Printer Friendly

Alternative Measures to Improve Demand Forecasts.

Abstract

Real estate appraisers typically rely on an application of economic base theory to generate forecasts of property demand. This article posits that a dynamic version of economic base theory may generate more accurate forecasts in current times when the structures of metro economies are often in great flux. The study examines dynamic measures using statistical analysis to evaluate their forecasting merit. The results indicate the Modified Lilien Index and the Hirschman-Herfindahl Index may be useful indicators for forecasting employment growth. The findings are drawn from recent research supported by The Appraisers Research Foundation.

The Problem

In Level C market analysis, appraisers are expected to make quantitative forecasts of relevant demand indicators--most prominently employment--in order to forecast absorption or to use reliable third-party sources for employment estimates. (1) These forecasts of demand usually apply economic base theory, which analyzes the economic structure or industry mix of a metro area. (2) Once the metro-level analysis has been completed, appraisers can zoom in to analyze the market area that contains the subject property. (3)

Metro-level employment forecasts are more accurate when the economic base of the metro area is relatively stable. In most metros, however, the economic structure changes over time, often quite dramatically. How can forecasting accuracy be improved given continual structural change? One approach is to use a dynamic version of the economic base to study changes in economic structure that have occurred over the years. A sound dynamic analysis of the economic base could provide a more accurate basis for forecasting employment.

This article offers alternative indicators to forecast employment growth derived from an analysis of the changes in the economic base of selected US metro areas. The next section summarizes relevant discussions of dynamic economic base. Then, the discussion presents the empirical analysis that involves "back casting" employment growth with alternative measures. The final section suggests several ways to use this work in appraisal practice including a straightforward way to incorporate these assessments in discounted cash flow analysis.

Related Previous Work

In a dynamic and competitive global economy, metro areas that improve existing specializations or adapt to changing conditions by adopting new specializations should retain or improve their competitive advantage over the long term. They should outperform metro areas with stagnant economic bases--ones that are unwilling or unable to adapt and change. (4) Recent studies sponsored by the Land Economics Foundation and The Appraisers Research Foundation (TARF) have identified and tested dynamic economic base measures and their impacts on economic outcomes. (5)

An indicator that could improve employment growth forecasts is one that directly measures the amount of structural change that has occurred in a metro economy during a period. Such an indicator is the Lilien Index, applied here in its modified form. Two other indicators have been used for many years to study economic structure: the Hirschman-Herfindahl Index, which is an absolute measure of specialization, and the Krugman Specialization Index, which is a relative measure. These measures of economic structure can be calculated for various periods. Changes in these indexes over time indicate structural change. (6) These three indexes are defined and described in Appendix A.

Empirical Results

The research conducted for TARF, which led to the recommended predictors of employment growth, is summarized in the following subsections addressing the units of analysis, the database, the measures and correlations, the regression analysis, and the empirical findings.

Units of Analysis

The US Census Bureau defined 366 metropolitan statistical areas (MSAs) in 2009 and added 16 more in 2013. For this research, 52 large MSAs were examined; the New York and Los Angeles MSAs were excluded because they tend to be outliers that skew statistical results. Exhibit 1 lists in alphabetical order the MSAs studied.
Exhibit 1 Metro Areas under Study

Atlanta, GA                 Milwaukee, Wl
Austin, TX                  Minneapolis-St. Paul, MN
Baltimore, MD               Nashville, TN
Birmingham, AL              New Orleans, LA
Boston, MA-NH               Oklahoma City, OK
Boulder, CO                 Orlando, FL
Buffalo, NY                 Philadelphia, PA-NJ-DE
Charleston, SC              Phoenix, AZ
Charlotte, NC-SC            Pittsburgh, PA
Chicago, IL-IN-WI           Portland, OR-WA
Cincinnati, OH-KY-IN        Providence, RI-MA
Cleveland, OH               Raleigh-Cary, NC
Columbus, OH                Richmond, VA
Dallas-Fort Worth, TX       Riverside-San Bernardino, CA
Denver, CO                  Rochester, NY
Des Moines, IA              Sacramento, CA
Detroit, Ml                 Salt Lake City, UT
Durham-Chapel Hill, NC      San Antonio, TX
Hartford, CT                San Diego, CA
Houston, TX                 San Francisco-Oakland, CA
Indianapolis, IN            San Jose, CA
Kansas City, MO-KS          Seattle-Tacoma, WA
Las Vegas-Paradise, NV      St. Louis, MO-IL
Louisville, KY-IN           Tampa-St. Petersburg, FL
Memphis, TN-MS-AR           Tulsa, OK
Miami-Fort Lauderdale, FL   Washington, DC-VA-MD


