"Document Review and Analysis of Economic Forecasts"
Oregon Department of Transportation Document Review and Analysis of Economic Forecasts Prepared by: Parsons Brinckerhoff Team PB Americas Simpson Consult Sorin Garber Consulting Group Starboard Alliance Tangent Services June 12, 2009 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Table of Contents Objective of Task 3.3 Technical Report.......................................................................................... 1 Oregon’s Socioeconomic Characteristics ........................................................................................ 1 Population ............................................................................................................. 1 Economy ............................................................................................................... 1 Industrial Sectors ................................................................................................... 2 Interviews with Economists.......................................................................................................... 5 Evaluation Criteria ...................................................................................................................... 6 Input from Other Groups or Studies.............................................................................................. 6 Freight and the Economy Working Group .................................................................. 6 ODOT Rail Industry and Economic Study................................................................... 7 Sources of Data.......................................................................................................................... 7 Identification of Existing Relevant Economic Forecasts .................................................................... 8 Assessment of IMPLAN and REMI Data ..................................................................... 8 Recommendations ...................................................................................................................... 9 Alternative Option # 1: Global Insight (GI) State/County Forecasts .............................. 9 Alternative Option # 2: Oregon Economic and Revenue Forecast (OEA Forecast)......... 12 Alternative Option # 3: TPAU Statewide Integrated Model Economic & Demographic Forecast (TPAU SWIM2 Forecast)........................................................................................ 16 Appendix A: Summary of Interviews with Economists ................................................................... 19 i Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 List of Exhibits Exhibit 1: Population Forecast (U.S. and Oregon) ..............................................................................1 Exhibit 2: Real Gross Domestic Product, All Industries (U.S. and Oregon), Annual Percent of Growth.......2 Exhibit 3: Industry Contribution to Real Gross State Product, Oregon...................................................3 Exhibit 4: Value of Products by NAICS Industry, Oregon 1997 – 2007 (millions of 2000 dollars) ..............4 Exhibit 5: Manufacturing Industry Value of Outputs by Disaggregated Industrial Categories, Oregon 1997- 2006 (millions of 2000 dollars) ..................................................................................................5 Exhibit 6: Interviewee List...............................................................................................................5 Exhibit 7: Economic and Demographic Forecasts Reviewed for In-Depth Analysis ..................................8 Exhibit 8: Global Insight County Forecast Output ............................................................................. 10 Exhibit 9: OEA Forecast Output ..................................................................................................... 13 Exhibit 10: OEA Forecast Error Tracking – Actual Year 2000 versus Forecasts for Year 2000 (from March 1998 to December 2000)........................................................................................................ 14 Exhibit 13: Forecast Evaluation Summary ....................................................................................... 18 ii Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Objective of Task 3.3 Technical Report The objective of this memorandum is to evaluate several existing economic and demographic forecasts and present recommendations for the identification and development of economic forecasts which would be suitable to support Oregon Department of Transportation’s (ODOT) current Statewide Freight Plan, and to support future freight planning activities. Oregon’s Socioeconomic Characteristics Population Based on U.S. Census forecasts, the U.S. population is expected to grow 29% between 2000 and 2030. Over the same time period, the U.S. Census projects a 41 percent increase in Oregon’s population1 – significantly higher than the national average of 29 percent. The US Census Bureau projects that Oregon will have the 10th highest percentage change in population among the fifty states and District of Columbia between 2000 and 2030 (see Exhibit 1). By 2030, Oregon’s population will be 25th when compared to the total populations of the other forty-nine states and District of Columbia – compared with 28th as of 2000. When comparing the US Census Bureau and Oregon’s Office of Economic Analysis forecasts for Oregon’s year 2030 population, the totals differs by less than 1.5 percent (or about 57,000 people). Exhibit 1: Population Forecast (U.S. and Oregon)2 United States Oregon Oregon's Rank 2000 281,421,906 3,421,399 28 2030 363,584,435 4,833,918 25 Change (2000-2030) 29.2% 41.3% 10 Economy Exhibit 2 summarizes US and Oregon economic growth (year-by-year growth rates) for the past ten years, measured in terms of Gross Domestic Product (GDP) for the US and Gross State Product (GSP) for Oregon. Where the green “Oregon” point is above the red “US” line, Oregon’s economy grow faster than the US overall for that specific year. There were four years in the past ten where Oregon grew more rapidly than the U.S. as a whole. In terms of overall growth between 1997 and 2007 Oregon’s economy has outpaced that of the US, growing 50% while the U.S. economy grew 33%. The graph below shows that while the timing of Oregon’s growth/decline cycles is similar to the US in general, it also suggests that Oregon’s economic cycles tends to be pronounced – Oregon’s growth years tend to have a substantially higher rate than the US, and similarly for the decline years Oregon’s economic decline rates tend to be measurably higher than the US as a whole. Oregon’s economy tends to fluctuate more than the US because as Oregon’s exports are concentrated in a few industries – computers and electronics, lumber/wood products, and agricultural, as compared to other states, as well as Oregon’s ties to the Asian and Indonesian economies. 