An In-Depth Look at the Selection Process

GETTING TO THE HEART OF YOUR STOCK PICKING PROCESS

Many equity investors consider themselves to be stock pickers, and they generally describe their portfolio construction as a “bottom-up” approach. Their selection process typically begins with a multi-level screening that eliminates stocks with inappropriate or unattractive characteristics, and then shifts to identifying favorable characteristics. This is often represented visually as a multi-level “funneling” effect, that begins with a large universe (perhaps the performance benchmark) and eventually produces a manageable universe of candidates for the portfolio. These stocks reflect the qualities and factors that define the investor’s value-added approach.

The next step is to perform in-depth analyses (perhaps both fundamental and technical) to identify the most favorable investments. Depending on the strategy, there may be a mix of factors at work, expecting the result to reflect the broad characteristics of the performance benchmark, but with a higher return. It is the selection of these individual stocks that is key to the expected outperformance.

The performance industry calls the excess return generated by this process the “stock selection” effect. This is separate from the decision to increase or decrease short-term exposure to the economic sectors that make up the benchmark. That decision (the “allocation effect”) is the context in which these stocks are selected. The premise of this performance evaluation method is that any short-term changes in the structural aspects of performance are addressed in the allocation effect, leaving a purely idiosyncratic return that is attributed solely to stock selection.

But is this true? Is performance attribution really that simple? Or might there be more to this problem?

In this paper, we will dig deeper into the question of where assets are allocated, and how this affects performance. We focus on understanding the characteristics of these economic sectors. Our analysis goes beyond simply noting the weighting in each sector - we also examine how accurately the portfolio’s stocks represent these sectors in the portfolio.

This analysis requires a deeper understanding of the performance benchmark, and a more precise evaluation of how well the portfolio demonstrates the characteristics of each economic sector. We can already see that significant insights (and surprises) await, as we examine the portfolio’s characteristics, relative to its benchmark.

UNDERSTANDING THE BENCHMARK: KEY TO PERFORMANCE ANALYSIS

Our case study involves a large cap equity portfolio, and our benchmark is the 200 largest stocks in the S&P 500.

While many consider the S&P 500 a proxy for large company stocks, it was originally intended to represent the entire US economy – and so it includes mid and small companies. To avoid the problem of benchmark mismatch with our portfolio, we use a benchmark that includes only larger companies; this is a better fit with the portfolio’s stocks.

The largest stock in the S&P 500 is 250x the size of its smallest stock. This degree of issue concentration is not representative of prudent standards of diversification. We try to minimize this problem by selecting only the largest stocks for our benchmark, while ensuring that the benchmark contains enough issues to represent all the economic sectors (some of which hold smaller issues.) Our customized benchmark reduces the stock concentration issue by 70 percent (from 250x to 75x.)

There are 11 economic sectors in the US economy. The disparity in size between the sectors reflects the size of the publicly-held stocks found in those parts of the economy.

The benchmark earned a 1-month return of 1.85 percent, led by the Consumer Discretionary sector. The greatest detractor from performance was the Healthcare sector, which lost 7.55 percent. (Going forward, we present these sectors in alphabetical order.)

UNDERSTANDING THE BENCHMARK: KEY TO PERFORMANCE ANALYSIS

The stock portfolio maintains a neutral position in terms of its sector allocations. Since its sector allocations mirror the benchmark, there are no active sector allocation bets, nor any allocation effect in the performance attribution analysis.

The portfolio also reflects a high-conviction strategy of concentrated holdings. It currently has positions in 12 securities that provide exposure to every sector. Of course, there is a dramatic difference in active share relative to the benchmark; this is the case for all concentrated strategies. Each sector holds a single security, except for the Technology sector, which splits its exposure equally across a pair of securities. (Given this sector’s 33% weighting, it is necessary to diversify the holdings to reduce issue concentration risk.)

The portfolio outperformed its benchmark by 346 bps over the 1-month performance period, earning 5.32% vs 1.85% for the benchmark. This is noted in the standard performance attribution analysis provided here.

The portfolio’s returns were higher than the benchmark in seven out of 11 sectors, with outperformance as high as 1,000 bps in the second-largest sector, and equally-substantial underperformance seen in one of the two smallest sectors. The strongest contributors to excess return were Communication Services and Technology, which account for nearly half the portfolio’s market exposure and 70% of its excess return.

LOOKING DEEPER INTO STRUCTURE OF ECONOMIC SECTORS

The economy’s 11 economic sectors are broken down into 32 Industry Groups. For example, the Healthcare sector is divided into four Industry Groups: Research, Drugs, Healthcare Services and Medical Devices.

These Industry Groups are then divided into 72 Industries. In the Healthcare sector, the Research group is divided into two industries: Biotechnology and Diagnostics & Research.