Database

The Woods and Poole Economics, Inc., database was accessed to conduct the analysis. Woods and Poole Economics is an independent firm in Washington, DC, that produces long-term economic and demographic forecasts for states, regions, MSAs, and other subareas of the United States. It has consistent historical data for MS As since 1969 as well as annual forecasts through 2050. Woods and Poole's projections are based on economic base theory. Its data have been widely used since the mid-1990s when the Bureau of Economic Analysis in the US Department of Commerce stopped making these projections. (7)

The US Bureau of Census has developed the North American Industry Classification System (NAICS) for "classifying business establishments for the purpose of collecting, analyzing, and publishing statistical data related to the US business economy." (8) For this study, eighteen NAICS sectors were selected that could potentially contribute to a metro area's export base. These sectors include utilities, manufacturing, wholesale trade, retail trade, transportation and warehousing, management of companies, administrative services, arts, entertainment and recreation, accommodation and food services, information services, professional services, financial services, educational services, health care and social services, other services, federal civilian earnings, federal military earnings, and state and local government. (9) The economic structure indicators were calculated using earnings in constant 2005 dollars for these eighteen NAICS sectors.

Measures and Correlations

Exhibit 2 shows the dependent and independent variables on which the study focused. The dependent variable is employment growth, measured as a ratio of the 2016 employment level divided by the 2010 employment level. The three indexes were calculated forl970 through 2010. The correlations between the indexes and employment growth were examined to select the best measures for the regression analysis. Independent variables found to be significant beyond the 5% level were treated as the most promising factors to consider.

The Modified Lilien Index (MLI) for 19902000 performed better than the MLIs for 1970-1980, 1980-1990, and 2000-2010. The Hirschman-Herfindahl and Krugman Specialization Indexes comparing change from 1970 to 2000 had higher correlations with employment growth than change from 1980 to 2010. Interestingly, measures for earlier periods performed better than ones based on more recent data for 2010.

Regression Analysis

In the study, ordinary least squares analysis (OLS) was used to simulate forecasting. The statistical tests for collinearity and heteroscedasticity indicate that the OLS results are not biased. After considering all combinations of the three independent variables, the model shown in Exhibit 3 was found to have the highest adjusted R-squared in explaining the variation in the 2010-2016 employment growth rate.

Empirical Results

Together, the 2000/1970 HHI ratio and the MLI for 1990-2000 accounted for almost 45% of the variation. The HHI ratio was significant at the 1.0% level; the MLI was more significant at the 0.1% level. (10)

The three metro areas that experienced the greatest amount of structural change in the 1990s, as measured using the MLI, were Charleston, SC, Austin, TX, and Seattle, WA. At the other extreme were Milwaukee, WI, Birmingham, AL, and Pittsburgh, PA. As for change from 1970 to 2000, as measured in the HHI, Durham, NC, ranked first, followed by San Jose, CA, and Las Vegas, NV. Rochester and Buffalo, NY, and Charlotte, NC/SC, experienced the least amount of change during the study period. (11)

Applications for Appraisal Practice

The findings from this research suggest that appraisers and real estate market analysts may consider using indexes of structural change in their market analyses. Specifically, the Modified Lilien Index and the Hirschman-Herfindahl Index are worth considering when forecasting employment growth. Appendix B offers a methodology and provides numerical examples showing how one can forecast employment growth with these two indexes.

One important caveat pertains to the time frames under study. Although historical measures for 2010 or earlier were found to impact outcomes after 2010, these relationships may not hold in the future. Therefore, appraisers should conduct back casting tests with different time frames as more recent data become available.

