## Composite Factors

Companies that are caught manipulating their earnings tend to see their stocks plummet in value. Is there a way to detect earnings manipulation only by looking at the financial statements?

#### The M-score

Professor Messoud D. Beneish studied the characteristics of earnings manipulators and used this to create a model that is pretty good at detecting this type of companies. In his most recent paper, he demonstrates that the model correctly identified a large majority (71%) of the most famous accounting fraud cases that surfaced after the model's estimation period in advance of public disclosure. The model attained widespread recognition after a group of MBA students posted the earliest warning about Enron's accounting manipulation using the Beneish model a full year before the first analyst reports.

While very few companies get indicted for accounting fraud, the M-score helps predict a firm's prospects.

A typical earnings manipulator as defined by Beneish is a firm that is:

These companies are hazardous to invest in as they're very likely to be overpriced (because of their high recent growth trajectory) and they exhibit several problematic characteristics (either lower earnings quality or more challenging economic conditions). These companies are more likely to disappoint investors in the future.

To the extent that the pricing implications of these accounting-based indicators are not fully transparent to investors, firms that “look like” past earnings manipulators will also earn lower future returns.

Beneish, Lee & Nichols in Fraud Detection and Expected Returns

Beneish initially described his M-score as a detector for companies that manipulate earnings. (click here to read his original paper.). In his more recent work, he reveals that the M-score is also an excellent predictor of future stock returns.

He summarized his main findings as follows:
1. The firms with a higher probability of manipulation (M-score) earn lower returns in every decile portfolio sorted by size, book-to-market, momentum, accruals, and short-interest.
2. The predictive power of M-score is related to its ability to forecast the persistence of current-year accruals. High M-score firms have income-increasing accruals that are much more likely to disappear next year and income-decreasing accruals that are more likely to persist.
3. The predictive power of the M-score is most pronounced for low-accrual (ostensibly high quality-earnings) companies.
4. The variables that relate to the predisposition to commit fraud (higher sales growth, change in assets quality, and increase in leverage) , rather than the variables associated with the level of aggressive accounting, are the primary drivers of the incremental power of the model.
5. Abnormal returns are witnessed in the three-day windows centered on the next four earnings announcements.

#### How do you calculate the M-score?

The M-score is based on eight variables, of which some are designed to capture the effects of manipulation while others show preconditions that may prompt firms to engage in such activity. While Beneish takes data from the fiscal years, we use the last trailing twelve-month (TTM) numbers as the current year (year t). For year t-1, we take the TTM results for the 12 months before year t.

1. Days Sales in Receivables Index (DSRI): The ratio of days sales in receivables during the last year (t) compared to the year before (t-1). A disproportionate increase in receivables relative to sales may be suggestive of revenue inflation.
$Days Sales In Receivables Index = Net Receivables Net Sales or Revenues Net Receivables y-1 Net Sales Or Revenues y-1$
2. Gross Margin Index (GMI): A value greater than 1 indicates that margins have deteriorated. This signals poor prospects and might lead to earnings manipulation.
$Gross Margin Index = Gross Margin y-1 Gross Margin$
3. Asset Quality Index (AQI): Asset Quality is the ratio of non-current assets other than plan, property, and equipment as a proportion of total assets. An AQI greater than 1 indicates that a firm has potentially increased its involvement in cost deferral.
$Asset Quality Index = 1 - ( Current Assets + Net Property, Plant & Equipment Total Assets ) 1 - ( Current Assets y-1 + Net Property, Plant & Equipment y-1 Total Assets y-1 )$
4. Sales Growth Index (SGI): Growth does not imply manipulation, but growth firms are more likely to commit fraud because their financial position and capital needs put pressure on managers to achieve earnings targets. In addition, controls and reporting tend to lag behind operations in periods of high growth. Any perception of decelerating growth can significantly impact the value of the stock and be very costly to manage. A value greater than one increases the probability of earnings manipulation.
$Sales Growth Index = Net Sales or Revenues Net Sales Or Revenues y-1$
5. Depreciation Index (DEPI): The rate of depreciation in year t-1 / year t. The rate of depreciation is equal to depreciation / (depreciation + net property, plant & equipment). If this value is greater than 1, this means that the rate at which assets are depreciated has slowed down. Either management revised the estimates of assets useful lives upwards or adopted a new income method.
$Depreciation Index = Depreciation Depletion Amortization y-1 Depreciation Depletion Amortization y-1 + Net Property, Plant & Equipment y-1 Depreciation Depletion Amortization Depreciation Depletion Amortization + Net Property, Plant & Equipment$
6. Sales General and Administrative Expenses Index (SGAI): The ratio of SGA to sales in year t / year t-1. Analysts would interpret a disproportionate increase in sales as a negative signal about the firm's prospects. Beneish expects a positive relation between SGAI and the probability of manipulation.
$SGA Index = SGA Expenses Net Sales or Revenues SGA Expenses y-1 Net Sales Or Revenues y-1$
7. Leverage Index (LVGI): The ratio of total debt to total assets in year t relative to year t-1. A value greater than 1 indicates an increase in leverage.
$Leverage Index = ( Long Term Debt + Current Liabilities ) Total Assets ( Long Term Debt y-1 + Current Liabilities y-1 ) Total Assets y-1$
8. Total Accruals to Total Assets (TATA): Total accruals is calculated as the change in working capital accounts other than cash less depreciation. This ratio proxies the extent to which cash underlies reported earnings. Higher positive accruals (less cash) indicates a higher likelihood of earnings manipulation.
$Total Accruals to Total Assets = ( Net Income - Cash Flow from Operations - Cash Flow from Investments ) Total Assets$

