All about the EIA

As a US agency, EIA data is heavily, though not exclusively focused on US information. In recent years though, given the importance of US oil and gas production, that has been really useful and the EIA has worked hard to adjust the type of data it collects to reflect current market relevance.

Other notable features of EIA data? The EIA has been around since 1978, which means its is able to provide some very long and consistent data series (at least by energy data standards). This is often a problem in relatively young markets – data series are too short, have breaks or gaps in the data, or change methodologies periodically which make them really hard to analyze in a useful way. The EIA is also highly transparent about its methodologies. There is extensive documentation on the website, and when all else fails email them or call them up – there is a list of experts by topic on the website along with contact info and they are typically very responsive. Also, most of the EIA’s data is collected using surveys of companies. Because it is mandated to conduct these surveys, participation is very good (though the EIA also has to be careful to protect individual company data or info, which makes some data hard to collect in regions where there is a high concentration of ownership, for example).

The weekly EIA statistical reports: important but over-scrutinized

The EIA is perhaps best known for its Weekly Petroleum Status Report released at 10:30 am on Wednesdays, and its Weekly Natural Gas Storage Report is released at 10:30 am on Thursdays (absent holidays, etc., a schedule of which can be found here for oil and here for gas). The market tends to focus heavily on these releases.

Why? Historically, any energy fundamentals data with a frequency greater than monthly was really hard to come by, so any information that was provided weekly was incredibly useful. This has changed in recent years as many market participants increasingly rely on more frequent, or even real-time private data from satellites, pipeline flow measurements, and other new technologies. Much of this new data is exclusive due to cost, though does have a way of leaking out into the broader market consciousness. Some is relatively more accessible via services like Reuters and Bloomberg. But EIA data is still in rare company as some of the only official, government energy data available more often than monthly.

Why might the market’s obsession with the weekly stats be overkill? Mainly because a single snapshot of a single week may or may not be indicative of a broader market trend, either because of the margin of error in the data, or because of unusual events or circumstances in a given week that push the data temporarily off trend. Also, as this data has had more and more eyes (and algorithms) on it in recent years, the knee-jerk reaction in the market when the data is first released may not be an accurate or nuanced interpretation of the data – that may come minutes or hours later.

Interpreting the Weekly Petroleum Status Report

The weekly oil statistics – specifically from the Weekly Petroleum Status Report (WPSR) – are best known for reporting the week-over-week changes in crude and refined product inventories. These are the numbers that pop up in headlines and to which the market reacts at 10:30 am on Wednesdays. However, there is actually a broader slate of data released at 10:30, and another tranche of data released at 1:30 pm the same day.

The EIA effectively reports a full supply/demand balance for total US crude and for regional, or PADD-level crude. Some of the values reported are based on a weekly survey of the industry, and some are modeled or imputed. This is an important distinction that is discussed more fully below.

A crude balance compares total crude supply to total crude disposition, to calculate an implied change in inventories. EIA estimates domestic crude production, and surveys crude imports – together these make up total crude supply. The EIA also surveys refinery inputs – which effectively represent crude demand — and estimates crude exports. Together these two values represent crude disposition. An SPR stock level is also reported (though does not change regularly). One can calculate an implied commercial stock change from these numbers, which may or may not match exactly with the surveyed, or reported stock change for reasons that are discussed below.  

For refined products, EIA surveys refinery production by major product, stock change by major product, and total imports of all refined products (not broken into individual product categories). Total product exports are estimated. This is all used to calculate for major product categories a number called “product supplied”.

What exactly is “product supplied”? Product supplied is often interpreted as, or even referred to in short-hand as oil “demand”. But while product supplied is an indicator of the trend in oil demand, it is important to understand the difference. Product supplied is, in effect, just a balancing item – refinery production, plus imports, minus exports, adjusted for stock change. It is an imputed value, not a surveyed number. Product supplied is calculated based on underlying series that are themselves based either on surveys that are subject to revision, or in some cases just modeled based on historical data. So product supplied is a model based on other models – it is only as good as the underlying data on which it is based, and essentially compounds uncertainty from those other series.