The gist of this analysis suggests that appraisers may find it helpful to extend the idea of comparables from the property level to the metro level. Appraisers identifying other metro areas with similar specializations and economic structure may find it informative to communicate with colleagues in these locations. For example, Charleston, Columbia, and Greenville, despite having about the same population size and being located in South Carolina, have very different metro economies and therefore specializations that are not comparable. On the other hand, it may be worthwhile for appraisers working in the Raleigh, NC, Austin, TX, and Nashville, TN, metros to collaborate and share their long-term forecasts because these metro areas have similar specializations.

Appraisers interested in identifying comparable metro areas can take the following steps. First, compute location quotients (LQs) for all important export sectors in the metro area. Second, identify all MSAs of similar size based on employment and population; economic structure tends to correlate more with size than regional location. Third, compute LQs for these export sectors in all potentially comparable metro areas. Next, select the metros with the most similar profile of LQs. Fifth, introduce qualitative factors to make the final selection of comparable MSAs.

The final suggestion is that appraisers incorporate their evaluation of the economic structure of the metro area in which the subject is located to gauge long-term market risk. To encourage this application, the 52 MSAs in this study were evaluated to assess long-term risk. As shown in Exhibit 4, the metro areas were categorized into three risk groups: stronger lower-risk metro economies, metros with moderate risk, and higher-risk metro areas.

When using the income capitalization approach to value, the terminal (going-out, exit, or residual) capitalization (cap) rate could be adjusted to account for the long-term economic prospects of the metro area. In metro areas with moderate risk, the estimated terminal cap rate would remain unchanged. In lower-risk metro areas, the terminal cap rate could be decreased by, say, 25-45 basis points. Conversely, in higher-risk areas, the cap rate could be increased by 25-45 basis points. With this approach, the metro area's economic prospects would have a direct impact on the value of its income-generating properties.

Appendix A Measures of Economic Structure

In the current study, three measures of economic structure are examined to estimate the dynamism of each metro area's economic base. The Modified Lilien Index (MLI) computes the absolute difference between change in earnings (or employment) in one sector to that in all sectors for a time period, therefore directly measuring structural change overtime. Higher MLI values mean greater flux among sectors and thus more dynamism in the metro economy for the time period. In the analysis, the MLI calculated from 1990 to 2000 was the most effective forecaster of employment growth from 2010 to 2016.

The Hirschman-Herfindahl Index (HHI) is an absolute measure of specialization, while the Krugman Specialization Index (KSI) is a relative one that measures how different one MSA's economy is from the composite economy of all metro areas. In the analysis, the change in these specialization measures from 1970 to 2000 was used to predict employment growth from 2010 to 2016, the assumption being that more change reflects more dynamism. Specifically, increases in the HHI ratios indicate higher levels of organization in the metropolitan economy as sector specializations increase. Larger absolute changes in KSI indicate greater differences between one metropolitan economy and the combined economic structure of all metropolitan areas. The mathematical formulas for these three indicators are shown below.

Modified Lilien Index (MLI)

MLI = SQRT ([SIGMA] [W.sub.i] * [[In([x.sub.it]/[x.sub.it-1]) - In([X.sub.it]/[X.sub.it-1])].sup.2])

Where: [SIGMA] = sum over all sectors

X = share of earnings in 2005 dollars

i = one of 18 two-digit sectors

t = time period (for example, 2000)

t - 1 = previous time period (for example, 1990)

X = total earnings

W = average share of total earnings for sector i in time periods f and f - 7

To calculate the MLI: (1) compute the ratio of metro earnings in sector i in time f by earnings in that sector in time period t - 1 and the ratio of total earnings for these two time periods; (2) take the natural log of each ratio; (3) for all i sectors, subtract the sector-specific log from the log for all sectors; (4) square this difference to eliminate negative values; (5) weight each difference by the average share of sector earnings for the two time periods; (6) sum the weighted shares over all sectors; and (7) take the square root of this sum.

Hirschman-Herfindahl Index (HHI)

HHI = [SIGMA] [X.sup.i.sub.a]

Where: [SIGMA] x and i are defined the same as above for MLI

a = coefficient usually 2 (squared)

Note, the higher the coefficient value the greater the weight given to sectors with a larger share of earnings.

