Alaska Department of Revenue Commonwealth North Fiscal Policy Study Group 2014 Anchorage, Alaska – December 18, 2014 John Tichotsky, Ph.D. ( Cantab .), Chief Economist & Audit Master Alaska Department of Revenue
Who forecasts Alaska Revenue?
Economic Research Group…with commercial analysts
Economic Research Group and Commercial Analysts are also within the structure of the Tax Division proper…. 4
A New Tax Division Director, Ken Alper, with strong skills in oil and gas. T he Economic Research Group also reports to… Deputy Commissioner, Jerry Burnett, with long- term experience and technical skills in finance and revenue, also oversees Economic Research Group and Commercial Analysts. DOR has a new Commissioner, Randall Hoffbeck, as former oil property assessor for DOR understands the role Economic Research Group does and can play… 5
Alaska Revenue
Revenue Categories for Alaska 7
Revenue Categories for Alaska 8
FY 2014 Total State Revenue, by restriction and type 9
Total State Revenue History and Forecast 10
Alaska Revenue Forecasting
Why do we forecast revenue at DOR? 12
Forecasting methods from the past…. The Delphi Method converging toward a “correct” answer Oracles were thought to be portals through which the gods spoke directly to people. Usually the oracle was in a “frenzy.” An “intuitive” or “qualitative” approach.
Thanks to Percy Jackson, my children know what I do at work. 14
Divining the future has many specialties… Seers ( μάντεις ) manteis practice haruspicy or extispicy The ancient science of interpreting signs sent by the gods through bird signs, animal entrails, and other various methods. More of an empirical approach than the Oracles…
Forecasters in my early life Is knowing the future enough? What if you really could How do you make knowledge of the model the future? future useful?
What we want is a Rational Basis for Forecasting Fundamentally Scientific Approach Economic analysis based on physical realities Empirical lessons of the past Recognizing the role of change. What relationship does the future and past have? Includes abstract analogues Question is well framed, maybe re-framed from “conventional wisdom” Right mix of the qualitative and quantitative economics for policy Economic analysis being policy relevant
What do we mean by scientific? “Clearly, scientific education ought to mean the implanting of a rational, skeptical, experimental habit of mind. It ought to mean acquiring a method — a method that can be used on any problem that one meets — and not simply piling up a lot of facts. … George Orwell – What is Science? Tribune , 26 October 1945
What Makes a Good Forecast? Forecasts must be objective and justifiable Revenue forecasts should represent our best estimate of what revenue will be, not what we want it to be. All forecasts should be based on the best available information and should fit with past results. Our methodology is heavy on statistical models, but the results should make sense logically. Otherwise you have a broken model! We have the most confidence in short-term forecasts. Uncertainty increases with time. We need to be able to respond when “fat tail” events occur. Forecasting methodology should be transparent While our data may be confidential, we always publish our forecasting methodology and explain any changes. This is important to earn trust in the forecast from decision-makers and the public. It is also important for credit rating agencies. 19
What Makes a Good Forecast, continued… “Correct forecasts are not proof that the forecast method is correct” “Trends can change” “Garbage in /garbage out” “Forecasts are always wrong,” but probability and ranges can be right Developing a consistent and USEFUL approach, that may not always yield exact answers…. Goldilocks forecast: Neither complexity nor simplicity is the goal
What do We Forecast at DOR Mostly Petroleum and Nonpetroleum Revenue We directly forecast Petroleum Revenue the largest component, accounting for 88% of state unrestricted revenue in FY 2014 “Petroleum Revenue” includes severance taxes, royalties, corporate income tax, and all other revenue from oil companies We directly forecast Nonpetroleum Revenue We use someone else’s forecast for Investment Revenue We take the Federal Revenue that is authorized for spending It is typically 20%-30% more than actually gets spent. DOR compiles all different revenue streams and compiles them in the annual Revenue Sources Book 21
