Don’t put all your money into a single asset.

But there’s a deeper problem.

Assets that seem different from one another in normal times may start moving together when markets come under stress.

When the economy shifts, interest rates rise, or risk appetite wanes, the relationships between stocks, bonds, currencies, gold, and high-risk assets can change.

This raises an important question in portfolio management:

Is it enough to diversify our investments once, or should a portfolio’s allocation change as market conditions change?

That is the question a recent paper on arXiv set out to explore by building a multi-asset portfolio that adapts to what are known as market regimes.

What is a market regime?

Markets don’t move at the same pace or carry the same risks all the time.

There may be a period of growth and stability, followed by a sideways market, then a downturn accompanied by rising volatility.

The study divided the market into three regimes:

Bullish, neutral, and bearish.

The researchers used a statistical model called a Hidden Markov Model to try to identify the prevailing regime based on information such as stock returns, volatility, bonds, and currencies. Crucially, they used only information available at the time each decision was made, to avoid using hindsight in the test.

Why might diversification change?

Because correlations between assets are not constant.

Two assets may move independently under normal conditions, but during a liquidity crisis, investors may start selling both at the same time.

This means that a portfolio that looks diversified on paper may become less diversified just when you need it most.

That is why the study did not use market regimes to forecast future returns. Instead, it used them to estimate the risk structure and relationships between assets differently for each regime.

The portfolio included five types of assets in Indonesia: government bonds, stocks, gold, currencies, and Bitcoin.

What did the portfolio do when risks rose?

In the regime the model classified as bearish, the average weight of stocks fell to around 6.2%, while gold’s weight rose to about 28.5%, and currencies accounted for around 39%.

Bitcoin’s weight remained very low in all three regimes because of its high volatility.

The point is not that gold or currencies will always rise during crises, but that the model sought to redistribute risk based on the relationships it had observed historically.

Did the strategy work?

The answer calls for some caution.

During the test period from 2021 to August 2025, the proposed strategy recorded a maximum drawdown of 4.44%, compared with 15.27% for the equal-weight portfolio and 21.17% for the buy-and-hold strategy.

But in exchange for this protection, the strategy delivered an annual return of only around 1.14%, compared with 12.97% for the equal-weight portfolio and 14.48% for buy-and-hold. The simpler NCO model, which does not use market regimes, actually had a lower maximum drawdown of 2.19% and a higher annual return of 5.71%.

This illustrates an important rule:

Reducing losses isn’t free.

Sometimes it comes at the cost of giving up part of the upside.

When can frequent adjustments become a problem?

The study itself found another issue.

The model changed its market-regime classification roughly every month during the test: 55 transitions over 56 months. This led to high portfolio turnover and additional trading costs. The researchers also acknowledged that some regimes’ characteristics changed over time. In fact, the regime classified as “bullish” later recorded a negative average stock return during the test period.

This is a very important point.

Markets don’t read books, and a model trained on what a “bull market” looked like in the past may find that future conditions are different.

Diversification is more than a list of assets

The most important lesson from the study is not that investors should change their portfolios whenever a statistical model changes.

Rather, true diversification is about how assets relate to one another, not just how many you own.

Owning five assets does not necessarily mean having five independent sources of risk.

So, when building a portfolio, the better question may not just be:

What do I own?

But:

How will these assets behave together as market conditions change?

The goal of a portfolio is not always to achieve the highest possible return, but to find the right balance between growth, potential losses, the cost of adjustments, and the investor’s ability to tolerate risk.

Disclaimer: This material was prepared under the supervision of a Yamama Insights editor with the assistance of AI tools for financial education purposes. It is not a recommendation to buy, sell, or hold any security, and expresses the views of its authors, not those of the platform.