The other important thing to understand is that the numbers that EIA reports weekly (and monthly) are part of the primary oil balance – really the first steps in the supply chain, and reflecting the barrels that come out of the ground, off the boat, or out of the refinery. There are secondary and tertiary balances further down the supply chain, as refined products make their way to an enduser. There are pipelines, wholesaler tanks, and even individual consumer-level storage (such as a heating oil tank in the basement of a private home). None of that gets measured in the weekly EIA data. There is a fairly significant time lag and a data gap between a refined product coming out of a refinery and, say, going into the tank of a car. For that reason, “product supplied” usually gives a decent idea of the direction of the trend in oil demand when smoothed over a longer period of time, but the information that can be divined from a week or two of data in isolation is very, very limited. Resist the temptation to read too much into these numbers!

Analyzing the Weekly Gas Report

The Weekly Natural Gas Storage Report is somewhat more straightforward in that it only reports total US and regional gas storage level, as opposed to other components of the gas supply/demand balance (which are reported monthly). The tricky part is that in 2015 the EIA changed its reporting regions. Prior to 2015 there were 3 gas regions. As you might imagine, as the fundamentals of North American gas supply and demand changed radically over the past decade, those 3 regions really lost their analytical relevance. The EIA now reports 5 gas regions: East, Midwest, Mountain, Pacific, and South Central. (See maps below). One could argue that the breakdown of these regions still leaves a lot to be desired, but it at least provides more granularity than the data used to. The tricky part is series continuity; this change took place in 2015, and the EIA made estimates for the new regions prior to 2015, back to 2010. The old series for the old regions is still available, and obviously goes back much further. Analytically it a challenge, but that is the nature of a methodology change and this one really was necessary.

Another important thing to know about the gas storage data reported on Thursdays is that it reports working gas storage. In a nutshell, working gas is the layer of gas storage available to the market at any given time. This is as opposed to base gas, which is effectively a permanent layer of gas in storage that is necessary to maintain pressure and deliverability and under normal circumstances wouldn’t change. If you would like to learn more about how gas storage works, the EIA has an excellent primer on its website called the Basics of Underground Natural Gas Storage.

One final topic on natural gas storage: For the South Central region (the US Gulf, including Texas), the EIA reports working gas in storage in salt caverns and nonsalt facilities. Salt caverns are identified separately because those facilities have the ability to inject and withdraw gas much more quickly than conventional, nonsalt facilities. This makes them more flexible in response to market conditions, so how much gas is in storage in those facilities versus others matter for how dynamic the market – and therefore pricing – can be in response to changing fundamentals.

As previously mentioned, the EIA only reports gas storage data on a weekly basis, and does not report other supply/demand fundamentals weekly. However, the EIA does have a page called Natural Gas Weekly Update that provides some third-party supply/demand data and brief commentary around short-term fundamentals, so if you are a casual watcher of natural gas markets this is a useful resource.

Surveys of analyst expectations about the weekly stats — noise or useful tool?

Some financial or trade media outlets publish surveys of analyst expectations ahead of weekly oil and gas statistical releases. How useful are they as a prediction, or measure of consensus? In a nutshell, analyst estimates tend to be much more accurate for the gas storage number than for oil inventories. Predicting gas storage changes is a relatively more straightforward exercise. Even a simple model using free public data should get close, and with paid real time pipeline flow data and only slightly more expertise, these estimates can get incredibly precise. Not every estimate should be considered equally, but if you analyze past performance, you can get down to a shortlist of gas analysts to follow who generally do a very good job. In the rare weeks when the actual storage number really is a surprise relative to those expectations, that means there is something important to analyze.

Oil inventory estimates have far more limited value, especially in aggregate. In a survey of analysts, a range of plus or minus two to three million barrels is par for the course, so averages are typically useless. Even in cases where analysts do go through the full exercise of creating and analyzing an oil balance to inform their projections, there is so much more timing and quality uncertainty in the oil data that consistently pinning the single week inventory changes is no small task. In fact, there are weeks when correctly forecasting the surveyed components of the crude balance (refinery inputs and crude imports) and successfully matching the estimated components of the balance (domestic production and crude exports) to the EIA’s numbers yield an implied stock change number no where near the actual surveyed stock change value. In some weeks the balance simply doesn’t add, because of all the types of uncertainty that can creep into it. More on that below.

Bottom line: There is a useful way to use gas storage estimates, and storage numbers that fall well out of consensus expectations usually signal something important. But treat oil inventory change projections with caution and focus on nuanced analysis of the full slate of weekly – then monthly – oil fundamentals statistics released by the EIA to get relevant clues about the state of the oil market.

Why does the EIA also report a 4-week rolling average for some series?

