Q: What is the history of the fund? How has it evolved over the last decade?

A: The Core Equity fund was launched in 1929. Of course, it has evolved over time as managers moved in and out. But we have had the same quantitative investment process since 2013, when I took over the management.

We used to have a fundamental team as well, but now we are a purely quant fund with no fundamental analysis overlay. That made the fund more disciplined, because the quantitative process requires more discipline by its nature. In terms of style, this is a large-cap blend or core fund. It doesn’t have any growth or value bias and is not going into names or characteristics that differ from the universe that it operates in.

Q: What core beliefs drive your investment philosophy?

A: We realize that the large-cap core category of funds is a core holding for most individual investors, often representing 35% to 40% of their portfolios. That’s why we have to minimize the risk and achieve steady performance. We are active investors, who aim to deliver performance in the top decile, but the important aspect is that we do it in a risk-controlled manner.

The correlation with other funds is remarkably low and that’s evidence that the fund substantially differs from its peers. We are happy with our risk-adjusted numbers and the performance relative to other managers in the large-cap space.

Q: What differentiates the fund from its peers?

A: The key difference is in the stock selection and portfolio construction process. We have an intuitive and highly disciplined approach resulting in a portfolio that is true to its benchmark. In stock selection, we use 35 dynamic, industry-specific, multi-factor models, designed to perform over a 12-month investment horizon. They change over time because factor effectiveness changes.

In portfolio construction, we make conscious effort to control the tracking error and to adhere to sector neutrality. That’s an important differentiator because a manager can be a good stock picker, but if he or she doesn’t construct the portfolio in a reasonable way, there will be unintended risks that may eat away the stock picking ability. From my point of view, portfolio construction is as important as stock picking. These aspects make us distinct and lead to low correlation with many large-cap funds.

Q: How do you develop the multi-factor models? Could you explain the process?

A: The factors themselves are not unique; they are commonly used by stock researchers, but we have a distinct way of transforming and using them. In addition, we use academic papers and conferences to find factors that might lead towards performance. Overall, we compile a list of factors that might have potential and assess if they work on a standalone basis over a 12-month investment horizon.

Importantly, we examine if they should be tested across the entire large-cap space or are industry-specific factors. It makes sense that factors that drive performance in biotechnology, for example, might differ from the factors that drive performance in utilities. We separate the investable universe into industry groups and have 35 industry clusters. Recognizing that each of these clusters would be driven by different factors, we test the factors for each of them going back to 1989. The idea is to see which factors have the ability to pick stocks over a 12-month investment horizon.

The next step is combining the factors that have stock-picking ability to create a multi-factor model that works better than any of the individual factors alone. Based on the data, we look for the best combination of these factors. That becomes our multi-factor model for the specific industry. We have 35 multi-factor models or one model for every industry cluster.

These models are specific not only to the industry, but also to the 12-month investment horizon. The models evolve based on our continued back testing because the factors that were working years ago may no longer work today. It is intuitive that models should evolve as the business changes. Contrary to static models, we have an ongoing, dynamic process. That’s another differentiator.

Finally, when we construct the portfolio from the stock selection model, we use the rankings for individual names within each of the 35 industries. We know how those rankings performed in the past and, on that basis, we form a view on the expected outperformance for every stock in the universe. Then we use an optimization tool to create a sector-neutral portfolio, where the tracking error is controlled and is one of the lowest in the active investment large-cap space.

Q: What is the rationale behind using a 12-month investment horizon?

A: A shorter investment horizon would result in higher turnover. The factors that drive performance would be different. We view the 12-month period as a relatively long-term horizon and we think of ourselves as long-term investors. We are not market timers or day traders. With the same process, we have been managing individual mandates for 18 years. The 12-months horizon has positive tax consequences and is preferential in terms of capital gain considerations.

From a fund perspective, it works towards reducing turnover and is consistent with the longer-term view of the world that we and most investors have. However, we believe that the fund should be evaluated on the basis of business cycles, or over the period of three to five years.

Q: What factors do you include in the stock-selection models?

