Forecasting Local Employment

1
Forecasting Local Employment

• There are two general approaches to forecasting
  employment
• 1) The Unique Economy The local economy differs
  so substantially from the economies of the region
  or nation that it must be treated as a unique
  entity. --Analysis using pattern areas is
  thought to be of little use. --In these cases
  you might use the extrapolation/curve fitting
  techniques to make forecasts about this economy
  on an industry by industry level.2) The
  Integrated Economy The local economy is closely
  intertwined with the regional and national
  economy. Local growth is assumed to be closely
  related to the growth of the nation, state, or
  region. --Using projections of national or state
  employment, you can derive local employment
  projections the analyst uses the constant share
  or shift share techniques to make projections.

2
Analyzing the Unique Economy

• If an analyst determines that a local economy is
  indeed a unique economy, then they might choose
  to utilize the extrapolation technique to
  generate a forecast about that economy.
• Example

3
Analyzing the Unique Economy Contd

• Using this data we can generate projections using
  the extrapolation technique. A Linear Curve
  yields the following results

4
Analyzing the Integrated Economy

• If the analyst determines that an economy is
  indeed well-integrated with the regions or the
  nations economy (the changes experienced in the
  local economy reflect changes in these pattern
  areas), then typical methods of forecasting
  employment include
• 1) Constant Share (CS) technique
• 2) Shift Share (SS) technique
• 3) Variations on the SS technique
• These techniques utilize historic employment data
  for a larger pattern area to project the
  employment of a component sub area.
• A primary assumption of these techniques is that
  the growth of a smaller area will emulate the
  growth of its pattern area because a similar set
  of factors affect both of these geographic
  regions (an integrated economy).
• The usual set of comparisons
• For Counties 1) County to Region 2) County to
  State
• For Cities 1) City to County 2) City to Metro
  Area 3) City to State

5
Constant Share Approach Basics

• The constant share projection approach assumes
  that the local share of a larger regions
  economic activity remains constant. We assume
• eit remains constant for all industries
• Eit
• This approach assumes that this share is constant
  (a constant share) that it is invariant over
  time. This can be true only if employment in the
  industry grows at the exact same rate both in the
  local economy and in the reference region
  economy.
• To make our local employment projections we use
  the following formula
• eit (1 Rit-t) eit
• where eit projected local employment in
  industry i in time t
• eit local employment in industry i in time
  t Rit-t projected growth rate for the
  reference region for the time period from t to
  t

6
Constant Share Approach Continued

• To calculate the Reference Region Growth Rate for
  each industry we use the following formula
• Rit-t Eit - Eit OR (New - Old)
• Eit Old
• We then assume that the Local Region Growth Rate
  is the same as the Reference Region Growth Rate.
• Example Food Products Employment in Chapin
  County
• FP Emp. State of Florida 2003 44,400
  (actual) 2015 48,200 (projected)
• FP Emp. in Chapin County 2003 1,700 (actual)
• 2015 ??? (projected)
• Florida FP Growth Rate 2003-2015 (48,200-
  44,400) /44,400 .08559
• Chapin FP Projected Employment (1 Rit-t)
  eit
• (1 .08559) 1,700
  1,846

7
The Shift Share Approach

• The Constant-Share Approach assumes that the
  local economy will maintain a constant share of
  a reference regions employment. This is very
  rarely the case however.
• The Shift-Share Approach was developed in an
  effort to address this shortcoming.
• In some cases, in industry i the local economy
  will grow faster than the reference region
  economy, in other cases it will grow slower. The
  Shift-Share Approach uses an observed trend
  (faster rate of growth or slower rate of growth)
  and applies this trend into the future.
• To project local employment using this approach
  we modify the Constant-Share projection formula
• eit (1 Rit-t sit-t) eit
  where
• eit projected local employment in industry i
  in time t
• eit local employment in industry i in time t
• Rit-t projected growth rate for the
  reference region for the time period from t
  to t
• sit-t projected local employment shift for
  industry i for the time period from t to t

