> Reports+Data > Regional Statistics

Forecast Methodology

Metropolitan Council Forecast Methodology, March, 2006

I. Purpose of Council forecasting

Provide a reasonable forecast of future growth in the region to:

• Engage stakeholders in analysis and discussion about issues related to regional growth.
  
  
• Prepare long range plans for regional systems.
  
  
• Guide local communities in preparing their comprehensive plans.
  
  
• Provide other agencies, businesses and market researchers forecast data for their planning work.
  
  
• Provide a basis for monitoring growth in the region.
  
  
• Better understand the forces influencing regional growth in order to more effectively respond to unexpected change.
  

Forecasts of population, households and employment are a major quantitative element of long-range planning. The Council’s forecasts are the basis for the development of the Council’s regional system plans. They are also quantitative parameters provided to local communities to guide local comprehensive plans, but are used with flexibility. The most current information on development trends and proposals is considered before plans are implemented.

II. Regional Methodology

The Council’s forecast process has continuously evolved in response to Council policy direction and available data. Generally the Council’s forecast process consists of two parts. The first step is to prepare forecasts for population, households and employment for the region as a whole. These overall forecasts are then allocated to subareas within the region: quadrants, planning areas, cities and townships and ultimately traffic analysis zones (TAZs).

The Council’s regional forecasts provide a picture of what we can expect to occur given what we know today. The Council has never viewed the regional forecasts as a goal or a limit. They are intended to serve as a “control” for subsequent allocation of growth within the region. Obviously forecast accuracy is a goal, but for most planning purposes a forecast that proves to be too high is better than one that turns out to be too low. Although the forecasts are provided as specific numbers, they are probably best viewed as representing a range of growth possibilities.

The regional forecasts start with an age-based population forecast, which is then converted to households and labor force (which provides a check on the employment forecasts).

1. Population Forecasts
   1. The Council employs a basic cohort-component model, which is universally used for forecasting large, independent functional regions such as the Twin Cities metropolitan area. The model uses birth, death and migration assumptions by age to project future age distributions for five-year intervals for three to five decades into the future depending on Council planning needs. This basic method has been used since the mid-1970s, although some forecast updates have been made using local input or more current trend information such as the Census rather than reapplying the forecast model. The most current official Council forecasts are for 2010, 2020 and 2030. They were adopted on January 14, 2004, as part of the Regional Development Framework. They were revised for the Systems Statements in August 2005 and revised again in March 2006 based on local response to the Systems Statements.

      • Births. The number of future births are determined in the model using assumptions about future age-specific fertility rates. These rates are used to determine the total fertility rate, which is the number of children a female will have in her lifetime. For the Blueprint 2030 forecast model application the total fertility rate was assumed to be 1.9 over the entire forecast period. The metro area total fertility rate was 1.83 in 1990. It dropped to 1.79 in 1995, but rose to 1.9 in 1999.

        From 2000 through 2004 births rose steadily over what had been projected in the model using the 1.9 assumption. There were 5,008 or about 2.6% more births than projected, with the largest deviation in 2004. The question is whether the added 5,000 births reflects an increase in the fertility rate or was due to in-migration of adults of childbearing age. However, a fertility increase resulting in 5,000 more persons in five years doesn’t have much impact on a forecast expected to reach 3,000,000 by 2010.

        If the increase was due to more childbearing adults, it could nearly triple the impact in the short run of just a fertility rate increase. This assumes each child represents a set of parents. While some of these new couples may account for two births, they also might have brought other children adding to the population. There is no definitive way to determine the reason or reasons for these added births or what impact they have had on the population until Census data becomes available after 2010. However, Council population and household estimates for 2004 are tracking close to the results of the original forecast model output from 2002. Subsequent revisions to local forecasts have resulted in increases to the regional total (see section below).

      • Deaths. The age-specific survival rates were very slightly increased over the forecast period. The survival rate is much more stable than the fertility or migration rates, and thus has less impact on the forecasts. In 2000, deaths were up slightly from forecasts, but from 2001 through 2004 deaths have fallen increasingly below the projections. For the five years since 2000 the net difference was 5,260 or 5.9% below what had been forecast, with the greatest deviation occurring in 2004. It is not possible to determine whether the lower number of deaths represents a higher survival rate or more out-migration of elderly who are living and dying elsewhere. An increase in survival rates would result in at most an added 5,000 more people in the region, but this would only be a temporary gain. If the drop in deaths were due to out-migration it would not result in any population gain and could reflect a higher out-migration of elderly who are still alive but not here.

