1.4.1 BMU & PEU determination

Determining the universe of private companies is a combination of assessing private companies' likelihood of being regarded as investable and their eligibility to be investable. For estimating the likelihood we focus on the characteristics of private companies that resemble those that private equity funds have invested in the past. To determine the eligibility, we lay down the rules that make the companies investable. These criteria when combined with the index methodology lead to three distinct buckets of private companies as shown below:

Broad private Market Universe or BMU

The constitution of the broad private market universe relies on simple rules that make a private company eligible for investment. These rules include:

1. The company should not be dissolved, liquidated, or declared bankrupt at the time of consideration.

2. Is a for-profit private company with a capital structure that does not have any security (equity or bond) traded publicly at the time of consideration.

3. If it is a subsidiary of another private company, only the parent is eligible for inclusion provided their financial information is available.

4. Is not entirely government-owned or controlled.

5. Has non-negative sales with the average sales in the past being over USD 1 million and at least two fiscal years of financial accounting data that is accessible.

6. Has a business and industrial activity description.

7. Exclude companies that are specifically available to be included in the infrastructure indices of inframetrics.

8. Information on the key factors that are used to price the private company using our factor model approach is available. Missing information is imputed provided there is sufficient information in the sector, year, and country available and the factor being imputed is not one of the primary factors affecting its valuation.

Private companies that pass these filters are considered to be part of the broad private market universe.

Private Equity-backed Universe or PEU

How do private equity managers select their targets? Not all companies can be suitable to be professionally managed. Some may lack professionalism. Some businesses may not be scalable, and some may lack the profit margins that allow private equity-owned transformations. Thus, we attempt to focus more on companies that can accommodate PE ownership strategies, rather than focus on all the private companies in the privateMetrics® database.

Size Filters

To implement size filters, we first take a look at the distribution of a large number of transactions in private equity markets. Below are the histograms of raw observed prices in USD $ millions and the log-transformed prices. We can observe that both the distributions are truncated at the left due to prices being strictly positive. Also, the raw prices resemble an exponentially decaying distribution, whereas log-transformed prices resemble somewhat of a normal distribution.

^{Histogram of Transaction Prices in Private Markets}

To understand whether the distribution in estimated shadow prices is drastically different, we replicate such plots with all the priced companies, dropping outliers at the top and bottom 1 percentile. Note that these are estimated shadow prices in the universe of private companies put together in the privateMetrics® database, where the estimation is based on factor prices that sufficiently explain the variation in transactions, projected onto the financials of the companies in the privateMetrics® database. The plots reflect a similar distribution. However, the log scale shows a bimodal distribution, which is unsurprising given the variation in priced companies across countries. The notable takeaway from this exercise is that there exists a large proportion of priced companies in the privateMetrics® database that are smaller than the typically observed/modeled transaction in PE markets, i.e., the mass of firms below 0 in the log scale. Thus, any size treatment should ensure that these “small” companies are excluded.

^{Histogram of Estimated Shadow Prices}

To achieve, that we filter out the prices in the shadow universe based on the distribution characteristics of transaction prices. Specifically, using the mean and standard deviation of the log of transacted prices, we compute the range to be within μ ± 3.72σ, which roughly spans 99.99% of transactions. In terms of raw prices, these filters result in excluding private companies with prices approximately below USD 20 million and above USD 21 billion.

Profitability Filters

Performing an exercise similar to size filters is not feasible for profitability as private equity focuses on acquiring companies that are ripe for transformation, thus not shying away from unprofitable companies. So outliers present opportunity rather than discouraging PE. Thus, we focus on mapping the profitability profiles of the two samples, i.e., the transactions and the privateMetrics® database.

To get a sense of the profitability profiles in the two samples, we segment the ratio of EBITDA to Sales into deciles in each sample and compute the average within each decile. The results are produced in the below plot. Ignoring the outliers in transactions, what stands out is the steepness in how profitability increases as we move across deciles in the Transaction sample, whereas it is quite flat in the privateMetrics® database.

In terms of outliers, the privateMetrics® database is fairly well-behaved, even in contrast to the transactions sample. However, there are a large number of low profitable companies that are dragging the slope of the curve above, and these warrant being excluded.

In other words, there are a lot of “boring private companies”, which Private Equity managers are never going to consider. Once these are excluded, the slope of the privateMetrics® database will be similar to that of the transactions sample.

Furthermore, given that profitability varies systematically with sectors, we propose to identify boring companies in each sector so that the average profitability across deciles in that sector in the privateMetrics® database resembles that in transactions in the sector. This is made clear in the below plot, where repeating the exercise by PECCS™ industrial activity, produces differences in how much the profitability profiles deviate by sector.

Now to achieve a profit profile that resembles that of the transaction sample, two key steps are required:

1. Filter: Identifying an appropriate filter that can move the profit profile towards the transaction sample for each sector.

2. Metric: Computing a metric that can summarise the extent of alignment of the two curves, ignoring the outliers.

For step 1, we experiment with filters beginning from various starting points and ranging for different bandwidths. Given the evidence in the full sample, we play around with filters ranging from -20% profitability to +15% profitability and alter bandwidths in increments of 1 % till a maximum bandwidth of 15%. This gives rise to several combinations of filters which when applied to the universe allow us to compare the profit profile with that of the profit profile of transactions.

For step 2, we simply sum up the squared differences between the two curves, ignoring the first and the tenth decile. We call our metric SSD or the sum of squared differences in profitability between the two samples, and we proceed to minimize the SSD in each sector. SSD is computed as below:

where μ is the average in each sector i of the profitability Profit measured as the EBITDA to sales ratio, within the decile (d) defined by the same profitability ratio in the two samples, namely the Deals (or transaction) sample and the Universe (or privateMetrics® database), respectively. We exclude the first and the tenth decile in the sum to ignore the effect of outliers in the Deals sample. For the Universe sample, the mean takes into account the filter specifications by excluding companies whose Profit falls within the range of ft_st or filter starting value and ft_st+bw or filter starting value plus the bandwidth of the filter.

Finding the optimal filters then becomes a minimization problem, where we identify both a global minimum, which is the minimum SSD value across all combinations of filters, and several local minima, which are the minimum SSD values for each bandwidth of the filter. Local minima are useful substitutes as in some sectors, the global SSD minimum might leave out too few private companies in the database, in which case we can choose one of the local minima of SSD.

The results of these optimizations at the sector level are reproduced in the images below. The z-axis represents the SSD value while the x- and y-axes correspond to the filter starting and ending values. The plotted plane represents the values of SSD that correspond to each filter with the color of the plane ranging from dark to light with the value of SSD. The red dot indicates a global minimum SSD for each sector and the black dots represent the local minima at each filter bandwidth for each sector.

Thus, this approach leads to the identification of several filter combinations at the sector level for EBITDA/S, which can lead to a profit profile of the privateMetrics® database that almost resembles that of the transactions data in private markets.

In conclusion, we choose a filter that balances coverage requirements with having a low SSD, leading to the identification of the PEU or Private Equity-backed Universe.