- Email: [email protected]
Regional logistics hubs, freight activity and industrial space demand: Econometric analysis
Research in Transportation Business & Management 11 (2014) 98–104
Contents lists available at ScienceDirect
Research in Transportation Business & Management
Regional logistics hubs, freight activity and industrial space demand: Econometric analysis Christopher Lindsey a, Hani S. Mahmassani a,⁎, Matt Mullarkey b, Terry Nash b, Steven Rothberg c a b c
Transportation Center, Northwestern University, 600 Foster Street, Evanston, IL 60208, United States CenterPoint Properties, 1808 Swift Drive, Oak Brook, IL 60523, United States Mercator International, LLC, 620 Kirkland Way, Suite 206, Kirkland, WA 98033, United States
a r t i c l e
i n f o
Article history: Received 4 November 2013 Received in revised form 24 May 2014 Accepted 6 June 2014 Available online 20 June 2014 Keywords: Logistics hubs Freight flows Industrial space Economic development
a b s t r a c t There has been a continuing interest among transportation researchers and the logistics industry in the relationship between the consumption of industrial space and freight transportation activity. With the growing importance of logistics and supply chain economics to global industries, firms organizing their industrial activities and locating their warehousing and operational centers must increasingly consider the availability, quality and cost of a range of transportation services. Accordingly, the development of logistics facilities in conjunction with regional freight transportation hubs has become an important element of the overall industrial economy, predicated on the notion that robust freight activity is a good indicator of the consumption of industrial space. In this study, we conduct an econometric analysis of a longitudinal data set consisting of twenty metropolitan markets observed annually from 1997 to 2007. From those results, we develop a methodology to score and rank metropolitan markets according to their potential for industrial space consumption based on macroeconomic, demographic, and freight flow variables. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction The relationship between regional consumption of industrial space and freight flows has been of increasing interest to urban geographers, transportation researchers, and logistics and supply chain managers. With growing importance of logistics and supply chain economics to global industries, firms organizing their supply chains (through the location of warehouses and operational centers) increasingly consider the availability, quality, and cost of a range of transportation services as major decision factors. This makes regional freight hubs attractive to developers of major logistics facilities, predicated on the notion that robust freight activity is a good indicator of demand for industrial space. This approach to the development of logistics facilities — i.e., concentrating on regional freight hubs — has become an important, but rarely studied, element of the overall industrial economy. Generally, the geography of freight and logistics distribution, and its associated locational dimensions, has seen limited systematic investigation and scholarly research (Hesse & Rodrigue, 2004). Motivated by the above considerations, this research explores the relationship between the consumption of industrial space and freight flows in U.S. metropolitan markets, and uses this relationship to help assess the relative potential demand for industrial space in these markets. ⁎ Corresponding author. Tel.: +1 847 491 7287. E-mail address: [email protected] (H.S. Mahmassani).
http://dx.doi.org/10.1016/j.rtbm.2014.06.002 2210-5395/© 2014 Elsevier Ltd. All rights reserved.
The interrelation between freight flows and industrial land consumption is considered in conjunction with the macroeconomic and demographic factors that influence consumption. Accordingly, the study objectives are to (a) identify determinants of industrial space consumption in metropolitan markets, and (b) develop an indicator of relative metropolitan market potential that would help predict likely development opportunities. This research contributes to transportation business and management by providing some insight into industrial space consumption using relatively simple indicators of economic activity within a rigorous statistical framework. It provides a methodology for assessing development opportunities provided by different markets as indicated by macroeconomic, demographic, and freight flow variables. The results of this work have implications for (a) management professionals formulating development investment strategies, (b) logistics planners considering location opportunities, and (c) planners in metropolitan and state agencies interested in predicting likely development patterns or in formulating economic development strategies in a spatially competitive environment. The remainder of the paper is organized as follows: First, some brief background information about the problem is given. Next, the research methodology is outlined along with a discussion of the data used for the study. This is followed by a discussion of the findings of the research. Lastly, the managerial implications and research contributions are discussed.