      • Migration. Net migration was initially set at 12,000 per year reflecting historic age-related migration trends. The migration amount was adjusted every five years to reflect the region’s changing population and age composition. The annual net migration gain was forecasted to be about 13,700 by 2030. Since 1950 net migration has been fairly steady with two exceptions. Net migration averaged close to 10,000 per year in the 1950s, 1960s and 1980s, but averaged 13,400 per year in the 1990s and minus 4,000 annually in the 1970s.
   2. A number of alternate assumptions regarding births, deaths and migration were tested in the model to determine how sensitive the forecasts are to possible variation. Using various assumptions that have all been experienced in the past several decades results in an enormous range of forecasts. But major forecast errors occur only when trends consistently deviate over a long period of time. The expectation is that these trends will not be consistent over time, even for the assumptions that were input to the model, but that their impacts will average out over time. History has shown that this is not always the case--the baby boom and their subsequent household formation being prime examples. Measuring the history of fluctuating and divergent trends is a lot easier than predicting them.
      
      
   3. Revisions to the regional forecasts are sometimes made as a result of adjustments to the city and township forecasts. Communities are given an opportunity to comment on their forecasts.The Council is able to make appropriate adjustments, consistent with policy. Revisions to the 2002 Blueprint forecasts were made in 2004 for the Regional Development Framework based on local input. Although the focus of this process was on the household and employment forecasts at the local level, significant local household changes required adjustment to the community population forecast. The cumulative impact of such changes was to raise the regional population forecasts.
      
      Revisions were also made to a number of communities as part of the Systems Statement process in 2005 and 2006. The System Statement informs each community how it is affected by the Metropolitan Council’s policy plans for three regional systems – transportation, water resources (including wastewater collection and treatment), and regional parks and open space. For more on the Systems Statement see: www.metrocouncil.org/planning/framework/timeline.htm.
      
      The initial System Statement revisions, adopted in August 2005, were intended to resolve inconsistencies between the Development Framework forecasts and subsequent Council actions. Additional revisions to 21 communities were made based on local responses to the Systems Statements. The impact on the regional forecast of the 2005-06 revisions was to add 114,000 people to the 2030 model forecasts prepared for the Blueprint 2030. While this is a significant increase it still only represents three percent of the 2030 population, well within the range of reasonable expectations given the uncertainty of long-range forecasting.
      
      The forecast revisions had greater impact on the 2010 and 2020 forecasts. The current 2010 forecast is 93,000 above the 2002 model results for 2010. The forecast differences relate more to the timing of growth than the long-term (2030) outcome. Although the revised 2010 forecasts may not be reached, the 2030 forecasts are still well within a reasonable range of regional growth expectations. From a planning point of view it is better to plan for a little more growth to cover the inevitable fluctuation in growth trends. This is particularly true at the local level, where some places will exceed expectations while others will fall short. Planning for roads and wastewater treatment facilities requires substantial lead-time. An adequate supply of developable land with urban services available is needed to keep the cost of housing and land from increasing too fast.
      
      
   4. The regional forecasts are compared to national population forecasts and historic trends to provide a reasonableness check on the demographically derived forecasts.
      
      
      • The growth trends projected by the Council’s forecast model and subsequent revisions are reasonably consistent with the metro area’s historic share of national growth. Since 1940 the metro area has averaged between .5% and 1.5% of the nation’s growth per decade. The metro area's share of U. S. growth was 1.08% in the 1990s. As a share of the total national population the metro area has increased from .76% of the nation's total population in 1940 to .94% in 2000.
        
        The Census Bureau released new preliminary national forecast figures in 2004. These were up from previous forecasts based in part on a higher Census count in 2000. The Blueprint 2030 forecast share of the revised United States population growth forecast from 2000 to 2030 was 1.14%. The revisions to the Systems Statements based on local input increased the metro area’s share of national growth to 1.28%, and a share of U.S. total population in 2030 of just over 1 percent. The current revised forecasts for 2010 would account for 1.53% of national growth for the decade. While this appears to be high it is only slightly above the region’s share of population growth in the 1960s.
        
        
      • The Council's population forecast model produced growth trends consistent with those of the past 50 years. The region had its greatest growth of any decade in the 1990s, adding 353,000 people (4,000 over the previous peak that occurred in the 1960s). The 2002 forecast model (Blueprint 2030) projected 318,000 growth for the current decade. Although well below the previous decade this was more than the 303,000 growth of the 1980s. The forecast revisions for the final revised Systems Statements pushed that number up substantially to 411,000.
        
