Inference

Overview

When portfolio holdings are not available, some components use inferred logic to predict what a portfolio’s holdings would be on a particular date. The two inference options are as follows:

•Infer forward and

•Infer backward.

Inference is available in the following components:

•Equity Attribution

•Allocation components

A buy-and-hold strategy is assumed and the weights are return-adjusted. Forward inference uses holdings from a date prior to the date of the missing holdings information. Backwards inference uses holdings from a date after the date of the missing holdings information.

What is the purpose of inference?

In a perfect world, a meaningful attribution portfolio analysis is performed when portfolio and benchmark holdings are always available. In reality, users may decide to report on time periods with beginning dates that fall on days where holdings information is not available. When this happens, holdings data from a nearby date are used to infer what the holdings would be on the desired date.

What is the methodology of inference?

Holdings’ weights are inferred based on the buy-and-hold assumption. In other words, it is assumed that no purchases or sales are made between the date where holdings information is available and the desired date. Thus, the only difference between the two dates is the holdings weights (expressed as percentages) of individual securities, which have grown or diminished by the relative returns between the security in focus and those of the rest of the holdings.

In a forward-inferring situation, if a security performed better than the rest of the portfolio between the date of the most recent holdings data and the desired date, its weight will increase. Similarly, if its return is worse than the rest of the portfolio between these two dates, its weight will decrease. If it has the same performance as the rest of the portfolio between these two dates, its weight will stay the same.

In a backward-inferring situation, the opposite is true because the date of the most recent holdings data is later than the desired date; thus, a better performing security’s weight is diminished when inferring to the desired date.

When you choose to infer forward, pay special attention to the earliest time periods in the selected date range because holdings information may not be available prior to these periods. If that is the case, holdings from a later date (backward inference) must be used.

Similarly, when you choose to infer backward, pay special attention to the most recent time periods in the select date range because holdings information may not be available later than these periods. If that is the case, holdings from an earlier date (forward inference) must be used.

An Example

Suppose you need to perform an attribution analysis for the time period April 1 to June 30. You have portfolio holdings information for February 28 and April 30, but not for March 31.

The Morningstar solution: Using the buy-and-hold assumption, use nearby dates to infer the holdings data for March 31.

To infer forward, the calculation uses holdings data from February 28 (the last date prior to March 31 on which holdings information was available).

To infer backward, the calculation uses holdings data from April 30 (the next date after March 31 on which holdings information was available).

Last Updated: August 8, 2017

The information contained herein: (1) is proprietary to Morningstar and/or its content providers; (2) may not be copied or distributed; (3) is not warranted to be accurate, complete or timely; and (4) does not constitute advice of any kind. Neither Morningstar nor its content providers are responsible for any damages or losses arising from any use of this information. Any statements that are nonfactual in nature constitute opinions only, are subject to change without notice, and may not be consistent across Morningstar. Past performance is no guarantee of future results.