Every February, the State’s attention seems to turn toward the North Slope. People become interested in whether or not oil production has increased, and what the experts expect to happen next. Inevitably, this leads to questions about “fair share” and “competitiveness.”

People voice different interpretations of the data they are presented. People like me have to try to explain how a reduced decline rate is actually an increase in production. And then there’s the speculation about what caused the changes we see.

Eventually we start our annual debate about oil taxes. But of course our confirmation biases prevent us from hearing what others say. The conversations fill with hyperbole and rhetoric. The lobbyists and consultants flood into the Capital building, spinning the numbers to fit their talking points.

And the whole system is complicated enough that you can find data to support any position you have. You can focus in on a single variable to validate your beliefs, even if the larger picture washes out the findings. Or you can just torture the data until it confesses to the finding you want to be true.

In the end, the picture is so muddy that no one really knows what the data are saying. How can we know which tool to use if we don’t understand the problem?

That is what I am trying to help fix through this endeavor. I want us to have informed debate based on factual information. That’s what this blog is all about. No spin, no agenda, just the facts.

Oil Industry

So let’s start by looking at the oil industry. There is no question that the oil industry is important to Alaska’s economy. It provides good paying jobs to thousands of people. Those paychecks help support the rest of the economic system. The tax and royalty revenues the industry contributes are what pay for almost all of the government services we enjoy. And the money that those resources generated in the past is what provides us with our dividend check each October.

The health of the oil industry is vital to our economy. And understanding what is going on in the industry must be a starting place for any broader conversation. That is why it surprises me that we don’t have a better handle on the data. And without a good understanding of the data, how can we speak intelligently about what those data mean? Why do we provide so little effort into understanding what is going on and what will happen next?

Forecasts

Right now, the only agency that provides a forecast of the oil production is the Department of Revenue (DOR). Yet, their data is confidential and their forecasts are weak. Historically, actual production has come in below their forecast 96% of the time.

While I helped to improve the situation back in 2012 (when I worked for DOR back then, I was part of the reason we introduced “risking” to the forecast), there is still work to be done. Now that DNR is helping generate the production forecast, it is a little better. But there is still room for improvement.

DOR also provides a price forecast, which combines with the production forecast to project future state revenues. As I said before, the oil price forecast is not very good either (ironically they tend to miss in the opposite direction).

The State relies too much on these forecasts for them to be unreliable. Especially now, as budget cuts, taxes, and reductions to the PFD are perennial issues that threaten government shutdowns.

We need an unbiased, non-partisan, third-party to intervene and interpret the data. Someone that can double-check the numbers, understand the issues, and inform the conversation. That’s the role I’m trying to fill.

Data Problems

But the problem isn’t just in the forecasts. It starts with the data. If we can’t agree on what history was, how can we possibly agree on what the future may be?

For example, look at the production data for the last 15 years from the various sources that provide it.

Let me help reconcile why these numbers are different.

Data Sources

In case you want to double-check my numbers, let me explain how I got these data. The DOR RSB figures come directly from its annual publications. DOR receives these data as part of tax filing requirements, and therefore the data are confidential in detail.

The DOR website also provides daily throughput numbers, which it receives from Alyeska every night. You can get them here.

The DNR data come from its website. These data are provided to fulfill the royalty requirements of the lease agreements. These data are very detailed and are not confidential. However, they only provide the detailed information back to 2012 on the website. This data is in a report format, so it’s a pain to use for anything else.

AOGCC has a fantastic data extraction tool you can access here. These data are provided to the Alaska Oil and Gas Conservation Commission at a very detailed level by the producers. These data are not confidential except in very specific circumstances.

And EIA publishes its numbers here, but they are delayed compared to other sources.

Reconciling Production Data

So what was production in 2017? These different sources provide answers ranging from 527,912 to 479,986 barrels of oil per day. That’s a 48,000 barrel per day difference between sources. How can we be certain if actual production rates differ from forecasts due to errors in forecasting or errors in reporting? And if you try to combine data sources (like using AOGCC data to forecast what DOR will report), good luck.