The idea here is simply to smooth out some of the aberrations in isolated weekly data points, to give a better idea of the actual trend over time. There are lots of reasons why even accurate weekly data points can give the wrong idea about the broader trend in physical supply/demand fundamentals. Remember that weekly data provides a very specific snapshot in time, based on an industry survey with a very specific cut-off time (the cut off for the oil survey, for example, if 7 am on Fridays). That means that whether a Very Large Crude Carrier (oil tanker) delivers just before or just after the cut-off, for example can actually have a meaningful impact on weekly crude import statistics, let alone if stormy weather delays a few tankers by a day or two. For natural gas, one day of warmer or colder than forecasted weather can throw a weekly gas storage number off trend, even if temperatures revert to normal very quickly. A surprisingly small inventory build in one week may be followed by a large one the following week, and that will come out in the wash of a 4-week average. But three surprisingly large or small stock builds in a row, for example, will shift that average, and signal that there is something more meaningful going on.

Why are there (sometimes big) revisions between weekly and monthly energy data?

Because of the limitations of weekly data, it is really important to also pay attention to monthly data releases and how they change the fundamental picture. Sometimes they get ignored and this is at an analyst’s peril. The EIA’s weekly and monthly surveys target similar information, but fewer companies are surveyed on a weekly basis. The EIA targets 90% of volume per data category with its weekly survey, and surveys companies from biggest to smallest until reaching 90% of volume. Inclusion in the weekly survey is based on companies’ recent historical activity, and that is reassessed monthly, so the list of companies that are asked to respond to the weekly survey can change. Depending on the data category, reaching a 90% volume target may require surveying most of the companies relevant to that category. But in categories with some big players, it may be that only half the companies surveyed monthly are also surveyed weekly.

The EIA also does have to impute, or estimate data where companies fail to report on time, or provide an incomplete submission, or have clear errors in their submission that could not be verified or resolved by the data reporting deadline. That said, EIA does pursue late or missing submissions (the response rate typically ends up at 98-100%) and does follow up on what appear to be errors in how a survey was filled out. Published revisions to EIA weekly data are very rare; revisions will instead show up in monthly reporting.

More importantly, not all of the data presented in the WPSR is actually surveyed weekly – as previously mentioned, some of it is estimated in the weekly report and only surveyed monthly. Models are based on historical data, so when a data series is changing rapidly or unexpectedly, models can underperform pretty significantly. Some of the data points that are modeled rather than surveyed are not terribly significant, like refinery processing gain, or inventories of “other oils”.  But there are three that are a very big deal: US crude production and US exports of crude and refined products. Historically these were small numbers, or didn’t change much from week to week – hence the absence from the company survey. Now, of course, these are three of the most important data series for the oil market at large. The EIA is well aware of how important timely data in these categories is and does periodically add or remove series from the data it reports. In the meanwhile, the really important thing to remember is not to read much – or anything! – into the crude oil production and crude/products exports data in the weekly report. Wait for the monthly oil reporting of surveyed values for these categories.

Be aware that EIA data may be revised again in annual data publications, and could in theory be revised subsequently though that is not very common.

How to compare energy fundamentals to “normal”

It is hard to interpret what energy fundamentals mean for the market without putting them into some sort of statistical context. Often, energy fundamentals are compared to the same period of the previous year. A year-over-year comparison gives a quick idea of change over time in addition to level, but any single point in time may have been influenced by anomalous circumstances that skew the comparison.

However, in data series that are trending up or down, a five-year chart still gives an incomplete picture. Consider analyzing data relative to both the overall directional trend over a relevant historical period, and to recent historical seasonality. This is particularly important for energy demand analysis, since “normal” demand often does have a structural underlying up or down trend. The section below gives an example of this style of analysis.

It is also useful to consider the broader context of a single fundamental series. For example, if inventories are high relative to recent history, does it matter? It may be that inventories are nearing the limits of overall commercial storage capacity, in which case “rejected” inventory could flood the market and pressure prices. On the other hand, higher inventories may reflect additions to storage capacity (which may occur for a variety of reasons), which would not necessarily be bearish for prices. Looking at inventories as a percentage of storage capacity — using data available from the EIA — helps fill in this picture.

Similarly, if inventories look low relative to recent history, that may or may not be important depending on whether there have been significant changes in demand for that energy product. For example, when heating oil was phased out of use in the US for regulatory reasons, inventories fell but so, of course, did heating oil demand. For that reason, in many fundamental analyses (including on the EIA website) you will also see inventories shown in terms of “days of [forward] demand cover” — i.e. the inventory level divided by current or future expected daily demand.