1 Similarly, the Oregon Office of Economic Analysis projects a 42% increase in population statewide. Source: Oregon Office of Economic Analysis, Demographic Forecast, State and County Population Forecasts and Components of Change, 2000 to 2040, 2004. 2 Source: US Census Bureau, Population Division, Interim State Population Projections, 2005. 1 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 2: Real Gross Domestic Product, All Industries (U.S. and Oregon)3, Annual Percent of Growth 10% United States Oregon 8% 6% 4% 2% 0% 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 -2% -4% Industrial Sectors In 1997, the “real estate and rental and leasing” industry constituted 16% of GSP in Oregon. This industry was the largest contributor to Oregon’s economy. However, in 2007 the manufacturing sector’s share of total gross state product was 34%, which was a significant increase from 15% in 1997 (see Exhibit 3). While manufacturing has more than doubled as a contribution to GSP, “real estate and rental and leasing” and construction have seen a drop in their contribution to GSP from 1997 to 2007. These trends by industrial sector are graphically shown in Exhibit 4. 3 Source: US Bureau of Economic Analysis, U.S. Department of Commerce 2 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 3: Industry Contribution to Real Gross State Product, Oregon4 North American Industry Classification System 1997 2007 (NAICS) Industries Real estate and rental and leasing 16% 12% Manufacturing 15% 34% Wholesale trade 8% 7% Construction 8% 3% Health care and social assistance 8% 7% Retail trade 7% 7% Professional and technical services 6% 5% Finance and insurance 5% 5% Transportation and warehousing, excluding Postal Service 4% 3% Administrative and waste services 4% 2% Management of companies and enterprises 3% 1% Information 3% 5% Agriculture, forestry, fishing, and hunting 3% 3% Other services, except government 3% 2% Accommodation and food services 3% 2% Utilities 2% 1% Arts, entertainment, and recreation 1% 1% Educational services 1% 1% Mining 0% 0% 4 Source: U.S. Bureau of Economic Analysis, U.S. Department of Commerce 3 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 4: Value of Products by NAICS Industry, Oregon 1997 – 2007 (millions of 2000 dollars) $50,000 Agriculture, forestry, fishing, and hunting $45,000 Mining Utilities $40,000 Construction Manufacturing $35,000 Retail trade Transportation and warehousing, excluding Postal $30,000 Service Information Finance and insurance $25,000 Real estate and rental and leasing Professional and technical services $20,000 Management of companies and enterprises Administrative and waste services $15,000 Educational services Health care and social assistance $10,000 Arts, entertainment, and recreation Accommodation and food services $5,000 Other services, except government $0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 The observed manufacturing sector growth is largely comprised of the growth in the “computer and electronic product manufacturing” sub-sector. As illustrated by Exhibit 5, this sub-sector grew by 1119% from 1997 to 2007, which is by far the highest growth industry in Oregon. Wood product manufacturing is the second biggest manufacturing sub-sector. 4 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 5: Manufacturing Industry Value of Outputs by Disaggregated Industrial Categories, Oregon 1997-2006 (millions of 2000 dollars) $45,000 Wood product manufacturing Nonmetallic mineral product manufacturing $40,000 Primary metal manufacturing Fabricated metal product manufacturing $35,000 Machinery manufacturing Computer and electronic product manufacturing $30,000 Electrical equipment and appliance manufacturing Motor vehicle, body, trailer, and parts manufacturing Other transportation equipment manufacturing $25,000 Furniture and related product manufacturing Miscellaneous manufacturing $20,000 Food product manufacturing Textile and textile product mills $15,000 Apparel manufacturing Paper manufacturing Printing and related support activities $10,000 Petroleum and coal products manufacturing Chemical manufacturing $5,000 Plastics and rubber products manufacturing $0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Interviews with Economists As part of this task order, the consultant team conducted interviews with several key Oregon-based economists. The objective of these interviews was to gain a more comprehensive understanding of what different agencies (and economists) use for their specific economic forecasting efforts, and to gain their insights and opinions as to what the future may bring for Oregon’s economy. Exhibit 6 lists details of those interviewed. Exhibit 6: Interviewee List Tom Potiowsky, Oregon State Economist, Office of Economic Analysis (OEA) and Governor’s Economic Revitalization Team (GERT) Scott Drumm, Port of Portland Michael Anderson, Economist, Oregon Economic & Community Development Department (OECDD) Tim Duy, Department of Economics, University of Oregon Marion Haynes, Oregon Business Association Questions posed to the interviewees revolved around the current and future Oregon economy, its major industry sectors now and how they may change in the future, its economic geography, and the impact of future trade and global economic factors. Some general findings of the interviews include: • Many of the economists use, in part, economic data and forecasts from the Office of Economic Analysis (OEA) 5 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 • Some of the economists use economic data generated by Global Insight to feed parts of their overall economic forecasts, but there were concerns about accuracy of certain components (tending to be focused on those that disaggregate data to certain modes or to a sub-state level). • Agricultural products, specifically grain, tend to be stable. Oregon’s economic cycles are affected by demand for grain. Oregon has a high trade rate with Asia and the Pacific Rim. • Natural resources and mining may not recover to pre-Recession levels. Exports may not recover for some time. • Capacity issues at