Fortunately, there are good reasons for the different data reporting. But, you need to understand these differences before you use the data.

AOGCC is the Source Data

It turns out, AOGCC is the most reliable data source. It is reported down to the well level, on a two month delay. And these data are publicly available. From what I can tell, these data are the source data for everyone else.

EIA Data

Let’s start with the EIA. This discrepancy was the most concerning to me, since it is consistently lower than the others. I dug around their disclaimers, made a few phone calls, and played with the AOGCC data until I figured it out.

The EIA uses the AOGCC production data, but does not include the production of Natural Gas Liquids (NGLs). They also report the numbers as calendar year averages.

Once I made these adjustments, the data line up almost exactly (any accountants out there might still wonder about the remaining difference. It bothers me too, but let’s just call it a rounding error).

	AOGCC FY	EIA	AOGCC CY without NGLs
2002	1,039,515	953,641	953,693
2003	1,018,419	946,756	946,641
2004	1,001,939	885,836	885,881
2005	928,892	844,595	844,691
2006	863,514	723,956	723,977
2007	756,736	707,142	707,146
2008	740,517	669,683	669,737
2009	714,000	637,644	637,724
2010	658,887	589,419	589,493
2011	618,026	550,937	550,530
2012	594,967	514,541	514,692
2013	546,445	499,997	500,244
2014	545,578	478,162	479,137
2015	501,941	464,518	464,933
2016	516,377	474,117	474,723

So, I don’t recommend you use the EIA data. It’s incomplete.

One down, 2 to go.

DNR Data

Let’s tackle the DNR question next. On its website, DNR makes available a production history file. That file is expressed in calendar year averages. When I compare those number to AOGCC numbers (also expressed in calendar year averages and including NGLs), they look to be very much the same. However, there is a 9,000 barrel difference in 2005 that cannot be explained. I believe this is a data entry error is one of the agency’s data (here is an example of why we need third-party corroboration).

CY	DNR Prod (CY)	AOGCC (CY including NGLs)
2002	1,032,745	1,032,745
2003	1,021,159	1,021,160
2004	961,730	961,734
2005	918,247	909,381
2006	781,373	781,385
2007	768,458	768,457
2008	726,361	726,415
2009	691,205	691,262
2010	638,269	638,258
2011	600,704	600,704
2012	564,710	564,800
2013	548,767	548,639
2014	522,789	522,778
2015	509,581	509,562
2016	518,997	518,825

Production and Transportation Differences

The other file available on DNR’s website is titled “TAPS throughput.” Looking at the difference between throughput and production should be enlightening.

CY	DNR Prod (CY)	DNR TAPS (CY)	Difference
2002	1,032,745	1,000,916	31,829
2003	1,021,159	993,276	27,883
2004	961,730	935,108	26,621
2005	918,247	891,104	27,143
2006	781,373	759,081	22,292
2007	768,458	740,170	28,288
2008	726,361	703,551	22,809
2009	691,205	672,028	19,178
2010	638,269	619,655	18,614
2011	600,704	582,395	18,309
2012	564,710	547,202	17,508
2013	548,767	533,832	14,935
2014	522,789	512,827	9,962
2015	509,581	507,802	1,779
2016	518,997	517,188	1,809

TAPS is the only way for oil to move off the North Slope. So, the difference between production and throughput should be the amount of oil that is used on the North Slope. There are only two ways this happens. First, a small amount of oil is refined at the topping plants at Kuparuk and Prudhoe Bay. Second, NGLs make for good enhanced oil recovery (EOR) fluids. When a field doesn’t have the native NGLs it needs to perform EOR operations, it gets them from somewhere else. For years, this was happening between Prudhoe Bay and Kuparuk.

Looking at the table above, the volume of produced oil that stays on the North Slope looks to be decreasing steadily as production declines. That is, until 2014. At that point, something changed. I happen to know what that change was.