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
1.1. Background The theorized attractiveness of hub markets to firms making location decisions is firmly rooted in the field of regional science (Marshall, 1920) and transport geography (Bowen, 2008; Cidell, 2010; Sivitanidou, 1996). Marshall (1920) observed that the economic advantages gained from geographic proximity encourage firms to cluster together. Sheffi (2013) argued that logistics clusters are a unique type of cluster characterized by a collection of firms with logistics-intensive operations. Though they share many of the same characteristics that generally make industrial clusters attractive, logistics clusters particularly rely on the operational advantages gained from transportation and asset-sharing. Access to transportation infrastructure plays an influential role in the location decisions of businesses across a wide range of industries (Targa, Clifton, & Mahmassani, 2005, 2006). For instance, in an empirical analysis of Wal-Mart store locations, Holmes (2011) found that a dense network of stores allowed Wal-Mart to achieve significant distribution costs. However, for businesses in the industrial sectors, accessibility to goods' movement infrastructure is particularly important (Sivitanidou, 1996). Both Bowen (2008) and Cidell (2010) observed that the distribution of firms' logistics facilities is heavily correlated with accessibility. Lindsey, Mahmassani, Mullarkey, Nash, and Rothberg (2014) showed that there is a positive and statistically significant relationship between the demand for industrial space and freight flows. The results suggested that as freight flows increase, so too does demand. As centers of logistics activity and knowledge, hub markets allow firms to achieve lower transportation costs and greater supply chain flexibility (Nuzum, 2006). In hub markets, shippers have better opportunities to generate full truckload shipments (as opposed to less-thantruckload shipments and partial loads) and fewer empty miles, both of which allow carriers to offer better rates. Regarding supply chain flexibility, multimodal freight hubs allow customers' access to all of the major modes of freight transportation (rail, trucking, and air), which mitigates the risks associated with service disruptions in a single mode. The contribution of this work is a methodology that identifies the metropolitan markets that are most desirable for industrial and logistics real-estate investment. The basic idea is that those areas that are regional logistics hubs are prime investment markets. This is supported with a methodology that uses a few relatively simple and easily observed variables to develop a single indicator. Developers of logistics facilities desire indicators because they help to distinguish better investment opportunities from inferior ones. We develop our indicator using the parameter estimates of an econometric analysis of real estate, demographic, macroeconomic, and freight flow data. 2. Research questions, methods, & data This section discusses the specific research question we are examining, the methodology by which we will conduct this examination, and the data to be used in the study. 2.1. Research questions This study hypothesizes that the total consumption of industrial space in a metropolitan market can be formulated as a function of a few relatively simple demographic-, macroeconomic-, and freight flow-based barometers of consumption; this relationship can then be used to formulate a single indicator that investors can then use to help identify and compare prime investment markets. Industrial space consumption is measured by gross absorption, the total amount of consumed space in a market (NAIOP, 2005). Gross absorption can be considered a temporary equilibrium between supply and demand. It reflects the supply of land suitable for developing industrial and logistics facilities and the demand for those facilities. It is an appropriate measure for this research because it is
99
more useful from a developer's perspective than the year-to-year fluctuations in the difference between occupied and vacated space — net absorption (NAIOP, 2005). Gross absorption accurately reflects market size and can better inform investment strategies predicated on capturing market share of large and burgeoning markets; that is the goal of this research. 2.2. Data The study is conducted using data on the real estate, demographic, macroeconomic, and transportation characteristics of several metropolitan markets. The study markets are comprised of regionally and nationally significant metropolitan areas. They can be characterized as either having significant levels of freight activity, large consumer bases, or both. The markets include: ∙ Atlanta, GA ∙ Boston, MA ∙ Chicago, IL ∙ Cincinnati, OH ∙ Cleveland, OH ∙ Dallas, TX ∙ Detroit, MI ∙ Edison, NJ ∙ Houston, TX ∙ Indianapolis, IN
∙ Los Angeles, CA ∙ Minneapolis, MN ∙ New York, NY ∙ Oakland, CA ∙ Orange County, CA ∙ Philadelphia, PA ∙ Phoenix, AZ ∙ Riverside, CA ∙ St. Louis, MO ∙ Seattle, WA.
