of Financial Studies, 39(3), 702–743. https://doi.org/10.1093/rfs/hhaf050.
* Images generated using ChatGPT




Comments
Post a Comment