This article is targeted towards actuaries and risk managers working at insurance companies. This article must not be construed as investment advice or product recommendation.

The recipe for building smart beta index is quite straightforward. Here it is:

Step 1: Find a stock picking formula.

Step 2: Apply the stock-picking formula to past market (stock) data and create a historical chart of the index.

Step 3: If this newly built index is going up, slow and steady, and has low correlation to the S&P 500 Index, you are successful. Otherwise, repeat Steps 1, 2 and 3 again.

Let’s examine the steps in detail.

Step 1

The stock picking formula needs to be broadly systematic and not discretionary. There are many well-defined formula already available to index designers that can be modified slightly and used. Here are a few popular formulae:

Factor Models: Factor models are derived from a mathematical framework known as Capital Asset Pricing Model. It is simply a method of selecting stocks which have particular qualities such as low size (small market capitalization vs high market capitalization) or low price (cheap v/s expensive).

Risk-adjusted return: This is a very popular framework and is also known as mean-variance optimization. Investors essentially are looking for stocks with low volatility (low risk) and high returns (high quality). Here, volatility can be replaced with another measure of risk. Determination of ‘low risk’ and ‘high quality’ can be a matter of perception and may vary a lot with the specific market and the investment thesis. One investor may consider a high rate of reinvestment of profits as favorable, while another investor may consider it as unfavorable and may want to receive dividends instead.

Thematic or sector-specific formula: Often, investors find themselves drawn to a particular industry when that industry has a great potential of future profits. Examples are biotech industry, renewable energy, climate technology, or internet technology.

Typically, the stock picking formula will determine its rebalancing frequency. The parameters of the formula may become stale quickly (within days) or may persist for months. So, the formula needs to be ‘reapplied’ to the most current stock data on a periodic basis, leading to rebalacing (refresh) of index constituents.

Step 2: Back-casting

Now, the formula devised in Step 1 is applied to the stock market data, which results in a ‘back-casted’ index. In other words, hypothetical historical values of the index are calculaed by applying the formula to the stock data. This back-casted index is then compared to the broad market index and evaluated for its suitability as a smart-beta index.

Step 3: Analysis of the new index

It is generally presumed that if the (hypothetical) past price returns of the newly built index have the needed qualities of smart beta, viz, 1) stable returns and 2) low correlation with the board market, then such qualities may continue into the future. While past performance is not a guarantee of future outcomes, past performance is still the best indicator we have for the utility of a newly developed smart beta index.

If the newly developed smart beta index does not sufficiently fulfil the requirements of providing stable returns and low correlation, we go back to Step 1 and repeat the whole process.

Key Considerations

Overfitting

Occasionally it happens so that the formula obtained in Step 1 fits perfectly to the past stock data by pure coincidence, and creates a perfect outcome. Brute-force testing of all possible permutations of formula parameters often leads researchers into this trap of finding such over-fitted parameters. Therefore, index designers need to run the stock-picking formula through rigorous variations and scenario analysis to ensure that there is a low chance of overfitting.

Broadly Systematic

The formula needs to use quantitative criteria for stock selection, so that the formula can be applied to the stock data systematically, without much manual discretion. For example, ‘quality of CEO’ is a criterion that is difficult to apply systematically. On the other hand, ‘Price-to-Book value’ ratio is a criterion that can be easily applied to stock data systematically.

Minimize transactions

Ideally, at every rebalancing event, the new portfolio weights should differ only slightly from the current (stale) weights. If on the other hand, the new weights are very different from current (stale) weights, or perhaps entirely different set of stocks get picked in the rebalancing, then too much stock trading will need to be done by index managers. These stock transactions will dramatically increase the index manufacturing and hedging costs.

Transactions also depend on rebalancing frequency. Once a month is considered a good enough rebalancing frequency. Daily rebalancing might cause too much trading, while annual rebalancing would be too slow.

Prefer stocks with high liquidity

Index managers want to pick from stocks that have high liquidity. In general, stocks with higher market capitalization have higher liquidity – meaning, they trade more often on the stock exchanges and on liquidity pools, and hence are easier to buy and sell.

Prefer stocks with low dividend yield

This may be counterintuitive, even for experienced investors. But the key thing to notice here is that the index payoffs earned by retirees is tied to the gain in price of the stock (and not to dividends). Let’s understand in detail.

If the business is profitable, the profits will accumulate inside the business, thereby elevating the value of the business and also the stock price. If instead, the business decides to pay out the profits to its shareholders via dividends, then the stock price will go down after the dividend payout. Most indices track the gain in price and exclude dividends. So profitable stocks that pay high dividends will have lower price return. Hence, policyholders prefer indices that are built with stocks that have low dividend yields.

At Annuity Risk, we help insurers pick the right smart beta indices from an ocean of available options. We offer unbiased advice on the suitability of the index considering insurer's overall product portfolio, hedging costs and the risk management framework. We stand to gain nothing from favoring one manufacturer or component over another.