According to the Petroleum News, ConocoPhillips converted the Oliktok Pipeline from NGL transportation to fuel gas in late 2014. That pipeline was previously delivering NGLs produced from Prudhoe Bay over to Kuparuk for miscible injection (a type of EOR). By the way, the article also says that the pipeline is returning to NGL transportation this month (August 2018).

Royalty and Production Report Variance

Next, we need to resolve the differences in the production report and the royalty report on the DNR website. This one is actually simple. The royalty report only includes State leases (and only goes back to 2012).

CY	Royalty Report	Production Report	Difference
2012	542,551	564,710	22,159
2013	527,383	548,767	21,384
2014	503,295	522,789	19,494
2015	489,447	509,581	20,134
2016	485,918	518,997	33,080

The difference is production from Federal and ASRC lands. The increase in 2016 is production from CD5.

DOR Numbers

Now we are just left with the DOR numbers.

Unfortunately, DOR does not provide monthly data. So, let me convert the AOGCC data to fiscal year and see how DOR compares. We already saw that AOGCC data is the source data for EIA and that is matches the DNR reports. So it should line up with DOR too.

FY	DOR	AOGCC	Difference
2002	1,008,600	1,039,515	30,915
2003	987,000	1,018,419	31,419
2004	973,800	1,001,939	28,139
2005	911,300	928,892	17,592
2006	839,700	863,514	23,814
2007	734,200	756,736	22,536
2008	715,400	740,517	25,117
2009	692,700	714,000	21,300
2010	642,600	658,887	16,287
2011	599,900	618,026	18,126
2012	579,400	594,967	15,567
2013	531,600	546,445	14,845
2014	530,400	545,578	15,178
2015	501,000	501,941	941
2016	514,700	516,377	1,677
2017	526,500	527,912	1,412

These differences look familiar. It appears that DOR does not count reinjected NGLs as production. So, if you want to use AOGCC data to forecast DOR production numbers, you need to account for reinjected NGLs.

DOR Transportation and Production Variances

Ok, one last item. How close are the daily run tickets that Alyeska provides to the final production numbers that DOR reports in its annual publication? Since the DOR numbers pull the reinjection numbers out, they should be close. Let’s check.

FY	DOR	Daily Reports	Difference
2002	1,008,600	1,002,923	5,677
2003	987,000	986,350	650
2004	973,800	970,571	3,229
2005	911,300	909,769	1,531
2006	839,700	838,101	1,599
2007	734,200	731,798	2,402
2008	715,400	699,841	15,559
2009	692,700	691,130	1,570
2010	642,600	640,785	1,815
2011	599,900	597,813	2,087
2012	579,400	579,027	373
2013	531,600	530,920	680
2014	530,400	529,649	751
2015	501,000	499,971	1,029
2016	514,700	514,364	336
2017	526,500	526,440	60

It looks like the daily throughput reports are a very reliable indicator of the DOR production numbers (other than an anomaly in 2008). This means that you can use the daily data to update your forecasts in real-time without corrupting your data.

Takeaways

Data can be tricky. While there are very good reasons for the differences in the data, we need to be careful when we use them. Here are a few recommendations from me.

First, I recommend that we all use AOGCC production data when we talk about Alaska oil production. This is going to become a more serious issue as production from GMT, Willow, Liberty, and ANWR start to flow.

We need to agree on what counts as “production” when some is taxable, some is royalty bearing, some is neither, and some is reinjected into the ground. We are going to need a way to differentiate those production types without confusing the conversation.

Second, be careful to ask what the data mean whenever it is presented to you. Comparing fiscal year and calendar year data is a common mistake that can lead to spurious conclusions. Comparing different data sources can lead to similar confusion.

And finally, let’s get a better handle on the data we are using to make important policy decisions. I’ll be here trying to help, but only for as long as I can afford to do it. If you want to keep me working on things like this, consider making a donation to the cause.