8
Calculating the Shift Term

• The formula for projecting local employment using
  the Shift-Share
• eit (1 Rit-t sit-t) eitNote
  --When the shift term is positive (a greater
  local growth rate in industry i) projected
  employment is greater than the constant share
• -- When the shift term is negative (a smaller
  local growth rate in industry i) projected
  employment is less than the constant share
• The Shift Term is equal to the difference between
  the industrys local growth rate and the
  industrys regional growth rate
• sit-t rit-t - Rit-t
• The Projected Shift Term is unknown, so we
  typically use a historical shift term for which
  we have data available.
• The Shift Term is then held constant into the
  future when we project local employment. Often
  this term is referred to as a Constant Shift Term
  because we do not vary it over time (we assume
  that the shift does not vary over time).

9
Calculating the Shift Term Example

• Food Products Employment 2000 2003 2015
• State of Florida 47,300 44,400 48,200
• Chapin County 1,525 1,700 ???
  Projected value
• Calculating the Growth Rates
• FL FP Growth Rate 2000-2003 (44,400- 47,300) /
  47,300 -0.06131
• Chapin FP Growth Rate 2000-2003 (1,700- 1,525)
  / 1,525 0.11475
• Calculating the Three Year Shift Term (2000-2003)
• sit-t rit-t - Rit-t gt Shift Term
  Local Shift - Regional Shift
• sit-t .11475 - (-.06131) 0.17606
• Note, however, that we have a problem. The
  periods of time for our shift term and the
  projection period do not match
• Shift Term 3 Years (2000 to 2003)
• Projection Term 12 Years (2003 to 2015)
• We need to adjust our three year shift term to a
  twelve year period.

10
Adjusting the Shift Term Example

• To adjust a shift term to a different projection
  horizon, we use exponents.
• Calculating the Twelve Year Shift Term
  (2003-2015)
• Rim (1 Rin)m/n 1 where
• Rin the shift term for industry i in time
  period n
• n the length of the shift term period
• m length of projection horizon
• Rim (1 .17606)12/3 -1 0.91301
• Once we have adjusted our shift term
  appropriately, we can then calculate our
  employment projections.
• Projecting Chapin County Food Products Employment
  in 2005
• Chapin FP Projected Employment (1 Rit-t
  sit-t) eit
• (Rit-t from Constant Share) (1
  .08559 0.91301) 1,700 (1.99860)
  1,700
• 3,398

11
Evaluating these Projection Techniques

• These projection techniques are popular because
• --They are conceptually and computationally
  straightforward
• --They have low data requirements
• --They provide fast and easily accessible results
• Other more advanced techniques, such as
  Input-Output and Econometric models (to be
  discussed next week), are difficult to
  understand, more complex to develop, and
  generally require much more data.
• These techniques have been heavily criticized
  from the beginning. Why?
• 1) Each technique makes major, often
  incorrect, assumptions
• --Rarely does the local area economy mirror
  changes in the pattern area economy (so the
  Constant-Share is unsound)
• --Rarely do past economic trends continue in the
  same form in the future (so the Shift Share is
  unsound)
• 2) No Theoretical Foundation for understanding
  employment change
• Empirically, however, these techniques have been
  found to provide cheap, accessible, and
  relatively good results, although generally not
  as accurate as more advanced techniques
  (particularly I/O models).
• --Of particular interest to us, the Constant
  Share approach provides good, county-level
  projections with a few, small industrial sectors
  and limited local data.

12
Variations on the SS Technique

• A great deal of research has discussed the
  viability, theoretical foundations, and
  effectiveness of these techniques in projecting
  employment.
• In particular, researchers have noted the
  tremendous variability and fluctuation in local
  shift terms over time in single industries.
• For example, if Chapin County sees a new mall
  open, creating 2,000 new retail jobs between 2000
  and 2005, the shift share approach assumes that a
  new mall will open every five years in the
  County. This results in inflated employment
  projections in the retail sector.
• As a result of this work, several variations of
  the shift share approach have been suggested.
  These variations include
• 1) Average the local shift term across all
  industries to minimize the impacts of large short
  term increases or decreases in a single industry.
• 2) Use a generalized version of the shift share
  approach that includes a decay function to
  reduce the impact of a large local shift term
  over time.
• 3) Cap and/or floor the local shift term to
  reduce the impact of short term changes to the
  economy.