other ports have benefited Oregon’s ports in the recent past; however, they still compete with Oregon’s ports, especially for containerized products. • There were varying opinions as to what the future holds for manufacturing. There was general agreement that fuel and utility costs will play a role. • Wind farms and alternative energy sources may play a role in the long-term future. For a summary of the interviews, please see Appendix A. The purpose of this task was to recommend an existing forecast and forecast methodology for use driving the commodity flow forecast (not to develop the forecasts themselves). The information obtained from the interviews was helpful in evaluating the extent to which the forecasts under review reflected major factors affecting the Oregon economy now and into the future. Evaluation Criteria The gathered applicable national and statewide economic forecasts were evaluated based on the following attributes or criteria. It was determined that the forecast that best met the criteria, or encompassed these attributes, would be best suited to successfully driving commodity flow forecasting for the Oregon Freight Study: • Commodity flow economic input requirements • Consistency with ODOT’s needs including the Oregon Transportation Plan (OTP) requirements, including the requirement that the forecasts extend through the year 2035 • Reflection of export-oriented industries, industry clusters, new economic clusters • Necessary geographic and commodity/sector disaggregation of economic forecast data (county level data preferred) • Basic demographic information, including population (population and income are major drivers of retail sector goods flows) • Reflection of international and national trends impacting Oregon economy • Underlying basis of forecast and forecast drivers • Soundness of forecasting methodology • Variables included • Reflection recent economic conditions and changes • Cost of obtaining the forecast • Frequency of forecast update • Consistency in the use of economic forecasting among state agencies Input from Other Groups or Studies Freight and the Economy Working Group The Freight and the Economy Working Group met for its initial kickoff on March 12, 2009 and again on May 14, 2009. Two important pieces of information relevant to the economic forecasting task were revealed at the meeting: • The Portland Metropolitan Planning Organization (METRO) economic forecast will not be acceptable to the Working Group for use in the current freight study due to its limited regional 6 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 forecasting geography and several concerns about its range of accuracy if applied at the state level. • Oregon’s economy is more export-oriented than most other states, and thus more linked to the global economy than the U.S. economy. Oregon’s exports tend also tend to be less diverse than other states, focused on computers and electronics, wood products and agriculture. This tie-in to the global economy and lack of diversity in types of exports may help to explain why Oregon’s economy has more extreme ups and downs than the national economy as a whole. ODOT Rail Industry and Economic Study The consultant reviewed the study, Freight Rail and the Oregon Economy: A Background Paper, prepared for ODOT in March 2004. That study does not address the future of the Oregon economy or forecasting methodologies. However, it does provide extremely useful information about the relationship between the rail system in Oregon and Oregon’s economy on a detailed sectoral and modal basis. That analysis will prove extremely useful for subsequent work including the memorandum Documentation of the Relationships between Freight Transportation and Economic Development, which is currently being undertaken. Sources of Data Information and data to support the investigation and identification of an appropriate economic forecast or forecast options was collected in a number of ways. These included: • The project team has familiarity with existing commercial forecasting services and subscriptions, including the array of Global Insight5 economic and trade data forecasting products and services, IMPLAN6, Moody’s economy.com, Woods and Poole, and others. The project team has relied on these sources for numerous previous studies. • Discussions and coordination with the Commodity Flow/Freight forecasting group with the project team. Because a key criterion for the economic forecast is that it be able to drive updates or revisions to the Freight Analysis Framework (FAF2) forecast, it was necessary to develop both a clear understanding of the FAF2 forecast, as well as to discuss a methodology which would be used to update these forecasts (WOC #3). While the FAF2 forecasts are driven by Global Insight economic forecasts, the underlying forecasts are not transparent within the FAF data. • A survey of agencies within the state of Oregon which either produce or are consumers of economic forecasting services, as well as the major MPOs within the state. • Interviews with five economists or agency representatives. • A review of economic forecasting literature and on-line internet information. • Information obtained during the kickoff meeting with the Freight and the Economy Working Group. • Focused conversations (not the interviews as mentioned above, but separate conversations regarding this topic) with state agency representatives involved in economic forecasting, including: o Tom Potiowsky of the OEA; o Carl Batten of EcoNorthwest (consultant to TPAU for economic forecasting at the state level); and o Global Insight’s account representative for ODOT who is knowledgeable of the state’s current subscription services. 5 Global Insight is a private economic organization that provides economic and financial analysis and forecasts for countries, regions, and industries globally. 6 IMPLAN, (Impact analysis for Planning), is a static model based on input-output modeling structures, adapted for geographic areas down to the county level. 