        

        The population growth forecasts are expected to be smaller over the following two decades, in response to the continued aging (and eventual mortality) of the baby boom generation. The Blueprint 2030 forecast of population growth is about 329,000, from 2010 to 2020 and just 274,000 from 2020 to 2030. The final Systems Statement forecast are higher at 364,000 for the 2010 to 2020 period and just 270,000 from 2020 to 2030. With the exception of the 1970s, when the region grew by only 111,000, population growth in every decade since 1950 has exceeded 300,000.

2. Household forecasts

   The population forecasts by age are converted to numbers of households based on a very consistent (since 1980) relationship between household formation and age. Households are the key element in the Council’s allocation to cities and townships because they are the major consumer of land. Household growth has been more stable over time than the population trends. Household growth from 1970 to 2000 ranged from 146,000 in the 1990s to 154,000 in the 1980s. The household trends diverge from the population trends as a result of the demographic impacts of the continued aging of the baby-boom generation.
   
   The city and township revisions to the Council forecasts entailed not only changes in population but also increases in the regional household forecasts. The overall regional household changes were not quite as much as the population forecasts. These changes were largely based on local input and were the driving force behind most of the population forecast revisions. The household trends do not track closely with the population trends due to the continued aging of the baby-boom generation. The net impact was a gain of 29,000 more households in 2030 than originally projected using the population forecast model, about 2.5 percent. As with revisions to the population forecasts discussed above, the greatest share of the additional growth is forecasted to be in the current decade.

3. Forecasts of housing demand by type

   The split between single-family and multifamily (including townhouses) is made to further assist the Council in determining future land demand. It is based on the projected numbers of households by age of head. Only 30% of households headed by persons under age 30 were in single family housing in 2000. From ages 40 to 65 the percentage in single family was between 71% and 73%. For older cohorts, the share drops steadily.

   For all age groups with household heads under age 65, the percentage living in single family housing has been declining since 1980. These projections continue these declining percentages, but at a slower rate. The percentage of new units that are single family is projected to average about 51% over the next thirty years (53%, 52% and 47% by successive decade). While down from the peak years of the early 1990s when single- family units made up nearly three quarters of all new units, it is about the same percentage as occurred from 1970 to 1990.
   
   In the six years since the 2000 Census, building permit data indicates that the share of single family detached housing was only about 45 percent. One determinant of the shift in the housing-mix trends since 1970 was the aging of the baby-boom generation. Their continued aging will drive these trends though 2030. Because housing trends tend to fluctuate it is not clear whether the recent trend is an indication of a stronger long-term multifamily trend (beyond demographic forces) or just a short-term fluctuation.

   Employment Forecasts

   Employment forecasts have historically been done by projecting the region's share of national forecasted growth (from the Bureau of Economic Analysis), which has been quite stable.

   A separate process projects the future labor force using age/gender inputs from the Council’s population forecast model, applying assumptions that include workforce participation rates for each group. Other assumptions are necessary to convert the projected future labor force to regional employment, including unemployment, self-employment, multiple job-holding, and net in-commuting. In the past, the two methods have produced compatible results. For the Blueprint 2030 forecasts, only the second (demographic-based) method was used due to lack of recent U.S. employment forecasts. To date the federal government has not indicated when or whether they will produce new long-term job forecasts.

   The forecast revisions since Blueprint 2030 have resulted in less change than was the case for the population and household forecasts, largely because the revision process was driven by local input and there was not much local input regarding the employment forecasts. The 2030 revised Systems Statement employment forecast was 25,000 above the Blueprint 2030 forecast.

III. Subregional Methodology

The subregional methodology is described in more general terms. Within the basic structure, policy considerations can greatly influence the forecasts. The other major determinant of the subregional forecast methodology is available data. Radial quadrants and concentric planning areas are used to provide a general guide for further allocations of the regional household and employment forecasts to cities and townships.

1. Quadrants

   Household growth trends by quadrant are very stable over time, reflecting people’s general tendency to locate close to their existing location when moving. Annually, about half the region’s residential growth occurs in the southern part of the region and half in the north (excluding the central cities). The north share has declined in the past several years because there is less available land. Growth is still occurring in the north, but just outside the seven-county area, most notably in eastern Wright and Sherburne Counties. The west half of the region generally gets about two-thirds of the annual growth and the other third is in the east. The east share rises each time a housing slowdown occurs.

