Kirkendall: My first degree was a Bachelor of Science in mathematics from Ohio State University. After I graduated I got a summer job at North American Aviation, as a computer programmer in an Analog Hybrid Simulation Group; however, my title with this job was engineer, which I thought was rather amusing. At that time I didn’t know anything about computers, however, and they hired people with a background in mathematics because they knew that with that background you could pick up programming easily. And, as it turned out, they were right; I found the programming easy. In those days, the computers we used were DDP 2400 computers that had Fortran 2 on it and we used machine language as well. So statistics were involved in the job, even though I was there as just a programmer.

I got my Master’s degree at Ohio State and as part of that curriculum I took my first statistics courses, which I thoroughly enjoyed – probably because I thought they were more practical than the mathematics courses I had been taking.

Aczel: When you graduated, you went….

Kirkendall: After graduation my husband went into the Army and was transferred to Washington D.C. I got a job at Bellcom, which was the smallest company in the Bell system. Bellcom was organized to work on space projects for NASA, and I was hired as an applications programmer, under the condition that I would go ahead to graduate school and pursue a PhD. I was thrilled because it was what I wanted to do anyhow.

My new boss suggested that I should go to George Washington University and major in statistics. He evidently knew that it was a good, convenient program in statistics, and he thought that I would like it. At Bellcom I started moving from programming into applied statistics. My husband’s company, however, bid on a job in Vietnam, and when they got that contract, they wanted very much for some of the people who had worked on the job to go. My husband said that he would go if they hired me. As a result, we spent a year in Saigon, working as senior operations research analysts.

One of the interesting projects we worked on there was a survey used to determine the quality of life and the extent of Viet Cong influence in the country. This was done on a large management information computer system. The input was from questionnaires that were filled out by people who were living in the hamlets and villages scattered around the country.

We came back to Washington, and I spent a year completing my PhD in statistics. After that I went to work for the Mitre Corporation working on projects for the Federal Aviation Administration [FAA] concerning separation standards for aircraft and air routes. Separation standards are basically rules for how far apart routes should be spaced or how far apart aircraft should fly to keep collision risk at a minimum.

Finally, 15 years ago I came to the Energy Information Administration [EIA]. The person who hired me is the same one I work for now, Dr. Yvonne Bishop, who is a well-known statistician. This was my first job as a real statistician, and I was thrilled.

Aczel: What role does statistics play in the department’s business?

Kirkendall: Well, I can certainly talk about the role of statistics in the Energy Information Administration’s business, because EIA is the statistical agency for the Department of Energy. We collect and publish data from energy producers, suppliers, and consumers. We operate forecasting models which use either econometric models or linear programming methods. And, in addition, we are currently advising the Department of Energy on implementing customer satisfaction surveys. This is something a little bit new for us, and it has been an interesting exercise. And that’s one that I have been particularly involved with as well.

Aczel: Where does the EIA get data on such things as the usage rate for different kinds of energy?

Kirkendall: EIA conducts large-scale sample surveys to estimate energy use in homes, in commercial buildings, and in manufacturing establishments. Those are three separate surveys. We collect actual billing data for the sampled units from the utilities that provide the energy – from the electric power companies or, if natural gas is used, from the gas company that supplies it. We get good solid data on the actual energy used in either the home or the commercial building.

Those surveys collected a lot of information that could be used for explaining why certain buildings consume more or less energy than other buildings of the same type; that is, we collected information about the number of appliances they have and operate, the age of the refrigerators, whether they’ve replaced their windows, whether their hot water heater is wrapped, whether they have special lighting in commercial buildings or in manufacturing establishments – almost anything that might be a descriptor or might be related to energy savings.

Aczel: Does the department keep track of foreign sources and supplies? How and where does that data come from?

Kirkendall: We collect information on energy imports into the country by country of origin. These data are actually from the U.S. Customs Service. We also have a stocks-at-sea report for petroleum that is in tankers at sea and that is more long-range prediction of imports. The United States is a member of the International Energy Agency, and we therefore have to provide a lot of energy information to the IEA, as do all other member countries. As part of that affiliation, then, we have access to the data from the other member countries. We also collect some foreign data directly from other countries that are not members of the IEA. We do not go out and interview companies that work inside other countries. We rely on the foreign government to provide information to us.

Aczel: The Department does assessments of future energy needs for the United States. How are those done and what tools do statisticians and forecasters use in the process?

Kirkendall: EIA has two modeling systems – a short-term modeling system that predicts about a year and a half into the future and the National Energy Modeling System (NEMS), which predicts about thirty years into the future. The short-term modeling system uses some statistical time series methods and some econometric models. These tend to be smaller scale models – they could probably all be put on a PC (although I believe that some of them are still on a mainframe, just for the sake of convenience).

The National Energy Modeling System is a fairly new effort and that relies on econometric models and linear programming methods. And the modulars for NEMS are gigantic programs, some of which take hours to run, even on the mainframe computers.

Aczel: Do those models get revised? What do you see as the next major changes to the forecasting system?

