What The HG-Type Screen Does, In Detail

General Idea

The idea behind the HG-Type Screen is to find companies that are strongly growing with competitive advantages and good cash generating abilities and shareholder-friendly management, but are overlooked by large investors. A few key things to look for that might imply such a company are:

Frequently companies matching the above requirements are quickly identified by other investors and their share prices are pushed upwards beyond a good buy point. So it is important to identify such companies as early as possible after they start showing good financial performance.

For this reason, the HG-Type Screen looks at the most recent eight quarters of reported fundamental financial data for the trends above.

The database employed by the HG-Type Screen is AAII's Stock Investor Pro, a low-cost subscription database with weekly downloadable updates. The database is in a standard binary format intended for use by the AAII Stock Investor Pro software, but it can be converted to a simpler CSV-type format that can be used by a wider variety of screens. I have developed a screen that is easily programmable to perform arbitrary queries and operations on the company data, including linear regressions, comparisons, arbitrary computations, and even charting. Basically, anything, depending on how much effort is put in.

Let's get to the detailed description of how the screen works. The screen is built out of 15 separate "tests" each of which performs one specific test. A company passes the screen only if it passes all 15 tests.

All specific examples use data from ACME United Corp (ticker ACU -- a real company, not Wile E. Coyote's favorite mail order house) through the quarter ending 06/30/2004.

But first, a short discussion on dealing with share dilution.


Share Dilution

There are two mechanisms of share dilution. The explicit one is the increase in the number of outstanding shares reported by the company each quarter. The less obvious one is dilution due to other claims on the shares: warrants, convertibles, in-the-money but unexercised stock options, etc. The effect on existing shareholders of any increase in shares can be taken into account by expressing all key numbers for a company as "per diluted share" numbers, as long as both dilution mechanisms can be identified. The company's financial reports always include the outstanding shares explicitly so that is easy. The other dilution source can be backed into by looking at the ratio between two reported net earnings: Basic and Diluted. Basic Earnings divided by Diluted Earnings gives the approximate additional share dilution factor that should be applied.

Here is an example from ACU, over the four quarters of 2004:

Calculating Fully Diluted Shares
Q4
Q3
Q2
Q1
Reported Shares (millions)
3.368
3.303
3.273
3.282
Basic Earnings per share (as reported)
$0.32
$0.12
$0.07
$0.09
Diluted Earnings per share (as reported)
$0.29
$0.11
$0.06
$0.08
Dilution Factor (= Basic Earnings / Diluted Earnings)
1.103
1.091
1.167
1.125

Fully Diluted Shares (= Reported Shares * Dilution Factor)

3.715
3.604
3.820
3.692

Now we can put this all together and come up with the true revenue seen by existing shareholders, which is what really matters: the revenue per fully diluted share:

Revenue per Fully Diluted Share
Q4
Q3
Q2
Q1

Revenue (millions)

$12.3
$8.6
$8.1

$9.5

Fully Diluted Shares (millions)
3.715
3.604
3.820
3.692
Revenue per Fully Diluted Share
$3.31
$2.39
$2.12
$2.57

The limitations of this approach are two-fold: if earnings are only a few cents a share, the accuracy of the share multiplier ratio is not very good; and the diluted earnings only report a subset of the potential dilution effect of stock options and other convertible instruments, as only the dilutionary effect of in-the-money instruments are counted and reported. Further refinement would require investigating the company's SEC filings.

All per-share computations in the HG-Type Screen are per fully diluted share as described here.


Income Statement Tests

A few common approaches underlie all the Income Statement Computations. These approaches focus on two major things: finding trends accurately in the face of quarterly variations such as seasonality, and identifying recent performance above the trend (acceleration).

Finding Trends

While quarterly results may have strong seasonal variations, year-over-year ratios generally are considerably smoothed out. Even if Q4's results are always high because of Christmas and Q1 is always low, the improvement a company makes in its general revenue trend can be seen by looking at the year-over-year ratios. Here is an example for ACU, ACME United Corp:

Year-Over-Year Increases
Q4
Q3
Q2
Q1

2004 revenues (millions)

$12.3
$8.6
$8.1

$9.5

2003 revenues (millions)
$10.1
$7.2
$6.9
$7.9
Year-over-Year, 2004/2003
1.218
1.194
1.174
1.203
Percent increase, Year-over-Year
21.8%
19.4%
17.4%
20.3%

While there is clearly a strong seasonal component to ACU's quarterly revenues, the Year-over-Year percent increase is quite stable, ranging from 17.4% to 21.8%. This gives a good insight into ACU's performance that shows a good trend: ACU's revenue growth rate is consistently very close to 19-20%.

Examples of Growth Screening

To see visually how this works, here are two charts. Each chart has the data points and a line fitted to the points using a standard linear regression routine. First is the per-diluted-share quarterly revenue of ACI over the last 8 quarters:

Quarter 8 is the most recent quarter. Following is a chart of the year-over-year growth of revenue per diluted share:

Quarter 4 is the most recent quarter.

The key conclusions that the HG-Type screen draws from this data are:

ACU passes both the increasing revenues test (line slope is positive) and the accelerating revenues test (last point is above the line).

One bit of terminology: I will frequently use the term "base" in the tests. The base value is the value where the fitted line intersects the right axis. It serves as a more stable proxy for the most recent actual data point.

In the HG-Type Screen, this type of analysis is done for revenues, year-over-year revenue growth, and year-over-year operating income growth (all per diluted share). If you are thinking that operating income, which can be negative in some quarters and positive in others, may cause problems with this approach, hold that question until the next section.

To summarize, the income statement growth tests use linear regression line fits and some simple comparisons to look for some combination of:

Operating Income

True "structural" operating income is difficult to estimate accurately from most financial statement filings. In addition, for many companies expected to be good investments, recent operating income may be negative but on its way to strong positive performance. The HG-Type Screen only does a simple check on operating income--that it is accelerating above trend.

First, an attempt at a "structural" operating income per share is calculated as follows:

OpInc = (Gross Operating Income - Non-Operating Interest Expenses + Unusual Expenses - Unusual Income) / Diluted Shares

Second, since operating income may be subject to the same seasonal variations as revenue, a year-over-year approach needs to be taken. However, since operating income can be negative, using a ratio as is done for revenue will not work since the ratio may be positive if both quarters' operating incomes are negative, yet be negative when one is positive and the other is negative. This can lead to wide variations in the year-over-year ratios that are meaningless.

Instead, operating income's year-over-year values are not calculated as ratios, but as differences, between quarterly operating incomes as follows:

YoYOpInc1 = OpInc1 - OpInc5
YoYOpInc2 = OpInc2 - OpInc6
YoYOpInc3 = OpInc3 - OpInc7
YoYOpInc4 = OpInc4 - OpInc8

Therefore an improvement in operating income from one year to the next always results in a positive value. The regression analysis is done on the four YoYOpInc1 .. YoYOpInc4 values as described for the year-over-year revenue values above.

Income Statement Tests

With the preceding as background, here are the specific tests run on income statement data. All except the first use some form of the trend and acceleration analysis described above. Note that the gross margin tests use only gross margin data from the last 4 quarters, since the idea of the screen is to try to find recent upturns.

Name
Test
Parameters Used
Notes
Total Sales Total revenue (not per share) > $40 million Total revenue, most recent 4 quarters Minimum company size
YoY Sales growth Year-over-Year sales growth > 0 Year-over-year revenue/share growth, last 4 quarters Make sure year-over-year sales are growing
YoY Sales Acceleration Actual most recent Y-o-Y sales growth > Y-o-Y fitted base Most recent Y-o-Y revenue/share growth; Y-o-Y revenue/share fitted base Y-o-Y sales growth is accelerating over trend
Gross margin Fitted gross margin base > 20% Gross margin, most recent 4 quarters Minimum gross margin
Gross margin growth Slope of fitted gross margin > 1%/year Gross margin, most recent 4 quarters Make sure gross margin is growing
Operating income Last quarter's operating income > 10% of last quarter's gross income Operating income/share, gross margin base, revenue/share Operating income needs to be at least somewhat positive
YoY Operating income growth Year-over-Year operating income growth > 0 Year-over-year operating income/share growth ratios, last 4 quarters Make sure Y-o-Y operating income is growing
YoY Operating income acceleration Actual most recent Y-o-Y operating income growth > Y-o-Y fitted operating income base Most recent Y-o-Y operating income/share growth; Y-o-Y operating income/share fitted base Make sure operating income is accelerating over trend

Balance Sheet Tests

The goal of the balance sheet tests is to eliminate companies that are obviously unattractive from the balance sheet point of view, due to either high debt levels or poor current asset management (receivables, inventory, payables).

The main question on the debt test is what to compare the debt against. We are looking for companies that may not necessarily have great debt/asset ratios right now but have the ability to generate enough cash to pay off their debts in a reasonable time frame. In other words, we are looking for some rough measure of debt coverage. For this purpose, a "potential cash flow" is defined as the company's current gross income. There are of course many other variables that enter into an accurate debt coverage calculation, but many of these are too difficult to do given the limited data available in a screen, and generally require deeper analysis and judgment. The goal again is to eliminate companies that are obviously in debt trouble and leave the others to pass the screen so they can be analyzed in more detail.

Assessing current asset management is done by running two tests and passing the company if it passes either one. The two tests are:

Neither one of these tests is particularly stringent, but again serves to eliminate companies doing poorly on current asset management.

Here are the Balance Sheet tests in summary form:

Name
Test
Parameters Used
Notes
Debt Total long term obligations < current gross income Long term debt, other long term liabilities, TTM revenues, gross margin base (from gross margin line fit) Debt level is surmountable
Current Asset Management AR Turn Rate > 4 and Inventory Turn Rate > 2.5 Accounts Receivable and TTM revenues; Inventory and TTM Cost of Goods Only one of these two tests needs to pass
Foolish Flow Ratio < 2 Current assets, cash and short-term investments, current liabilities and short term debt

Maximum and Minimum Price

Even if the company is performing very well, we don't want to look at it if it is much too expensive. Again, the intent with testing for price is to eliminate companies selling for an obviously much too high price.

Calculating a reasonable maximum price runs into similar problems as calculating a debt coverage ratio. We don't know enough about cash flows in the types of companies we are looking at to make any accurate earnings or cash flow estimates to base a valuation on. Instead, the HG-Type Screen uses gross income again as a starting point. In this case, however, we want to be able to project the company's performance out a few years.

The strategy is to project gross income out 1 year by applying the average revenue growth rate (per diluted share, of course) for the last 4 quarters to the average revenues per diluted share for the last 4 quarters. This should represent the revenues per diluted share of the company 1 year from now assuming it continues growing at its current rate with similar share dilution. Then this projected revenue is multiplied by the fitted gross margin base to arrive at an estimate of the gross income per diluted share next year.

In short, the maximum price of the stock is calculated to be:

MaxPrice = Current Liquid Cash + (4 * Next Year's Projected Gross Income)

One final detail on this: to eliminate the occasional crazy numbers like 5,000% growth, for this maximum price calculation the screen limits the revenue growth rate to 100%, and the gross margin to 60%.

The minimum price test is simply that the share price is > $1.


Ownership Tests

The remaining tests assess stock ownership as follows:

Name
Test
Parameters Used
Notes
Insider Ownership Insider Ownership > 5% Insider ownership of float Management aligned with shareholders
Institutional Ownership Institutional Ownership < 50% Institutional ownership of float Relatively undiscovered by large institutions

Statistics

The screen collects statistics on each run to report how many stocks are analyzed and how many pass each individual test. On a recent run (database ending 07/15/05), the full Stock Investor Pro database consisted of 8768 stocks. Of these, the HG-Type Screen only considered 6742; all the missing ones were eliminated due to not having 8 quarters of data, insufficient data, zero revenues, zero share count, or zero gross income. Many of these problems may very well be SI Pro limitations, but it is actually surprising just how many flaky companies are publically traded.

Of the remaining 6742 companies, the table below lists the number and percent that passed each individual test. It is encouraging from a screen construction point of view that no single test dominated the screen results. Even though no single test passes fewer than 32% of the companies, the final screen result of 10 passing stocks is only 0.15% of the universe of 6742. It appears that all factors being tested for contribute in some relatively balanced part to the final results. In fact, multiplying all the passing percentages in the table below together results in 0.06%, or what 4 companies would represent. This means that if all 15 tests were perfectly independent of each other, 4 companies would pass the screen; the fact that only 10 pass means that while there is some correlation between the tests, it is not a lot.

At some future point it might be interesting to find correlations between test outcomes to determine which tests may be redundant and therefore whittle this screen down to a smaller, more essential set of tests. But for now there does not appear to be an obvious imbalance in the screen.

Test
Pass Number
Percentage

Sales Base > 0

6713 / 6742

99.6

Gross Margin > 20%

5625 / 6742

83.4

YOY sales growth > 0

3793 / 6742

56.3

YOY sales acceleration

3024 / 6742

44.9

YOY operating income growth > 0
3034 / 6742
45.0

YOY operating income acceleration

3199 / 6742

47.4

Latest operating income > 10% gross income
3976 / 6742
59.0

Gross Margin growth > 1%

2181 / 6742

32.3

Long Term Debt < Gross Income

4271 / 6742

63.3

Sales at least $40 M

4662 / 6742

69.1

Current Asset Management

5130 / 6742

76.1

Insider Ownership > 5%

4871 / 6742

72.2

Institutional Ownership < 50%

4347 / 6742

64.5

Price > $1

5746 / 6742

85.2

Price < Maximum

3697 / 6742

54.8