Difference in biodiesel production data sparks debate

My associate editor Erin Voegele and I wrote an article on the Energy Information Administration’s release of its 2011 biodiesel production statistics, and how those numbers were well short of the U.S. EPA’s EMTS figures. In the article we offer that different data sources and methods of collection were reasons for the discrepancy of around 200 million gallons, give or take.

“A guy without his head in the sand” commented on the article, saying:

“I find it hard to argue that a 42 percent increase from EIA December numbers to EMTS December numbers is just due to statistical issues. Or [a] 26 percent increase for full year comparison! EIA numbers show 112 million [gallons] produced [and] imported in December, EMTS shows 160 million. That’s 48 million more gallons in one month alone! Overall for 2011 that's 230 million [gallons] or 345 million RIN[s] more on EMTS than EIA predicts. I'd like to believe that the EIA is smart enough not to just ignore plants producing because they forgot to send in their survey responses. I would concur that there would be some discrepancy in December due to rush producing for year end 2011 to capture the [dollar tax credit], and I'm assuming that the EIA wouldn't fully account for that whereas EMTS does[,] as it captures every gallon registered as produced, since that's an RFS requirement (not an economic decision). Even adjusting the EIA number by [one-third,] say to take that into account[,] it still posts way below EMTS which only points to one thing. Massive amounts of RIN[s] being created for production that doesn't exist.”

I appreciated the comments from “A guy without his head in the sand,” even if he was implying that our heads were.

Then, a gentleman who identifies himself as Alex Henry, wrote:

“In reply to the post above:

As stated in the article, EIA had to switch mid-year from the M311K data collected by the Census Bureau to its Short-Term Integrated Forecasting System (STIFS), which uses an entirely different methodology. Whereas M311K used surveys to collect data on the production and utilization of fats and oils, STIFS uses economic data to forecast the behavior of biodiesel producers and consumers. Basically, STIFS plugs a variety of U.S. energy indicators into a massive system of regression equations. It has not, to my knowledge, been used to estimate biodiesel output before, since the EIA had been sourcing its biodiesel data from M311K before that report was discontinued last summer.

2. There are significant issues with the M311K data as a measure of total biodiesel production. First, the data for methyl ester utilization of individual types of fats and oils do not add up to the corresponding totals for methyl ester utilization of all fats and oils. Some of this must be because the reports do not have individual numbers for the methyl ester utilization of corn oil, canola (rapeseed), sunflower oil, and only infrequent data for lard. That is, the report contains rows for, e.g., corn oil methyl ester utilization, but under "Quantity" they list either a "(D)" or a "-." However, the gap between total [methyl ester] utilization of fats and oils and the individual totals for the feedstocks they have data on seems too wide to be accounted for in this way. For [the first half of] 2011, total [methyl ester] utilization was about 2.28 [billion pounds] (~304MMg), while the sum of the individual methyl ester utilization numbers for soybean oil, inedible tallow and grease, and poultry fat was just under 1.6 [billion pounds] (~213MMg). That would mean that 33 percent of biodiesel produced in those months [was] derived from corn oil, sunflower oil, canola oil, and lard, which seems high (I'd expect 20 percent at most). The Census Bureau does not address this discrepancy, beyond stating, for the entire report, "Note: Total may not equal the sum of the detail due to independent rounding." It's unclear where the extra volume in the total ME utilization figures comes from. As for the survey itself, the M311K report had 600 monthly participants and 580 smaller, annual participants. This was for all of the fats and oils in the report and all of their commercial uses, biodiesel being just one. They do not say how many of their survey participants are biodiesel producers.

3. The EIA 2011 production figure for ethanol was off by 500MMg compared to EMTS. (I'm assuming it's uncontroversial that ethanol RINs are too cheap to invite fraud.) Granted, it was 500MMg too high, and the total production level for ethanol is much larger than that for biodiesel, but it underlines the point that different methodologies can give different results, even with a single-feedstock fuel like ethanol.

4. The shift in methodology from M311K to STIFS also seems to be reflected in the standard deviations of the production totals in each set. The standard deviation for the M311K production numbers (January-July) is 17.9 (15.2 for the first five months). The standard deviation for the STIFS numbers (August-December) is 3.74. That seems like a significant difference, though of course there's no way to know if it reflects the change in methodology.

5. "[A]djusting the EIA by 1/3" would actually give us 1.14 billion gallons (868 x 4/3), which is more than the EMTS 2011 total of 1.11 billion gallons. All in all, there are good reasons to expect significant differences in the total production numbers reported by EIA and the EPA via EMTS. That's not to say there aren't phony gallons in that total, but 230MMg is way too high. I'd estimate 40MMg, which would mean 60MM bad D4 RINs for 2011, i.e. Absolute Fuels plus the [three to five] smaller rackets that RIN traders have expressed doubts about to the EPA.”

Alex thanks for the great, and detailed, comment. Even the guy without his head in the sand appreciated it, so much so that he changed his name to “A guy who appreciates a detailed response” in his reply to Alex’s comment.

“Thanks for the response—it makes sense and is appreciated. I'd much rather have read something in detail in the article rather than what comes across as their off-hand dismissal of the discrepancies. Hence why I then felt it necessary to point out what seemed to be the obvious. If nothing else there should have been some level of acknowledgement on the source of some the discrepancies. Just quickly to point out too, my 1/3 quoted was referencing December numbers alone—adjusting EIA December by 1/3 gives 150MMg, which is still below the 160MMg EMTS.”

*Note: I always appreciate constructive criticism from readers like the fellow whose head is not in the sand. But ‘off-hand dismissal of the discrepancies’? Dismissal of the discrepancies would be to report on the EIA numbers and not acknowledge the huge difference between its figures and those reported by EPA and promoted by the National Biodiesel Board.