A: Our list includes about 50 factors. They could be traditionally used by stock analysts or could come from academic research. Most of them are common factors like valuation, book-to-price, net margin, etc. Some are specific for an industry, such as revenue per passenger for the airlines or tier-1 capital ratio for the banks. We back-test these factors to see which of them seem to work for the particular industry.

We scale every factor by peer groups as well. If I like Microsoft because it has a high book-to-price ratio compared to the universe, does that make Microsoft attractive? Yes, if I consider the book-to-price to be an attractive attribute in the large caps. Probing further, if the book-to-price is higher than the average for the large-cap universe, but below the average for its industry or sub industry, what is the conclusion? It is all relative to the peer group and the comparisons are important.

Overall, we transform each of these 50 factors by variously defined peer groups based on the S&P classification system. For instance, we evaluate the performance of book-to-price scaling it by differently defined peer groups and relative to the sub-industry, industry, industry group and sector. If we apply book-to-price only to the US large-cap space to evaluate performance, we would have industry biases. But when we drill down to the sub-industry, we avoid these biases.

The correlation between these two factors, which are both book-to-price but transformed differently, is about 0.55. That illustrates that these factors have become remarkably different. For comparison, the correlation between asset classes like U.S. large-cap equity and emerging market equity is about 0.8. We still consider U.S. equities to be significantly different from emerging market despite the high correlation.

In our model, by virtue of transforming the same factor by differently defined peer groups, we have turned it into a totally different animal. One of our distinctive features is testing a number of factors and transforming them. The factors that seem to have impact are used to make the best combination of factors.

Q: How do you define “the best combination of factors”?

A: It is the combination that leads to the most stable alpha in the specific industry. Through optimization, we pick a smaller subset of factors from those that seem to work. For example, there are three versions of analysts’ estimate revisions – revisions over the past one month, three months and six months. When the past one-month estimate is an important factor in an industry, there are good chances that the stocks that ranked high based on this factor, will also rank high based on the past three-month and six-month estimate revisions. In some cases, the top three factors on a standalone basis would be these three estimate revisions. Nevertheless, because of the overlapping information, we would pick only one of them. The other two aren’t that relevant because they don’t give much incremental information.

Overall, the best combination of factors is not just the top performing factors; it is the best in terms of the correlation of factors delivering performance. Essentially, that’s our multi-factor model for each particular industry. We may have an industry where valuation accounts for 28% of the weight, profitability accounts for 33% and other factors account for the rest. It is a continually evolving, dynamic process, where we keep revising our models.

Q: Can you illustrate the process with some examples?

A: In pharmaceuticals, valuation is an important attribute that accounts for 52% of the weight in the multi-factor model, while sales growth accounts for 15%. Interestingly, profitability doesn’t have a big role. In fact, over the 12-month horizon, profitability turned out to be a negative attribute lately. Part of the explanation for this paradox may be that highly profitable companies attract more competition and are unable to sustain that profitability in the future.

That refers to other industries as well. Being highly profitable today in some industries is, in fact, linked to worse future performance. This may seem counterintuitive, but in certain industries the low barriers to entry attracts competition and takes away the advantages of currently highly profitable and high-growth companies.

In the case of pharmaceuticals, many companies hold patents, but these patents have a finite life. They might have high profitability today, while they have patent protection, but in the future they are likely to face competition from generic drug makers.

Regarding the equity chain, stock issuance and repurchases, the academic theory works in two ways. A company that issues stock might be sending a signal that it has good ideas and projects, so it needs capital to invest with high internal rate of return. In that case, issuing stock is a positive factor. On the other hand, companies that consider their stock to be overpriced may want to take advantage of the high price today and issue stock. Those are two conflicting things.

For example, stock issuance in the financial industry, particularly by capital market companies, is negatively related to their subsequent performance. In other words, the financial sector issues stock when it thinks the stock price is high and that becomes a negative signal. In capital-intensive industries like transportation, for example, stock issuance seems to be related to positive future performance. Such industries tend to issue stock when they have profitable projects that require capital investment.