Kirkendall: The short-term models are the older set of models and they are continually revised and improved. As people have worked on them they will notice that they haven’t done a good job forecasting in the past and they will revise them and make improvements.

The National Energy Modeling System is new and, in fact, it’s probably not even complete as yet. They have been under a serious time crunch to put this system together and I’m sure as they’ve done it, they know where there are weak places in the model and they will be continually working to keep that one up-to-date as well.

Aczel: The energy industry, or at least the oil industry, has been down and out the last few years. Does your Department or do these models predict a comeback?

Kirkendall: I can’t speak for the Department and I have not looked that closely at specific model forecasts. However, I would think that the models most likely predict a slow increase in the price of oil. Our supplies of oil are finite and we do not yet have replacement technology securely lined up, so it would make sense that the oil industry would experience a comeback, unless there is some grand, inexpensive environmentally friendly replacement fuel or technology discovered before then. However, we don’t see any of them on the horizon just yet.

Aczel: In the 1970s and 1980s the oil crisis was projected to stick around. What happened to it? How did good or bad forecasting or data analysis affect the results?

Kirkendall: Well, I’m not sure exactly what happened. I guess economists would probably tell you the supply of oil exceeded demand by a fair amount, but in early 1986, the price of oil plummeted. As a result U.S. production fell by 4 percent per month and it was no longer economically feasible for companies to operate marginal wells. There was little incentive to explore for new wells. In the Energy Information Administration, our official data series did not show this 4 percent drop. Because we get data on crude oil production too late to include in our monthly publications, we used forecasts that were just extrapolations. And these estimates were not revised for a full year because of our publication policy. These estimates worked real well up until early 1986 because crude oil production was relatively constant, nobody thought there would be a drop like that.

In 1986, however, the oil industry started complaining to Congress about their economic woes. Congress observed that EIA data did not show that there was a problem and our administrator was required to go testify to Congress. In response, of course, we were required to change our forecasting methods for crude oil production. We now make use of auxiliary data and we have changed our revision policies. Now our monthly crude production data accurately show changes when they occur.

I would like to point out that this is really a common problem in making forecasts. The methods that make the strongest use of past data will do a great job when there is no change; it will make a nice, smooth forecast. Unfortunately, these methods will tend not to show a significant change when there is one.

Other methods make strong use of the most current but perhaps slightly less accurate data. These estimates will have more variation when there’s no change but will show a major change when there is one. So what you really need in a system like this is to make use of both systems and then use some expert judgement to decide whether the change you see in the data is a real change or just an exceptional noisy data point.

Aczel: In the future is there likely to be an energy crisis again? Why or why not?

Kirkendall: I personally think it is likely that there will be another energy crisis. After all, the world is highly dependent on energy and there are many ways an energy crisis could occur. However, it is unlikely that anybody will actually forecast either the timing or the impact of the next energy crisis.

Aczel: Do politics and policy and data and statistics ever clash or work against each other? How do you resolve such differences?

Kirkendall: Politics, policy, data, and statistics frequently work against each other. This is one reason why the Energy Information Administration was established as an independent part of the department. In 1978, when we were established, Congress realized that data and statistics should not be influenced by politics.

In fact, though, this is related to current discussions of paradigms and paradigm shifts. They say that many researchers did not see data or evidence that doesn’t conform to their perception of how things are supposed to work. And so, this is perhaps the same example, people in politics want to see things turn out a certain way and perhaps they just dismiss data that doesn’t go along with that expectation. At least within the Department of Energy, since the statistical agency is separate from the policy making department, you can be reasonably assured of honest statistics.

Aczel: If you could talk to students about statistics and how they are likely to need or use statistical thinking, what would you say?

Kirkendall: Statistical thinking is more important than people realize. We are surrounded by numbers, measures of things, and opinions. All of these are statistics. They are reported to us by the media, by politicians, and by people who are trying to sell us stuff. Statistical thinking is critical to making sense of all these numbers.

Aczel: What are your views about the state of quantitative and statistical literacy in the United States these days? How would you advise students to view statistics and mathematics in their education?

Kirkendall: Actually I have a lot of hope for the state of quantitative literacy. The American Statistical Association has initiated an active program working with schoolteachers at various levels to help them learn about projects and teaching methods to help kids appreciate statistics. In addition, the Department of Energy also has encouraged scientific education in many of its programs. There are students and teachers who are able to spend time at the National Labs working on projects. DOE sponsors an annual science bowl for high schools.

I would advise students to take advantage of any of these programs they can find and to take math and statistics, and really any of the sciences; I think they will find them very useful in the future. I would also advise students to look for practical applications of statistics – to read books describing real problems, to read statistics in the newspapers, some of the descriptions where they throw lots of numbers at you, to see if they make sense, to see if you can find errors in the reasoning.

feedback form | permissions | international | locate your campus rep | request a review copy

digital solutions | publish with us | customer service | mhhe home

---

Copyright ©2001 The McGraw-Hill Companies.
Any use is subject to the Terms of Use and Privacy Policy.
McGraw-Hill Higher Education is one of the many fine businesses of the The McGraw-Hill Companies.