7 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 In connection with these discussions, the project team was provided with written documentation describing the economic modeling for OEA, as reported quarterly in the Oregon Economic and Revenue Forecast, as well as the Global Insight forecasts. Together, these data and conversations provided a good understanding of the content, advantages, and disadvantages of the various forecast sources, and provided a basis for a recommendation to ODOT. Identification of Existing Relevant Economic Forecasts The project team conducted a broad search and review of economic forecasts utilized by Oregon state agencies, selected Oregon MPOs, and other agencies such as the Port of Portland. Exhibit 7 lists those forecasts that were considered for evaluation. Exhibit 7: Economic and Demographic Forecasts Reviewed for In-Depth Analysis Forecast Name Forwarded for In-Depth Analysis If NO, Why? Global Insight State/County YES Forecasts OEA Forecast YES Statewide Model (SWIM), ODOT YES Transportation Planning Analysis Unit (TPAU) Greater Portland /Vancouver NO Specific to Portland-metro region Economic Forecast only and not a true representation of the state. Metro Regional Population and NO Specific to Portland-metro region Employment Forecast only and not a true representation of the state. Metro Report 2000-2030 Regional NO Specific to Portland-metro region Forecast only and not a true representation of the state. LCOG Coordinated Population NO Lane Council of Governments Only Forecasts MWVCOG: 2030 Population Forecasts NO Marion and Polk Counties Only IMPLAN NO Not appropriate to develop a forecast REMI NO Not appropriate to develop a forecast Moody’s Economy.com NO Does not offer a competitive advantage over Global Insight. Assessment of IMPLAN and REMI Data The project team determined that neither the IMPLAN nor REMI economic modeling services provide a suitable economic framework or forecast to support the freight planning work. IMPLAN provides information on the inter-industry relationships among the various sectors which comprise an economy, and the multiplier effects which result within an economy when additional demand is introduced. IMPLAN is a static model based on input-output modeling structures, adapted for geographic areas down to the county level; it does not incorporate relevant economic forecasting features. However, IMPLAN can be useful as a data base for current economic structures. REMI also provides inter-industry relationships which, together with other statistical forecasting methods, can be adapted to states or other large areas to forecast the economic impacts of transportation or other investments on the area’s economy. However, REMI does not make available its input output 8 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 information, and is thus not useful either for forecasting or for commodity flow analysis. REMI is much more costly to acquire than other economic forecasting models. The primary relevance of IMPLAN data for freight planning is its representation of inter-industry sales and purchases. IMPLAN “make and use tables” or other similar input output based data have been used in various studies (including the 2004 Commodity Flow study conducted for ODOT) to assist in understanding how goods flow among different industries and households, both as intermediate exchanges or to satisfy final demands. Based on these flows, freight forecasts can be made more robust spatially, as inter-industry exchanges can be correlated with the location of industries, to represent the flow of goods among counties. It is anticipated that either IMPLAN or other similar input-output (I-O) sources will be used to support the Commodity Flow Forecast process that is being undertaken under a separate work order. Recommendations The following is an in-depth evaluation of the 2 selected forecasts. It presents a series of alternative recommendations for economic forecasts that ODOT Freight Planning can adopt for purposes of guiding and driving freight analysis and commodity flow forecasting, both for the current Statewide Freight Study and into the future as ODOT freight planning activities continue. The recommendations include two forecast options for consideration by ODOT. A single long term recommendation was not identified because the decision of which forecast option to use involves coordination with parties beyond those involved in this technical report - including internal ODOT coordination and interagency coordination with other state agencies - as well as consideration of costs and other contractual factors related to acquiring subscriptions to commercial forecasting services. Alternative Option # 1: Global Insight (GI) State/County Forecasts Global Insight’s (GI) U.S. Regional service provides economic coverage of all 50 states and the District of Columbia, all 361 metro areas (including all 29 metro divisions), and the approximately 3100 U.S. counties. The data set’s regional forecasting evaluates each state and large metro area’s economy in detail with coverage of 300 variables for each state and 75 for each metro area. ODOT currently has access to this forecast database through their ODOT/OEA Global Insight license. GI State/County Forecast Characteristics As noted, the regional forecast service includes state, regional and county level forecasts. The state/county level forecasts are most relevant to the requirements of freight forecasting and analysis, since these forecasts capture activity at a small area level, which allows the freight planning and freight flows analysis to capture shifts in the economic geography within the state by industry sector. The GI forecast is a county level 30-year forecast, which has 40 variables that are updated annually. Exhibit 8 lists the GI county level output. Employment forecasts are classified by North American Industry Classification (NAICS) industries. At the state level, there are forecasts for up to 300 variables. Some of the additional variables that are available at the state level but not at the county level include gross state product, industrial output/production by sector, housing/real estate information, taxes and energy costs, and sales. 