Krugman Specialization Index (KSI)

KSI = [SIGMA] [absolute value of [x.sub.i] - [X.sub.i]]

Where: [SIGMA] x and i are defined the same as above for MLI

X = average value of earnings in a sector for all metropolitan areas

Appendix B Forecasting Employment with Measures of Dynamic Economic Base

Appraisers need to forecast absolute employment to estimate property absorption for a target year. In the current study, however, none of the measures of dynamic economic base proved to be correlated with absolute employment. What is correlated is the level of absolute employment in the base year. The correlation between absolute employment change from 2010 to 2016 and total employment in 2010 is very high at +0.824, far stronger than any other measure under study. Fortunately, the economic structure measures do correlate with the employment growth rate. Therefore, the best way to forecast absolute employment change for one metro area is in two steps: estimate the employment growth rate with economic structure measures for that metro area and then multiply that rate by the metro area's employment level in the base year.

For the subject property's metro area, appraisers need an equation that predicts the metro area's growth rate with the structural indicators of dynamic economic base. Then, they can insert the HHI and MLI values for that area into the equation to find the metro area's growth rate. We can treat the regression model in Exhibit 2 as an appropriate equation to show the relationship between the six-year employment growth rate (employment in 2016 divided by employment in 2010) and the HHI and MLI:

Employment Growth Rate = 0.9601 + 0. 12599 HHI + 0.33574 MLI

Substituting the average values for the HHI (0.7619) and the MU (0.2366) for all areas into this equation, we find that the average growth rate from 2010 to 2016 for the 52 metro areas is 1.1355. This growth rate can be compared to the one for all 382 metro areas based on Woods and Poole data. According to Woods and Poole data, the employment growth rate from 2010 to 2016 for 382 metro areas is lower at 1.1119 (total employment in 2016 divided by total employment in 2010).

To illustrate this process, let's examine data for the Denver, Baltimore, and Indianapolis MSAs, as displayed in Exhibit 5.

Beginning with the Denver MSA, Exhibit 5 shows Woods and Poole's 2016 employment level and the 2022 employment forecast for Denver. Since Denver's employment was 1,613,053 in 2010, its 2010-2016 growth rate was 1.1644 and the forecasted growth rate for Denver from 2016-2022 is 1.1070. Applying the regression equation above with Denver's values for the HHI and MLI, the estimated growth rate from 2010 to 2016 is 1.1907, which is 0.0263 higher than Denver's actual growth for 2010-2016 according to Woods and Poole data. This higher estimated growth rate is consistent with the evaluation of Denver as one of the stronger metro economies among the 52 in the sample.

If the estimated 2010-2016 growth rate of 1.1907 persists for the next six years, the 2022 employment forecast for Denver would be 2.236 million as shown in Exhibit 5. Therefore, the 2022 forecast on the basis of dynamic economic base measures turns out to be 157,000 employees higher than Woods and Poole's 2022 forecast. However, perhaps the forecast should account for the 5.7% reduction in the Denver growth rate predicted by Woods and Poole for 2016-2022 compared to 2010-2016. To make the Denver forecast less aggressive, the growth rate can be reduced by the difference between the Woods and Poole estimate for 2022/2016 (1.1070) and the actual Woods and Poole growth rate for 2016/2010 (1.1644). The more conservative 2022 employment forecast for Denver using the lower growth rate is 2.129 million, which is 49,000 higher than the Woods and Poole forecast.

This methodology can be repeated for Baltimore, a moderate-risk metro area, and Indianapolis, a higher-risk metro area. Compared to Denver's 1.1907 estimated growth rate, the estimated 2010-2016 growth rate for Baltimore is 1.1264, and the growth rate for Indianapolis is lower at 1.1039. With the regression equation, the initial 2022 forecasts are higher than Woods and Poole's by 56,000 for Baltimore and by 25,000 for Indianapolis. Reducing the growth rate in the same way as for Denver results in modified 2022 forecasts. For Baltimore, the forecast is essentially the same as the Woods and Poole forecast. For Indianapolis, the modified forecast is 16,000 employees lower than the Woods and Poole forecast. In comparison to the 2022 Woods and Poole forecasts then, modified employment forecasts based on the MLI and HHI indexes are higher for Denver, about the same for Baltimore, and lower for Indianapolis as shown in the last column Exhibit 5.
Exhibit 5 Comparison of Woods and Poole
Forecasts to Forecasts Based on MLI and HHI