These are examples of how the same factor shows different signs, depending on the industry. It is a nuance that would be missed if we don’t drill down to the industry level, because there is substantial difference in factor behavior between the industries.

The financial sector has always been difficult to model. During the 17 or 18 years of experience in wealth management, our performance attribution process has revealed outperformance in nine of the 11 S&P sectors. We haven’t done well in energy, but we have done really well in financials and information technology. Banking was difficult for us to model until we found that the tier 1 capital ratio is a surprisingly potent factor for future performance. It is a relatively newly available factor and the models have evolved accordingly.

Q: Regarding portfolio construction, do you also consider separate benchmarks for each industry?

A: No, our benchmark is the Russell 1000 Index and we use it to evaluate the performance of the entire portfolio. We aim to construct the best possible portfolio using the portfolio construction optimization tool. We define “best” as a sector-neutral portfolio of 90 to 100 stocks with tracking error of about 200 to 250 basis points, that has the highest expected alpha. That is how the capital gets allocated to each of the names.

Sector neutrality is important, because the entire process is geared towards picking the best stocks, not predicting which sectors will outperform. As a result, the sector weights cannot differ by more than 10% from the sector’s weight in the benchmark. If a sector represents 15% of the Russell 1000, its weight in the portfolio should be between 13.5% and 16.5%. There is no sector with double or zero weight and we maintain sector weights that are consistent with the benchmark. We have no reason to believe that we are good at predicting which sectors would outperform and we don’t want to make random bets and take random risk.

In the portfolio construction process, we allocate the weight to each of the names. We use Barra, a portfolio optimization tool to build a portfolio with beta of 0.99 to 1.01 to the benchmark and tracking error in the 200 to 250 basis points range.

We have limits of 225 basis points for individual active bets, because we don’t want to just load up one name. If a stock has a 300 basis point weight in the benchmark, we would be no more than 525 basis points in that name. It is a diversified portfolio with one of the lowest tracking errors in the space.

Q: Would you describe your sell discipline?

A: Selling is an optimization-based decision; there is nothing subjective about it. We produce the optimal portfolio each month and it may differ from the portfolio last month. The optimizer uses the most recent inputs and, based on the combination of the relevant metrics, we produce rankings. That data is fed to the optimizer today. If a stock that was attractive in the past becomes unattractive and is expected to underperform, it may lead to selling if it has been held for at least 12 months.

Twelve months is our investment horizon and we don’t want to compromise it. So, we consider selling names that are in the portfolio for at least 12 months. The optimizer reacts if the expected alpha becomes unattractive, if the volatility in a name increases or if it affects the portfolio’s tracking error. We may sell a name if it contributes to violating the sector guardrails. In the same manner, we would buy a name when it shows attractive expected alpha and risk characteristics, as well as suitable correlation with the other names in the portfolio.

Q: How do you define and manage risk?

A: Total risk is important at the overall level of an investor’s wealth and for the allocation to different asset classes. However, when investors have already decided to allocate a certain amount to U.S. large-cap stocks, the key is being true to the benchmark and focusing on relative performance. The risk of the portfolio should be seen in the context of the benchmark associated with the asset class. Therefore, we see tracking error, not total risk, as a proper measure of risk.

The tracking error is related to outperformance as well. If it is low, we can’t outperform or underperform by a huge amount.  In our terms, the tradeoff between risk and reward is the tradeoff between tracking error and expected outperformance.

Let’s say that investors, based on their expectations, decide to allocate 40% of their money to the US equity markets. If I, as a fund manager, am bearish on the US equity markets, I can keep 30% in cash. In that way only 10% of the investment goes into equities, even though the investor was expecting an allocation of 40% for that market. So I would be undermining the investor’s asset allocation process and that would be wrong on my part.

Keeping cash would be misleading because it would show lower total risk, but would increase the tracking error, which is the right metric to judge the portfolio’s risk. That’s why total risk is not the right risk measure for our fund; it is the tracking error to the Russell 1000 index.