9 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 8: Global Insight County Forecast Output Employment by Sector Service Providing Private Manufacturing Transportation, Trade, & Utilities Wood Products Textile Product Mills Information Non-Metallic Mineral Products Apparel Financial Activities Primary Metals Leather & Allied Products Professional & Business Services Machinery Paper & Paper Products Printing & Related Support Educational & Health Services Computer & Electronic Products Activities Electrical Equipment and Leisure & Hospitality Petroleum & Coal Products Appliances Other Services Transportation Equipment Chemicals Federal Government Furniture Related Products Plastics & Rubber Products State & Local Government Misc. Manufacturing Other Non-Durables Military Durables & Other Durables Unemployment Rate Construction, Natural Resources, Non Durables Labor Force and Mining Employment by Place of Agriculture, forestry, Fishing Food Manufacturing Residence Beverages & Tobacco Products Number of Unemployed Textile Mills Labor Force Participation Rate Income & Wages Demographics Real Wage & Salary Disbursements, Average Annual Wage (Non-Farm) Population (by age group) Total Wage & Salary Disbursements, Heads of Households (by age Average Household Income Total group) Non-Wage Income Total Households Per Capita Personal Income Household Average Size Personal Income Births Real Nonwage Income Deaths Real Per Capita Personal Income Net Migration Real Personal Income GI State/County Forecast Methodology Unlike most other regional models, in the GI regional model each sub-region is modeled individually and then linked into a national system – that is, this model does not forecast regional growth as simple proportions of U.S. totals, but instead focuses on internal growth dynamics and differential business cycle response. Specifically, the advantages of this model include: o Each state is modeled individually, with different model structures specified according to the characteristics of the state o National policy is explicitly captured o The comparative advantage of one state over another is explicitly modeled using relative cost variables. 10 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 There are three major components to the GI regional forecast models: 1) Export module. The equations capture inter-industry, interregional, and dynamic linkages by integrating input-output, spatial theory, and econometric concepts. 2) Local module. The income generated by the export sectors circulates and multiplies through the local economy and generates the greater part of regional employment. 3) Demographic module. GI State/County Forecast Review Process As a commercial forecasting service, GI has its own internal review process. Findings of these internal reviews are not published. Underlying Data Inputs GI’s state econometric models are linked to their Quarterly Model of the U.S. Economy, incorporating national demands for goods and services as drivers of economic activity within a state. GI State/County Forecast Accuracy GI forecasts dominate the economic forecasting market in the U.S., as attested by the fact that GI (national level) forecasts are used to drive ODOT and OEA state level forecasts7. Competitor forecasting services at a national scale include Moody’s economy.com and Woods and Poole. These sources were considered in this assessment but not recommended as they are not comprehensive, do not offer a significant competitive advantage over the GI forecasts, and would be inconsistent with the major forecasting sources which drive other ODOT planning activities. GI State/County Forecast Limitations Limitations of the GI forecast include the following: o Since the forecast model is proprietary to Global Insight, it cannot be independently evaluated for its internal model specifications, forecast accuracy, and forecast review process. o GI forecasts primarily use national and regional economic data to estimate state/county forecasts. Portland’s strategic position as a major export port (and its potential growth) over the forecast period is not captured by this forecast – i.e. the Port of Portland as an economic driver is not captured by this forecast. o GI forecasts have historically been conservative, including significant under forecasting of oil prices. o The forecast model as specified currently is not well-equipped to do in-depth inter-industry analysis. This is so because Oregon’s endogenous variables are explained by national counterparts, so interactions between Oregon industries are not fully accounted for in the model. o Like many econometric models, the GI forecasts are based on historical relationships that may change in the future. So fundamental structural changes in the economy, such as shifting from a manufacturing to a service-based economy or changing household consumption patterns, cannot be captured by this model. o County Level 7 GI provides some information on its website to demonstrate forecast accuracy and customer satisfaction, but this information should be viewed cautiously (see http://www.globalinsight.com/accolades). Project team experience with recent GI forecasts (e.g., for its Panama Canal demand forecasting and modeling contract) has shown that GI forecasts are very conservative, including significant under forecasting of oil prices in early to mid-2008. 11 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Recommendation GI is one of the world’s leading forecasting companies. Though GI’s forecast accuracy cannot be comprehensively evaluated, the regional forecast characteristics fulfill many of the criteria necessary to use in the Freight Plan. However, the forecasts are not entirely transparent in their methodology. Since, all of the forecasts evaluated in this memo use GI’s national forecasts as their primary forecast driver and since the current ODOT/OEA Global Insight license provides access to the afore-described forecast, it is recommended that the GI state/county forecasts be used as the primary forecast for ODOT’s freight planning purposes. Additionally, to capture the strategic importance of the Port of Portland, the study team recommends the use of port-specific export/import forecasts (such as ones developed by the project team) to understand the links between trade going through the Port of Portland and regional economic development. Alternative Option # 2: Oregon Economic and Revenue Forecast (OEA Forecast) The Oregon Economic and Revenue Forecast at a state/sub-state level have been developed by the Department of Administrative Services at the Office of Economic Analysis (OEA). OEA Forecast Characteristics This is a 7-year forecast, which is issued four times a year – March, June, September, and December. Currently, the forecasts run through 2015. Though updates are issued quarterly, the forecast model is re-calibrated and run once every two years. The next scheduled complete forecast update is June 2010, where the forecast will be extended out to 2017. Output is geographically disaggregated at the state, region, and county level. Outputs of the OEA forecast are listed in Exhibit 9. Forecasts are classified by North American Industry Classification (NAICS) industries. This level of disaggregation (both output and geography) sufficiently qualifies the OEA forecast as a potential option for use in the Freight Plan. However, one major concern with this forecast is its relatively short forecast range. A potential solution to overcome this constraint is discussed later on in the OEA Forecast Limitations section. 