                         Woods & Poole Forecasts

MSA                      Employment    Forecast
                            2016      Employment
                                         2022

Denver, CO               1,878,207    2,079,194
Denver modified          1,878,207    2,079,194
Baltimore, MD            1,813,551    1,986,510
Baltimore modified       1,813,551    1,986,510
Indianapolis, IN         1,271,050    1,378,495
Indianapolis modified    1,271,050    1,378,495

                          MLI - &
                         HHI-Based
                         Forecasts

MSA                       Forecast     Difference:
                         Employment   Col 4 - Col 3
                            2022

Denver, CO               2,236,381      +157,187
Denver modified          2,128,572       +49,378
Baltimore, MD            2,042,784       +56,274
Baltimore modified       1,986,546           +36
Indianapolis, IN         1,403,100       +24,605
Indianapolis modified    1,362,899       -15,596


Additional Resources

Suggested by the Y. T. and Louise Lee Lum Library

Appraisal Institute--Education Courses

https://www.appraisalinstitute.org/education/

* Advanced Concepts and Case Studies

* Advanced Market Analysis and Highest & Best Use

Federal Reserve of St. Louis, FRED Economic Data--Employment

https://fred.stlouisfed.org/categories/32444

Integra Realty Resources--Viewpoint, 2017 Commercial Real Estate Trends

http://www.atrustrealty.com/images/IRR_2017_AnnuaLViewpoint.pdf

Land Economics Foundation--Funded Research

https://lai-lef.org/funded-research/

Woods and Poole Economics, Inc.--Databases

https://www.woodsandpoole.com/our-databases/

Emil Malizia, PhD, FAICP, is president of Malizia & Associates, LLC. He also directs the Institute for Economic Development at the University of North Carolina, Chapel Hill. His research and practice focus on vibrant centers, downtown redevelopment and urban economic development, and entrepreneurship. Contact: malizia@email.unc.edu

Andrew Malizia is a research associate at Malizia & Associates, LLC. He holds a BA in mathematics from Dartmouth College. Contact: malizia.llc@gmail.com

Acknowledgments

We would like to thank Josh Drucker, Emil Evenhuis, Steve Fanning, Ed Feser, Hugh Kelly, Michael Odern, Henry Renski, and Leslie Sellers for their input. We also thank Lorin Bruckner for her assistance preparing the exhibits. We are very grateful for the support provided by The Appraisers Research Foundation and guidance from Tim Leberman.

(1.) See chapter 6 in Stephen F. Fanning, Market Analysis for Real Estate, 2nd ed. (Chicago: Appraisal Institute, 2014). In addition to economic base analysis, Fanning discusses shift-share analysis.

(2.) On economic base theory, see Hans Blumenfeld, "The Economic Base of the Metropolis," Journal of the American Institute of Planners 21 (1955): 114-132; Homer Hoyt, "Homer Hoyt on the Concept of Economic Base," Land Economics 30 (May 1954): 182-186; and Emil E. Malizia and Edward J. Feser, chapter 3 in Understanding Local Economic Development (New Brunswick, NJ: Center for Urban Policy Research, 1999).

(3.) Although beyond the scope of this article, stepping down forecasts to small areas is very challenging. Appraisers with access to CoStar can use its submarket data to help carry out this task. See www.costar.com.

(4.) For a relevant conceptual and empirical work see Michael Storper, Keys to the City: How Economics, Institutions, Social Interaction, and Politics Shape Development, Princeton, NJ: University Press, 2013), especially parts HI; and Enrico Moretti, The New Geography of Jobs (New York, NY: Houghton, Mifflin, Harcourt, 2012).

(5.) See Emil Malizia, Central City Decline from the Perspective of Long-Term Economic Change (Chapel Hill, NC: Land Economics Foundation, Spring 2016) available at http://bit.ly/LEFresearch; and Emil Malizia and Andrew Malizia, Improving Demand Forecasts with Alternative Measures of Economic Base (The Appraisers Research Foundation, August 2017). Although the TARF study examines both employment growth and income levels, this article focuses on the employment growth forecasts.