Oil Revenue Forecasting Three Factors for Production Tax Revenue Forecast REVENUE = (Net value * Tax Rate) – Credits Net value = (Price*Production)-Costs 1. Price 2. Production 3. Costs 1. Capital expenditures 2. Operating expenditures 3. Transportation cost 22
Alaska North Slope Crude West Coast Price Threat of regional warfare in Middleast in response to use Russia invades Russia seizes Syria uses chemical of Syrian chemical weapon Ukraine in support Crimea from the $130 weapons in its civil usage. of rebels. Ukraine. 95% Upper Bound: $124.22 $120 $110 High: $116.52 Average Price: $104.24 $100 $90 95% Lower Bound: $ 87.48 Weak demand outlook and Saudi Arabia annouces it will not defend $80 ample supplies of crude in $100 oil prices. Coupled with weak demand storage. and robust supplies oil prices collapse. Low: $66.65 $70 $60 =95% confidence level 23
Historical ANS WC Annual Average and Official Forecast 24
ANS Oil Production Forecast 25
Total Revenue Forecast – FY 2015 & 2016 ($ millions) Actual Fall 2014 Forecast Revenue Type FY 2014 FY 2015 FY 2016 Unrestricted General Fund Oil Revenue 4,755.3 2,019.2 1,636.1 Non-Oil Revenue* 508.5 502.3 528.2 Investment Revenue 130.2 30.0 32.4 Total Unrestricted Revenue 5,394.0 2,551.5 2,196.7 Designated General Fund Non-Oil Revenue* 289.6 323.1 322.1 Investment Revenue 66.3 20.4 35.8 Subtotal 355.9 343.5 357.9 Other Restricted Revenue Oil Revenue (Restricted royalties, CBRF settlements, etc) 927.6 512.9 465.6 Non-Oil Revenue (Taxes, licenses, fines, etc) * 183.9 229.2 230.4 Investment Revenue (Permanent Fund, CBRF, etc) 7,861.4 3,322.2 3,549.8 Subtotal 8,972.9 4,064.3 4,245.8 Federal Revenue Oil Revenue 6.8 5.0 5.0 Federal Receipts 2,511.9 3,126.4 3,126.4 Subtotal 2,518.7 3,131.4 3,131.4 Total State Revenue 17,241.5 10,090.7 9,931.8 *Except Federal and Investment 26 Source: Department of Revenue - Revenue Sources Book Fall 2014
How should a decision maker use a forecast? What are Forward Looking Statements? 27
We are tapping into Decision Science / Decision Analysis Use by petroleum companies Investment decisions Including how price forecasting is typically done Potential benefits to the state Decision Analysis methods better price, production, cost and revenue models
What is needed for effective decision support? 2009 National Academies report highlights six principles for effective decision support Begin with user’s needs 1. 2. Give priority to processes over products 3. Link information producers and users 4. Build connections across disciplines and organizations 5. Seek institutional stability 6. Design for learning Goal of a decision support program should be “to provide knowledge that people need to make better decisions and to do so in ways that enable and empower decision makers to use it appropriately.” Report available on : http://www.nap.edu/catalog.php?record_id=12626
The information & decision link Technical Information Strategic/ Decision Policy Analysis Decisions Produced by: To support: technical experts Policy-makers, decision- makers, clients (e.g. economists, accountants, scientists, engineers)
Decision Analysis: What is it? Approach and set of tools for structuring and analyzing complex decision problems and dealing with uncertainty Process for making logical, reproducible, and defensible decisions in the face of: ― Technical complexity ― Uncertainty ― Multiple, competing objectives A multi-disciplinary field drawing from statistics, economics, operations research, management science, psychology… http://www.decisioneducation.org/about-DEF/better- decisions
Methods “Decision analysis will not solve problems, nor is it intended to do so. Its purpose is to produce insight Distinguishing and promote creativity to help decision-makers make better characteristic of DA decisions.” - Ralph Keeney Probabilistic methods Requires solid grounding in probability theory Requires sound processes for probability elicitation Modeling Monte Carlo simulation Decision Trees
Monte Carlo Simulation – 1940s technology Developed by Stanislaw Ulam & John Van Neumann for the Manhattan Project. Codeword named after Monte Carlo Hotel where Ulam’s uncle borrowed money to gamble. 33
1978 = probabilistic forecast 1995 = shift from probabilistic Monte Carlo method to scenario with low, medium, high. 2000 = a single deterministic set of numbers…
Recommend
More recommend