12 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Exhibit 9: OEA Forecast Output Personal Income Other Indicators Employment Nominal Personal Income Per Capita Income Total Nonfarm Real Personal Income Average Wage Rate Private Nonfarm Nominal Wages and Salaries Population Construction Housing Starts Manufacturing Durable Manufacturing Wood Product Manufacturing High Tech Manufacturing Transportation Equipment Non Durable Manufacturing Private Non Manufacturing Retail Trade Wholesale Trade Information Professional & Business Services Health Services Leisure & Hospitality Government OEA Forecast Methodology The OEA forecasting model is comprised of 40 equations and additional identities. These equations are interrelated and as such, a change in one variable would generally affect others. Some variables are jointly determined through the solution of multiple economic equations – i.e., they have a simultaneous equation component. National forecasts purchased from Global Insight are an input and key driver for the Oregon state/sub- state OEA forecast. Many of the equations in the OEA economic model use Global Insight national data as explanatory variables. The Oregon-specific forecast is derived from these national forecasts. After the OEA forecast is initially estimated, the model imposes a series of “add factors” that enable OEA to do ex-post adjustments to the initial forecasts. Ex-post adjustments are necessary because the model equations (which are reliant on national trends) may not capture recent localized economic events (such as major development of ports, etc.). OEA Forecast Review Process The Department of Administrative Services’ Economic Advisory Committee and the Governor’s Council of Economic Advisors review the forecast prior to publication. The Economic Advisory Committee is composed of economists within the Oregon state government and the Council of Economic Advisors is composed of leading economists from Oregon’s business and academic communities. Underlying Data Inputs OEA has a shared 3-year subscription (shared with ODOT) to the Global Insight forecasting service. In most cases, the state/sub-state OEA forecast uses Global Insight’s national baseline (or most likely) forecast. Occasionally, based on local market conditions an alternative Global Insight national forecast (optimistic or pessimistic) is used in place of the baseline forecast. Each biennium, OEA reviews the 13 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 services of the major forecasting consulting firms and reaches a decision regarding the most appropriate provider. Global Insight’s national forecast features approximately 2000 economic, financial, and business concepts and indexes. The forecasting model is comprised of 1500 equations tracking different aspects of economic activity. It is driven by aggregate demand in the short run, while constrained by aggregate supply (labor, capital stock, productivity) in the long run. The 30-year baseline forecast is updated semi- annually. OEA Forecast Accuracy Exhibit 10 provides a summary of the short term (2 year) forecast accuracy for the selected aggregate employment variables. This color-coded exhibit (dark blue translates to higher error) tracks a comparison of year 2000 actual employment figures with a series of OEA forecasts of year 2000 employment. The analysis includes OEA forecasts from March 1998 through December 2000. As illustrated by the color coding, forecast accuracy improves as the forecast gets closer to the actual forecast year – i.e. the color turns from dark blue (20% or more error) to light blue or white (less than 2% error). Personal income, manufacturing, and non-manufacturing sectors generally have the largest forecast errors. Exhibit 10: OEA Forecast Error Tracking – Actual Year 2000 versus Forecasts for Year 2000 (from March 1998 to December 2000) Dec 1 Sept 1 June 1 March 1 Dec 1 Sept 1 May 15 March 1 Dec 1 Sept 1 June 1 March 1 2000 2000 2000 2000 1999 1999 1999 1999 1998 1998 1998 1998 OEME -0.5% -0.2% 0.3% 0.1% 0.8% 1.1% 2.1% 1.2% 1.0% 1.4% 1.5% 1.3% manuf -0.5% -0.1% 0.3% 0.0% -0.4% 0.6% 1.2% 1.0% 0.0% 2.8% 3.7% 4.9% lumber/wood -2.2% -0.4% 0.6% 0.4% 0.4% 2.9% 1.2% -0.4% 0.2% 0.2% 0.4% 2.0% nonelec -0.5% 0.5% 1.5% 2.9% 4.9% 6.4% 9.8% 11.3% 8.8% 13.7% 19.1% 25.5% elec -1.8% -3.1% -4.4% -3.4% -4.7% -6.8% -6.2% -2.1% -9.1% -2.6% 2.1% 1.6% instr -2.9% -2.9% -4.8% -1.9% 2.9% 3.8% 3.8% 8.7% 8.7% 14.4% 7.7% 6.7% transp eq 4.0% 5.9% 8.4% 5.9% 2.5% 0.0% -1.0% -5.0% -5.9% -6.4% -7.4% -9.4% food/kindred 1.2% 0.0% 2.5% -1.2% -4.9% 2.5% 5.3% 1.6% 2.5% 2.5% 1.6% 3.3% nonmanuf -0.4% -0.2% 0.3% 0.2% 1.1% 1.2% 2.3% 1.3% 1.2% 1.1% 1.2% 0.7% construct 0.9% -1.4% -0.6% -2.1% 0.2% -0.1% 2.1% -2.3% -4.9% -6.5% -5.0% -4.9% trans com ut -0.5% -0.3% -0.3% -0.4% 3.5% 2.0% 1.9% -1.9% -4.0% -2.6% -2.8% -3.9% wholesale 1.6% 1.9% 2.3% 1.8% 3.8% 5.1% 6.8% 7.1% 6.3% 7.9% 9.5% 9.0% retail -1.4% -1.0% -0.3% -0.1% -1.2% -0.9% -0.5% -0.3% 1.0% 1.6% 1.4% 1.6% fire 1.3% 0.9% 1.3% 1.4% 4.8% 5.0% 7.4% 4.7% 8.0% 6.8% 5.6% 3.9% services -0.8% -0.1% 0.6% 0.2% 0.7% 1.4% 3.3% 2.2% 1.6% 1.4% 1.9% 1.4% health -0.7% -0.9% -0.5% -0.2% 1.0% 0.8% 1.6% 2.7% 0.9% 0.3% 0.2% 0.1% nonhealth -0.8% 0.1% 0.9% 0.4% 0.5% 1.5% 3.9% 2.0% 1.9% 1.8% 2.5% 1.9% gov -0.5% -0.2% -0.2% 0.0% 1.5% 0.6% 0.5% 0.3% 0.5% -0.5% -1.8% -2.3% state -1.5% -1.0% -0.7% -0.3% -1.2% 0.0% 0.8% 3.7% 1.5% 0.2% -0.7% -0.7% local 0.1% 0.4% 0.2% 0.3% 2.5% 0.9% 0.3% -0.9% 0.9% -0.3% -2.1% -2.7% pers income pers income -0.2% -0.6% -4.9% -5.3% -6.0% -6.7% -5.7% -6.6% -7.3% -3.9% -4.0% -3.7% wage/salary -0.4% -1.6% -1.3% -1.3% -2.0% -3.6% -1.6% -3.1% -3.1% -1.5% -1.6% -1.8% div/int/rent 0.0% 0.0% -24.9% -24.9% -26.7% -26.7% -26.7% -27.6% -30.4% -23.0% -23.0% -21.2% COLOR %OFF white 0-1% 1.1-2% 2.2-5% 5.1-20% 20% - 14 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 OEA Forecast Limitations Several limitations to the OEA forecast were identified in an initial evaluation of this forecasting model: o It is a 7-year forecast, which is substantially shorter than what is required for the Freight Plan (which requires forecasts to run through 2035). This limitation could be overcome by extending the OEA 7-year forecast by “pegging” each Oregon-specific variable with the corresponding national variable from the Global Insight forecast. Since the OEA Oregon-specific