(6.) See Muhammed Ansari, Chiara Mussida, and Francesco Pastore, "Note on the Lilien and Modified Lilien Index," Discussion Paper 7198 (IZA Institute of Labor Economics, Bonn, 2013); Albert Hirschman, "The Paternity of an Index, American Economic Review 54 (1964): 761-762; Paul Krugman, Geography and Trade (Cambridge: MIT Press, 1991); Nicole Palen, "Measurement of Specialization: The Choice of Indices" (FIW Working Paper 62, Vienna, December 2010).

(7.) Woods and Poole Economics, Complete Economic Demographic Date Source Technical Documentation (2013). These data are available though university or research libraries for those who not subscribers.

(8.) US Census Bureau, https://www.census.gov/eos/www/naics/.

(9.) Excluded are primary sectors, construction, and real estate and rental services.

(10.) One interesting finding is that the structural indicators used to successfully back cast employment growth turned out not to be good back casters of Income levels. The factors that back casted income at the metro level were change in manufacturing earnings, the depth and talent of the labor pool, and patent production as an Indicator of entrepreneurship. These three factors measured in 2010 or earlier explained 67% of the variation in average household income in 2016.

(11.) The authors will make the entire data set available to readers interested in the results for other metro areas. Please contact Emil Malizia at malizla.llc@gmail.com.
Exhibit 2 Variables, Measures, and Data Sources

Variables              Variable Description        Years

Dependent

Demand Indicator       Employment growth rate      2016/2010

Independent

Change in              Ratio of two HHI values     2000/1970
Hirschman-Herfindahl
Index (HHI)

Change in Krugman      Absolute differences        2000-1970
Specialization         between two KSI values
Index (KSI)

Modified Lilien        MLI values for one period   1990-2000
Index (MLI)

Variables              Data Sources

Dependent

Demand Indicator       Mayors' Report *

Independent

Change in              Woods & Poole, earnings
Hirschman-Herfindahl   in 2005 dollars
Index (HHI)

Change in Krugman      Woods & Poole, earnings
Specialization         in 2005 dollars
Index (KSI)

Modified Lilien        Woods & Poole, earnings
Index (MLI)            in 2005 dollars

* US Conference of Mayors, Metro Economies:
Past and Future Employment Levels, prepared
by IHS Markit, May 2017.

Exhibit 3 Best Model for Employment Growth 2010-2016

                  Coef.     t-value   p-value

Intercept         0.96010   31.267    < 2e-16 ***

HHI 2000/1970     0.12599   3.174     0.0026 **

Modified Lilien   0.33574   4.615     2.85e-05 ***
1990-2000

Adjusted          0.4497
R-squared

Significance: * 5.0% level ** 1.0% level *** 0.1 % level

Exhibit 4 Risk Assessments for Metro Areas under Study

Lower Risk               Moderate Risk              Higher Risk

Austin                   Atlanta                    Birmingham
Boston                   Baltimore                  Buffalo
Boulder                  Charlotte                  Cleveland
Charleston, SC           Chicago                    Hartford
Denver, CO               Cincinnati                 Indianapolis
Durham-Chapel Hill       Columbus                   Louisville
Raleigh-Cary             Dallas-Fort Worth          Milwaukee
Sacramento               Des Moines                 New Orleans
San Francisco-Oakland    Detroit                    Oklahoma City
San Jose                 Houston                    Pittsburgh
Seattle-Tacoma           Kansas City                Providence
                         Las Vegas-Paradise         St. Louis
                         Memphis
                         Miami-Fort Lauderdale
                         Minneapolis-St. Paul
                         Nashville
                         Orlando
                         Philadelphia
                         Phoenix
                         Portland, OR
                         Richmond
                         Riverside-San Bernardino
                         Rochester
                         Salt Lake City
                         San Antonio
                         San Diego
                         Tampa-St. Petersburg
                         Tulsa
                         Washington, DC
COPYRIGHT 2018 The Appraisal Institute
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Peer-Reviewed Article
Author:Malizia, Emil; Malizia, Andrew
Publication:Appraisal Journal
Article Type:Report
Geographic Code:1USA
Date:Mar 22, 2018
Words:4077
Previous Article:Appraisers in Arbitration: What Areas of Service Might Valuers Provide?
Next Article:Using the Income Approach for Minority Interests.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters