Beneish Model SGAI

[SpareTime]

In the curriculum for the Beneish model it says that a higher SGAI (the increase or decrease of ratio of sales, general and administrative expenses to sales) suggests decreasing administrative and marketing efficiency, which may induce management to manipulate earnings. This all makes sense, but in the M-score formula SGAI has a negative coefficient (-0.172), which means that if SGA/Sales were to increase since the last period, it would make the M-Score smaller, which would indicate less likelihood of manipulation…which is the opposite of what it is telling me.
What am I missing here?

 

[SpareTime]

Same goes for LEVI. If increasing leverage makes earnings manipulation more likely, why does LEVI have a negative coefficient?

 

[ArsenalFan]

I noticed the same thing. Probably a mistake in the Curriculum. Have you reached out to them?

 

[MrSmart]

It probably means whether the firm is currently manipulating earnings, and not if it’s currently in an enticing position to do so.

 

[SpareTime]

This is frustrating. Everwhere I look on the internet it says the same thing - increasing SGAI indicates a higher likelihood of earnings manipulation because it may indicate management is capturing more firm value through higher salaries…yet if SGA in time t is higher than in time t-1, that would increase SGAI which would lower the M score. The formula is correct, unless everywhere on the internet is wrong…but nobody can explain why it has a negative coefficient.

 

[S2000magician]

If you read Beneish’s paper, you’ll see that he (explicitly) expected a positive coefficient on SGAI and (implicitly) a positive coefficient on LEVI.
The fact that he got negative coefficients, if it means anything, means that these two factors are compensating for excess values in the coefficients on other factors.
Maybe there’s a collinearity problem in the model; I don’t know.
In any case, I wouldn’t make too much of those coefficients in isolation. In isolation they probably mean nothing.

 

[SpareTime]

S2000magician wrote:
If you read Beneish’s paper, you’ll see that he (explicitly) expected a positive coefficient on SGAI and (implicitly) a positive coefficient on LEVI.
The fact that he got negative coefficients, if it means anything, means that these two factors are compensating for excess values in the coefficients on other factors.
Maybe there’s a collinearity problem in the model; I don’t know.
In any case, I wouldn’t make too much of those coefficients in isolation. In isolation they probably mean nothing.

Thanks Magician. I found his paper but I could only read the abstract. I do know that there was a second model that stripped out LEVI and SGAI (and maybe something else)…so maybe that’s part of it?
I figured it would be (hopefully) unlikely that I would need to remember the entire equation with all the coefficients, so I am guessing that knowing something like “if SGA expenses increase year over year, what would happen to the M score” type questions would be a good idea. I guess if a question like that arises I would have to answer that it would (all else equal) lower the M score and indicate less likely earnings management…even though this is the opposite of what the textbooks and intuition say.