forecast is a derivative of the Global Insight national forecast, and since Global Insight provides a 30-year forecast (under the same subscription), the afore-described extrapolation methodology is a reasonable option. Another option is to request OEA to extend the forecast by applying the existing methodology (or version with fewer variables). o The model as specified is currently not well-equipped to do in-depth inter-industry analysis because Oregon’s endogenous variables are explained by national counterparts, so interactions between Oregon industries are not fully accounted for in the model. o Like many econometric models, the OEA forecast is based on historical relationships that may change in the future. So fundamental structural changes in the economy, such as shifting from a manufacturing to a service-based economy or changing household consumption patterns, cannot be captured by this model. OEA uses an adjustment factor to correct for some of these structural changes. o The forecast model does not adequately reflect the factors that are changing the dynamics of international trade and the trade goods specifically coming out of Oregon; nor does it account for the regional economic effects of the continued growth in freight traffic at the Port of Portland. OEA is currently evaluating the possibility of adding the Baltic Dry Goods Index (which is a trade measure) as one of the leading economic drivers of the forecasting model. o OEA forecast model was developed in 1980. Though the model has evolved since then, it is likely that the model requires a complete audit to ensure the model structure has kept pace with the structural changes in the state economy. o The model demonstrates high forecast errors, especially for short-term forecasts of employment (which traditionally follow a fairly consistent linear trend). Perhaps the reliance on national drivers to inform the Oregon forecast is causing these errors – i.e. localized market dynamics are not effectively captured (Exhibit 10). Some of the reasons for high forecast errors may include: External events such as natural disasters, policy shifts, etc. – errors of this nature are exogenous and by nature hard to predict. Errors in the national forecasts developed by Global Insight – U.S. macro national forecasting model by Global Insight is not available for a detailed audit. Recommendation While there are limitations to the OEA forecasts as outlined in the preceding section, the OEA forecast is an option that could be explored further. OEA officials have expressed interest to work with ODOT’s Freight Mobility Group to customize/enhance its forecast to suit the needs of this group. If the OEA forecast is selected a potential forecast, then a complete audit of model specifications (including internal model specifications and specification between state and national economies) is recommended. 15 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Alternative Option # 3: TPAU Statewide Integrated Model Economic & Demographic Forecast (TPAU SWIM2 Forecast) The Oregon Statewide Model (SWIM2) developed by ODOT’s Transportation Planning and Analysis Unit (TPAU), includes a long-term economic and demographic forecasts at a county/zone level. The SWIM2 model’s economic and allocation components were developed by a PB model development team, including staff from EcoNorthwest and HBA Specto.8 TPAU Forecast Characteristics The Regional Economics and Demographics (ED) module in the TPAU SWIM2 model generates a 30-year statewide-level forecast of income by household and employment by industry. A second Production Allocations and Interactions (PI) module, allocates the model-wide employment and household forecasts into smaller zones via a spatial input-output allocation framework. Oregon and southwest Washington constitute roughly 500 zones that nest largely within counties and are used in the model’s sub-state allocation process. The ED module’s statewide output and employment forecast is disaggregated by the following industry sectors. Further sectoral specificity is available in the sub-state PI module allocation. These are classified by Standard Industry Classification (SIC) industries, which is an older classification system. A near-term goal of the SWIM2 model is to re-estimate the ED and PI modules to reflect the newer NAICS industry categories (used in Options 1 and 2). Exhibit 12: Output Employment Categories - TPAU SWIM2 Forecast’s Industry Structure – State Level Accommodations Government Administration Other Non-Durables Agriculture & Mining Health Services Personal, Other, and Amusements Communications & Utilities Higher Education Pulp & Paper Construction Home-Based Services Retail Trade Electronics & Instruments Lower Education Transport Fire & Professional Services Lumber & Wood Products Wholesale Trade Food Products Other Durables TPAU Forecast Methodology The TPAU SWIM2 economic forecast model consists of integrated input-output and macroeconomic models. The integration of the input-output model (using Oregon-specific 1998 IMPLAN multipliers) with the econometric model is done via “coupling” methodology. Coupled models pass information in both directions, often simultaneously. The core of the economic forecasting model is a set of simultaneous linear equations representing the input-output structure of the region’s economy. Additional linearized equations for various components of final demand and for the labor market are solved simultaneously with the input-output equations. Based on historical U.S. Bureau of Economic Analysis (USBEA) data the model establishes relationships between local and national variables. The estimated coefficients are then applied to Global Insight’s national forecasts to generate localized forecasts. The same Global Insight national forecasts, used in the OEA model are used in SWIM2. 8 The SWIM2 PI module allocation process also outputs commodity flows among roughly 500 state zones and imports/exports to 5 external regions consistent with the economic forecast. The commodity flow forecast includes 41 SCTG commodities. Since this report reviews economic forecasts only, this TPAU SWIM2 forecast component is not included in this discussion. 