The calculation is as follows:
$M-score = -4.84 + 0.92 * DSRI + 0.528 * GMI + 0.404 * AQI + 0.892 * SGI + 0.115 * DEPI - 0.172 * SGAI + 4.679 * TATA + -0.327 * LVGI$

#### Let's take an example we found in the stock screener: Microstrategy

This company has been selling analytics software for more than 20 years and has 2,000 employees. Recently the company decided to dramatically increase its debt level to place a significant bet on bitcoin. At present, the company holds $92,000 bitcoins, with a current market value of 3.7bn. Its market cap is 4,6bn. After six years of declining revenue, the company seems to be making a turnaround. It reported that its Q1 revenue was up 10.3%. EBIT increased fro$-0.1 to \$+10.9m.

##### The Beneish M-scorecard

As you can see, the total score is 0.19, which is above the -1.78 threshold. According to the formula, this makes it a suspect of earnings manipulation. Other services like gurufocus give it an even worse score of 8.35, but this is incorrect and we will explain why.

We can also see that it's suspect on four signals: asset quality, depreciation rate, leverage, and accruals. Let's dive into a bit more detail.

###### Asset quality index

As you can see in the screenshot below, the company significantly increased its non-current assets. This is easy to explain by the investment in bitcoins. As a result of this transaction, this ratio went up from 0.03 to 0.85. But here's where companies like gurufocus make a mistake. The ratio last year was so low that any increase would have a significant impact. And because the M-score is just the weighted sum of these eight factors, it can have an important overall impact.

To solve this issue, Beneish winsorizes percentile 1 and 99. This means that he replaces this score with the median score of percentile 2 and 98. As a result of this, we use 3.28 instead of 27.92. If we had used 27.92, the M-score would have been close to what gurufocus calculated.

The following suspect ratio is the rate of depreciation, which seems to have slowed down by 13%.

As the company financed its bitcoin purchases almost exclusively with debt, it increased its leverage significantly. So far, this bet has paid off nicely, but the bitcoin has plunged 40% since mid-April. Investors fear that regulators worldwide will crack down on these virtual assets, and Elon Musk voiced concerns about the adverse effects on the environment.

Finally, the accrual rate compared to assets was 78%. This means that a significant share of the company's income is not supported by cash flow.

This article should not be considered investment advice; in fact, Microstrategy has been very successful so far. The purpose is to show our Beneish M-scorecard and what kind of signals it provides.

#### M-score as secondary ratio

Our members typically use the Beneish score to filter the results of other screens. They typically use it when scanning new markets for value stocks since they're unfamiliar with the companies. Since a share of the companies discovered manipulating earnings would eventually see their stocks plummet in value, it provides extra security to filter these potential manipulators out of the screener.