The benchmark’s 200 securities are spread across these 72 industries. 

EXAMINING THE EFFECT OF ALLOCATING TO INDUSTRY GROUPS

The portfolio focused on the macro call of identifying the best industry group within each sector.

For example, in the Consumer Discretionary sector, the portfolio’s exposure was dedicated to the Automobile industry group, with a return of about 44% as compared to 11% for the overall sector.

The portfolio’s Industry Group allocations produced a return of nearly 8% as compared to the benchmark allocations, which produced a return of less than 2 percent. This focus created an allocation-based excess return of 6.08 percent.

THE EFFECT OF ALLOCATING TO INDUSTRIES

The portfolio also focused on allocating to the best Industry within each Industry Group.

An example is the Entertainment industry (within the Industry Group of Entertainment and Internet Communications.) This earned 16.32% which outperformed its industry group by over 1,000 bps. Less than half of the industries outperformed their industry groups, but the cumulative effect was a return of 10% which added 207 bps of additional excess return.

The combined effect of the portfolio’s targeted allocation to industry groups and to industries contributed substantially to excess return, as shown below:

UNBUNDLING ALLOCATION FROM TRADITIONAL ISSUE SELECTION

Having evaluated the effect of the hierarchy of allocation decisions at work, we now have the appropriate context for evaluating the success of our stock selection process.

The industries and industry groups within each sector are structural factors; these have a substantial effect on market exposure and excess return. Even if one were to argue against these industry exposures as decisions (calling them “byproducts” of stock selection) we still must recognize that these factors are drivers of excess return.

This is a matter of allocation, rather than issue selection.

As we see below, we must evaluate each stock return relative to its industry, and not to its economic sector.

Eight of our twelve stocks underperformed their industries, with two stocks matching their industries (because their industries only contained a single stock.)

A proper evaluation of our results would be:

“We picked underperforming stocks in the best industries and industry groups.”

Yet, the traditional attribution approach would consider this a triumph of stock picking:

We contrast the traditional, non-hierarchical attribution approach with one that recognizes the industry and industry group factors operating within each sector:

We cannot assume that every portfolio contains the identical structure of its benchmark sectors. When we do, we mistake allocation success for stock selection success. In this case study, ignoring the structural drivers of performance produced an error of 815 bps, mistaking a negative 4.68% stock selection effect for a positive contribution of 3.46% to excess return.

A REVERSAL OF FORTUNE

What if we reversed our circumstances, and selected outperforming stocks in underperforming industries?

We maintained the same high-conviction stock selection process, but applied this to the lowest-returning industry groups and industries within each economic sector (while maintaining a neutral exposure to each sector.)

In this second case study, the stock selection effect becomes overwhelmingly positive, accompanied by underperformance from the industry selections. The portfolio’s total return declined from 5.32% to 2.61% with a corresponding decline in excess return from 346 bps to only 76 bps.

In the performance summary table below, we see that the combined allocation effects produced 555 bps of underperformance that was more than offset by 631 bps of outperformance from the true selection effect. Almost tree-quarters of this “selection alpha” came from the portfolio’s two largest stocks.

Once again, we see the critical importance of calculating selection benefits by comparing each stock’s return to its industry peer group, rather than to its overly-broad economic sector.

INSIGHTS INTO ISSUE SELECTION AND PERFORMANCE ATTRIBUTION

When we ignore risk and structure, we inevitably misrepresent our skill in selecting investments.

This is unfortunate, given the attention paid to the stock selection process, and the substantial resources dedicated to investigating the merits of individual stocks. This problem becomes even more severe with concentrated, “high conviction” stock portfolios.

Are we being unduly harsh here? Would it be fairer to say that the attractiveness of certain stocks is partly due to enthusiasm for their industries? If so, then the performance attribution process must incorporate these structural factors, and then give credit to investment analysts for their industry recommendations, while holding them accountable for the stocks they select.

The purpose of performance attribution is to identify the decisions and factors that drive performance, and then measure the benefits that they deliver. Allocation is a key decision, and we demonstrated that a proper analysis of sector allocation requires a “drill down” into the exposure to the industry groups and industries that define these sectors.

Our approach identified the three levels that make up the portfolio’s economic sectors. After identifying differences from the benchmark, we measured the contributions that these differences made to the portfolio’s performance. The evaluation of true stock selection benefit is made at the industry level, with a one-to-one correspondence between a stock and its peer group.

“What’s in your selection effect?”

If there are unidentified structural factors at work in your portfolio, then your performance evaluation is likely to be delivering an incorrect assessment of your decision process - and the results it produces. As we noted from these two case studies, your stock selection may be substantially better (or worse) than you believe it to be. It is best to know this.

Shouldn’t your performance attribution process provide these insights?

.

.

.

Written in partnership with Stephen Campisi