16 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 There is an override function in the model, which enable forecasts to be adjusted. The override function is enabled by two exogenous data files, which can override data in the model in absolute and marginal terms. For example, over the past 20 years, the electronics industry has seen a boom in Oregon. Since the forecast model is based on historical data, it would forecast a similar high growth pattern in future years for the electronics industry, which is highly unlikely. In order to correct for this, the forecast model uses the override function and replaces the predicted data with exogenous values. The core economic equations and override function were adjusted to ensure the long-term forecast was reasonable. Since the ED module’s output economic forecast is at the entire study region level (Oregon and adjacent counties in surrounding states), a separate PI allocation module is then applied to disaggregate industry output (in dollars), employment, and households (by income group) to a sub-state level. The PI module allocation employs a spatial input-output allocation framework. TPAU Forecast Review Process Since the TPAU SWIM2 forecasts are just now completing calibration, these forecasts have not yet been formalized in an official capacity; however calibration results indicate reasonable results relative to historical data and the OEA 7-year statewide employment forecast. Use of the TPAU SWIM2 forecasts in an official capacity, such as for the Oregon Freight Study, would warrant a more formal review process that has not yet been established. Underlying Drivers – Bureau of Economic Analysis & Global Insight National Forecast Historical data used in estimating the TPAU SWIM2 economic forecast model comes primarily from the Bureau of Economic Analysis (BEA), while all forecast data inputs are from the Global Insight national forecast (obtained via the ODOT/OEA license). TPAU Forecast Accuracy All equations in the TPAU SWIM2 forecast model are estimated from a 10-year or more historical data stream, rather than calibrated to fit any particular year. Therefore, there are three validation requirements that must be tested– 1) a good fit to historical data; 2) a good fit to historical and 7-year OEA forecast data; and 3) a reasonable long-term forecast beyond the OEA 7-year time horizon. A valid long-term forecast is the primary objective, given the intended policy use of the model. A preliminary analysis of the forecast by TPAU has found the forecast to be: o A good match with BEA historical data, with the exception of data for the lumber, electronics, and higher education sectors. These three sectors, where historic trends used to estimate the model were unrepeatable, were adjusted in the model equations and override feature to produce valid forecasts in the long-term. o The forecasts’ SIC industry categories are not a good match with ODOT’s NAIC-based industry categories TPAU Forecast Limitations o The SWIM2 economic forecast is an aggregate study region wide forecast, and is geographically disaggregated to the county level using a separate SWIM2 allocation model. This two step process (which is developed by two separate entities) increases the potential for model specification-related errors. Also, it is difficult to perform valid statistical diagnostics tests, when the models are disconnected. o The forecast uses SIC industry classifications, which differ from most current models that use NAICS codes. Bridge tables are available to transform SIC industry classifications to NAICS. A near-term goal of the SWIM2 model is to re-estimate the ED and PI modules to reflect the newer NAICS industry categories. o The forecast has not been used in an official economic capacity, so it validity has not been fully evaluated for this purpose. 17 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 o There is insufficient data to conduct a forecast error analysis, which would be part of validating the model for an official economic use. o Like many econometric models, the TPAU forecast is based on historical relationships that may change in the future so fundamental structural changes in the economy, such as shifting from a manufacture to a service based economy, cannot be captured by this model. TPAU uses an override function to correct for this. Also, the underlying input-output multipliers are from 1998 with some adjustments to better reflect Oregon’s economic mix. The capital/labor ratios in Oregon are likely to have changed over the last 11 years, so using 1998 data increase the likelihood for errors in the model. o The forecast model does not adequately reflect the factors that are changing the dynamics of international trade and the trade goods specifically coming out of Oregon; nor does it account for the regional economic effects of the continued growth in freight traffic at the Port of Portland. Recommendation The limitations described in the preceding section, particularly given the limited experience and testing of this new model are significant; and as such, the TPAU forecast is not recommended as a short-term option to use in ODOT’s Statewide Freight Plan. This is certainly an option that could be considered in the long-term. The study team recommends that the TPAU forecasts to be re-evaluated once the TPAU SWIM2 forecast model is fully calibrated and after sufficient time and review processes have occurred to better assess the accuracy of the forecasts (via an analysis of actual versus forecasted values). TPAU staff has expressed interest in working with ODOT’s Freight Mobility Group to customize and/or enhance the forecast to suit the needs of the Freight Mobility Group; a continued dialog by these two ODOT Units is recommended. Overall Summary A summary matrix of the evaluated approaches is presented below (exhibit 13). Exhibit 13: Forecast Evaluation Summary Forecast Methodology Use “as is”, or need to Industry/Commodity Study team ranking Reliable/updatable Different Economic Projection through Validity/Accuracy for use in freight Ability to Run Cost to ODOT Categories Customize Scenarios planning Forecast Tested 2035 Option 1: Can be Can use “as is”, Global accessed but recommend Need to request Insight Yes Yes Yes NAICS via discussion with 1 from Global Insight State/County ODOT/OEA GI to customize Forecasts license forecast Need Need to request Option 2: further Customization from OEA; and No Yes Yes NAICS 2 OEA Forecast discussions required limited by with OEA underlying GI data Option 3: Need Need to request further Customization from TPAU SWIM2, TPAU SWIM2 Yes No Yes SIC 3 discussions required limited by Forecast with TPAU underlying GI data 18 Document Review and Analysis of Economic Forecasts Draft: June 12, 2009 WOC 2 Task 3.3 Appendix A: Summary of Interviews with Economists Attached to this memo. 19