2. Planning Areas

   Planning areas are roughly concentric rings defined by the Council, based largely on stage of development. There are currently six planning areas shown in the Council’s Regional Development Framework: Developed Area, Developing Area, Rural Center, Agricultural, Diversified Rural and Rural Residential. The planning area household growth trends reflect the historic outward growth of the region.

   Planning area trends are calculated as a share of their quadrant’s growth and these trends are projected to serve as starting point in the city-level allocation process. Ultimately they will reflect the supply of available land as determined through the city-level allocation. The most recent land supply data input to the Council’s forecasts is from the Council’s 2000 land use inventory. Redevelopment activities appear to be on the upswing, but the vast majority of growth in the region will require raw land at the region’s developing edge.

IV. City Level Methodology

The city level methodology continues the allocation process of the regional control forecast. It requires more judgement, because it entails balancing a number of diverse factors. The quality and detail of data available at any time have been a primary determinant of the city level methodology that was employed.

1. Household Allocation

   The first step is to extrapolate a city’s or township’s share of its quadrant and planning area’s past growth. Because these trends vary considerably from year to year, the trends are looked at for different time periods, with more weight given to the most recent trends to reflect emerging market forces.

   The trend-based forecasts are then adjusted to reflect the amount of developable land and how it is expected to be developed in terms of the share used for residential and nonresidential uses, and the mix and densities of single-family and multifamily land uses. This process is done in collaboration with local governments as part of the Council’s review of local comprehensive plans and any amendments that are subsequently done. The availability of local planned land use data has been of great value in enhancing this process. Over the past several decades the quality and detail of land use data has improved considerably. Continued improvements to the forecast database are expected, particularly in the area of redevelopment. This will require adjustment of the forecast methods. Continued improvements to the forecast database are expected, particularly in the area of redevelopment. This will require adjustment of the forecast methods.

   The current allocations also reflect the location and staging of regional infrastructure— wastewater treatment facilities and major transportation corridors. Efforts to protect key natural resources also play role in this process.

2. Population Conversion

   Household forecasts are converted to city/township populations using projected household size trends. The local community population conversions are adjusted to be consistent with the regional population forecast model output. Limited resources are devoted to this process, because local demographics are so idiosyncratic. That is why the Council urges use of household and employment forecasts, rather than population, for local planning purposes.

3. Employment Allocation

   The employment allocation process uses many of the same data inputs as the demographic allocation process, including past trends; land supply and land use patterns; infrastructure availability; and Metro Council and local policies and goals. Employment density assumptions are not used as directly as they are for the household forecast allocations to cities, since employment-generating activities vary substantially in land used per employee.

   Access to major transportation facilities is also an important consideration in the employment allocation process because employment tends to concentrate and intensify over time in accessible locations. The location and timing of new concentrations may be difficult to predict based solely on observations of market factors, but the forecasts also take into account communities' intentions for future development as described in their local comprehensive plans. The ratio of jobs to households is also considered in the context of the city’s stage of development.

4. City Level Adjustments

   As discussed previously, the Council sometimes implements interim revisions, without redoing the regional forecast model. This process typically involves soliciting local comments regarding city-level forecasts. Recent interim revisions have been focused on clarifying expectations for the next 10-year planning cycle.

V. Transportation Analysis Zone Allocations

For transportation planning purposes, the region has been divided into approximately 1,200 transportation analysis zones (TAZs). The Council asks cities and townships (where there is more than one TAZ) to allocate their overall forecast to TAZs within their community. The Council provides the community with the most current base year TAZ data for population, households and employment and the previous TAZ allocation. The Council determines the allocations for cities and townships that are unable to do them. The Council also adjusts the city numbers to meet Council forecasts if there are differences in community totals from Council forecasts.

VI. Forecast Accuracy

At the beginning of this summary, the purpose of the forecasts was stated as providing “reasonable” rather than accurate forecasts. Of course predictive accuracy is a desired goal, but in looking decades into the future there can be no certainty. Long-term accuracy usually has less to do with the methodology than on the stability of the trends on which a forecast model’s assumptions are based. The baby boom and bust, major changes in immigration patterns over the past half-century, and the political and economic conditions that caused them, were not anticipated by anyone. Similar dramatic changes in the future, which we cannot predict, could affect demographic trends similarly — or they could have different effects because a different generation will be involved. The most effective way to respond to this uncertainty is to monitor the trends. This will better prepare us to respond quickly and effectively to any changes that occur.