13
Employment Projections Data

• As with the ratio technique used to make
  population projections, to make employment
  projections for a local area, we need good
  projections for our larger pattern area.
• Sources for projection data include
• 1) OBERS projections Federally prepared,
  long-term
• 2) AWI projections Prepared at the
  state/regional level, medium term
• 3) Regional/Local Planning Entity projections
  Prepared at the regional or local level, short to
  long term.

14
OBERS Employment Projections

• Projection data used to come in the form of OBERS
  projections developed by the Bureau of Economic
  Analysis. OBERS were prepared for the nation,
  states, MSAs, and some other substate areas, by
  the Office of Business Research (OBE, now BEA)
  and the Economic Research Service (ERS, situated
  in the USDA).
• -OBERS projections were prepared for personal
  income and employment.
• -The projections were updated about every 5 years
  (not on a regular cycle).
• -They were prepared using a step-down procedure
  (NationgtStategtRegion)
• -At the National and State level projections for
  56 industry groups were provided. At substate
  levels the projections are for 14 more aggregated
  industrial groups. (See Klosterman pp. 171-172
  and Appendix C)
• -The convention was to use the more disaggregated
  state level projections to develop local
  employment projections.
• These projections are no longer prepared by the
  federal government for states and smaller areas
  the program stopped in 1996.
• However, projections are prepared every two years
  for national employment for a ten year horizon
  (2000-2010, for example). See the following web
  page for more information on these projections
• http//www.bls.gov/emp/home.htm

15
OBERS Employment Projections for FL
16
AWI Employment Projections

• Employment projections are now prepared by most
  states. In Florida, these projections are
  currently prepared by the Agency for Workforce
  Innovation.
• According to the AWI web site, these projections
  are prepared annually for a horizon of 8-10 years
  (2001 BY projections were for 2009, 2002 BY
  projections go to 2010).
• These projections are prepared for 1) the state
  as a whole 2) for the states 24 workforce
  regions 3) counties with more than 100,000 jobs
• These projections are produced via
  extrapolations of past trends combined with
  knowledge of current and expected events.
• --They use linear and multiple regression models
  that use population projections and national
  forecasts of employment by industry as predictor
  (independent) variables
• --They use eyeballing techniques (what they
  term graphical displays of historical industry
  employment) to aid in the selection of a
  forecast.
• The 2001-2009 forecasts are available in PDF
  format at
• http//www.labormarketinfo.com/Publications/publ
  ications.htm
• The 2002-2010 forecasts are available in Excel
  format at
• http//www.labormarketinfo.com/oes-proj/oes.htm

17
Floridas Regional Workforce Boards Map courtesy
of WorkforceFlorida.com
18
(No Transcript)
19
Regional Employment Projections

• In larger urban areas, employment projections are
  often prepared by regional or local planning
  agencies.
• These projections are generated by a wide array
  of techniques and are often of higher quality and
  greater utility to local planners because they
  are focused on a particular metropolitan area.
• Examples
• Southern California Association of Governments
  (SCAG)
• http//www.scag.ca.gov/forecast/retp.htm
• Toronto http//www.city.toronto.on.ca/to
  rontoplan/flashforward.htm
• Seattle
• http//www.psrc.org/datapubs/pubs/forecasts_2
  002.htm

20
Theoretical Issues (extra slide)
Despite the attractiveness of these techniques,
due to four reasons 1) Use of a pattern
area 2) Industry level analysis 3) Simple
calculations 4) Low data requirements these
techniques have some substantial issues. What
theoretical issues arise in the use of the
Constant Share and Shift Share? 1) We make one
of two major assumptions for each method 1)
that the local share of an industrys total
employment will remain constant over time
(constant Share) OR 2) that the observed change
in share (the shift in share) will remain
constant over time. Both of these assumptions can
be questioned. 2) These methods require
projections for pattern areas that are sound and
relatively on target.