While working on my previous article, estimating and presenting scenarios for Tesla's (TSLA) selling, general and administrative (SG&A) expense, I came across something that struck me as somewhat strange: quarter-to-quarter reporting of subcategories within SG&A had some inconsistencies. Let me first start with how I got there.
If we look at a typical quarterly report for Tesla, the item SG&A will be divided into subcategories. As an example SG&A from the 2015 Q1 report:
We see the total amount of SG&A for the quarter: $195.4 million. We see what it was in the same quarter a year ago: $117.6 million. Then the increase is explicitly mentioned: $77.8 million, and broken down into subcategories:
- Increase of $36.5 million in category "office, information technology and facilities-related costs to support the growth of our business as well as sales and marketing activities to handle our expanding market presence."
- Increase of $33.3 million in category "employee compensation expenses related to higher sales and marketing headcount to support sales activities worldwide and higher general and administrative headcount to support the expansion of the business."
- Increase of $9.9 million in category "professional and outside services costs."
Note that the numbers of the increase in the subcategories don't necessarily add up to the overall increase. In the example above: 36.5 + 33.3 + 9.9 = 79.7, which is more than the 77.8 overall number. This in itself does not bother me, since the report only says "consisted primarily of." In other words: there are smaller subcategories, in this case slightly offsetting the increase in the larger subcategories.
Since my idea was to look more closely at SG&A changes over time, my original intention was to see if the development of the subcategories would tell me anything. Alas, Tesla doesn't make that easy. As can be seen in the example, there are no total numbers given for the subcategories, only increases compared to the same period a year before. So to get to the total, you have to go back a year, and another, and another... etc. I went back all the way to 2010. By that time, the numbers are so much smaller than the later increases that I felt safe to make an estimate without it having a big impact.
Then I noticed something that caught my attention: subcategory numbers are sometimes not consistent between quarters. Let me explain by showing the SG&A reporting for 2015 Q2:
Here we see the numbers for the second quarters, followed by the numbers for the first half of the year. Call me naive, but I would expect: ∆Q1+ ∆Q2 = ∆H1.
So let's look at subcategory "office, information technology and facilities-related costs [..]": For Q1, we had an increase of $36.5 million. For Q2, it is $15.6 million. If I add this up, I would expect an increase over H1 of 36.5 + 15.6 = $52.1 million. Now compare this to what has actually been reported for H1: $37.3 million. Somehow there is a difference of a 52.1 - 37.3 = $14.8 million.
This is clearly not a rounding error, but a fairly substantial difference. The wording of the subcategory is identical in Q1, Q2 and H1, so there is no explanation in how the subcategory is defined. It is also not a matter of misclassification, because then the difference would show up with the opposite sign in another category, which is not the case for 2015. (There is one such instance, but all others cannot be accounted for this way.)
Once I found one unexpected difference, I decided to check for all quarters in all years. It turns out there are such differences not only for ∆Q1 + ∆Q2 vs. ∆H1, but also for ∆H1 + ∆Q3 vs. ∆9M. (9M = "nine months ended," i.e. three quarters together).
Below is an overview of all the differences I found. I skipped the ones with a difference below 5%. I also left out the ones where there might be an explanation in a difference of classification. For this reason, there is no entry for 2017: In Q1, there was a separate subcategory for SolarCity with a considerable amount. In later quarters, there was no such entry, so I assumed its contents were divided over the remaining subcategories, thereby accounting for all differences. For dramatic effect and because I don't know what the right number is anyway, I always calculated the relative difference as compared to the lowest absolute number:
So above we see that in 2015, there was a mismatch for the item "professional and outside services costs" of about $12.7 million. Taking the total number for the first three quarters (9M) gives us $30.8 million. Calculating the same item by adding the third quarter to the first half of the year (H1 + Q3) leads to $43.5 million. This constituted a relative difference of 41.2%.
I then figured that these differences might occur in other items that have subcategories. Indeed, it turned out that within R&D the same inconsistencies between quarters occur:
Some may notice that I only mention numbers for Q1, Q2 and Q3, but never for Q4. This is because Tesla does not give Q4 numbers for subcategories. After year-end Tesla only gives us the annual number. From the annual number, we can derive the Q4 number by subtracting Q1 + Q2 +Q3 ... but of course this only makes sense if we can rely on those quarterly numbers.
SG&A Subcategories:
Now that I have done the exercise of going through all the numbers, I might just as well give my estimate. Below are the results of going back year after year for the SG&A subcategories:
It looks like labor costs have slightly gone up relative to the total, whereas the other categories shown here - "Office, information technology and facilities-related expenses," "Professional and outside services" and "Stock based compensation" - have diminished slightly. Note that this should be taken with a grain of salt: due to the discrepancies I had to make some guesses and adjustments.
Conclusions regarding discrepancies:
The first conclusion is that SG&A and R&D quarterly subcategory numbers should be treated with caution when used for analysis.
Other than that I'm not sure what the appearance of these discrepancies exactly means and I don't want to make more out of it than it is. The amounts are not in the 100s of millions, so won't make a big impact either way. If I would have guessed what the cause of the discrepancies was, then one possible explanation might be the impact of foreign exchange differences. Another might be some sort of retroactive reclassification.
I would like to stress that I did not find such discrepancies between the quarterly total SG&A or R&D numbers. Any analysis based on those total numbers should not be affected. Also, note that the discrepancies occur in the unaudited quarterly reports and (by definition) not in the audited yearly reports.
Before writing this article, I emailed Tesla for clarification but did not receive a response.
If there are some accountants or auditors reading this... feel free to comment.
-- Special thanks to Patricia Torres Tato for feedback and sanity check.--
Disclosure: I am/we are short TSLA.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: I am short via long-dated, out-of-the-money puts.