Many of the included markets are historical freight hubs at both the regional and national levels. Los Angeles and New York are gateway markets for much of the U.S.'s imports and exports. The ports of New York/New Jersey and Los Angeles/Long Beach are the nation's busiest in terms of container traffic. In 2012, they accounted for approximately 45% of U.S. waterborne container traffic (U.S. Army Corp of Engineers, 2014). Chicago, IL has long been considered the hub of the U.S. freight rail system. Six of the seven Class I railroads converge on the metropolitan area. Additionally, the availability of intermodal services has spurred in the region the development of large logistics clusters that offer to shippers other essential freight services such as warehousing and last-mile deliveries. Examples include the Burlington Northern Santa Fe and Union Pacific facilities located in Joliet and Rochelle, IL, respectively. Metropolitan areas such as Atlanta, Dallas and Seattle can be characterized as regional freight hubs, especially in regard to rail and intermodal services. Most have two Class I railroads operating within their areas and a few others have relatively large ports (e.g. Houston and Seattle). These markets are the economic and distribution centers of their respective regions. Smaller markets such as Riverside and Edison contain a predominant amount of the freight infrastructure used to service very large and populous regions (i.e. New York and Los Angeles). All of the markets, with the exceptions of Edison and Riverside, represent relatively large consumer bases as well. Thus, the selection of markets was driven a number of criteria. Because there was a desire to have broad geographic representation, metropolitan areas from across the U.S. were targeted. Markets that are widely considered freight hubs, especially in regard to rail services, were included in the sample alongside large, economically significant metropolitan areas. Also selected were regional economic and freight hubs. Additionally, markets with large amounts of freight activity relative to their size were included. However, the sample size is limited by the availability and quality of the data, especially considering that every market is observed several years. 2.2.1. Summary of the data The primary source of U.S. multimodal freight flow data is the Commodity Flow Survey (CFS). CFS data largely focuses on the most economically significant metropolitan areas in each state. Industrial
100
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
real estate data is primarily collected by private firms and tends to focus on those areas with significant manufacturing and distribution sector activity. Real estate data was acquired from CB Richard Ellis (CBRE). The geographic scale of the real estate data is synonymous, though not always equivalent, to metropolitan statistical areas. Macroeconomic and demographic data come from the U.S. Bureau of Economic Analysis (BEA) (BEA, 2011a and 2011b). Data for the freight flow variables was obtained from the Bureau of Transportation Statistics (BTS) which regularly publishes the Commodity Flow Survey (CFS) (BTS, 2000, 2004, and 2011). The CFS estimates the amount and value of commodities flowing through states and metropolitan areas. The most recent surveys are from the years 1997, 2002 and 2007. This study uses CFS data to estimate freight flows as measured by tons and ton-miles through the study areas included in the analysis. Data for years in between surveys (i.e. 1998–2001 and 2003–2006) were interpolated using simple growth rates. Additionally, for areas that are confounded with others, because the CFS aggregates data at the MSA and CBSA levels, freight flow is allocated among entities based on their percentage of the greater area's overall industrial stock. Overall industrial stock is included in the real estate data. Gross absorption is used as the dependent variable in this analysis because the magnitude of a metropolitan area's industrial space market is an important investment consideration. For instance, it may be more desirable to an investor to capture a small amount of market share in Chicago, IL as opposed to a large amount of market share in Indianapolis, IN. Gross absorption measures are directly derived from the overall industrial stock and vacancy rates included in the data set from CBRE. The definitions and summary statistics of the variables included in the analysis are given in Table 1. Simple scatter plots of the independent variables along the x-axis and the dependent variable, gross absorption, along the y-axis reveal some useful information (see Fig. 1). Firstly, Gross Absorption is strongly positively correlated with both freight flow measures, Total Tons and Total Ton-Miles, and distribution-sector employment. Secondly, there appears to be a robust relationship between gross absorption and Import–Export Growth as well as the presence of a maritime port in a metropolitan market (see Fig. 2). Lastly, in relation to many of the independent variables there are clearly two regimes of gross absorption with two markets — Los Angeles, CA and Chicago, IL — exhibiting levels much higher than all other markets. Though this is evident in the scatter plots of many of the independent variables, it is perhaps the most interesting in the case of Population Gain. There is already a significant nonlinear relationship between Population Gain and Gross Absorption when those two outlier markets are ignored (see Fig. 3). Their presence in the data set introduces an additional level of complexity due to a relationship with Gross Absorption that seemingly bucks the overall trend. Focusing on the relationship between Population Gain and Gross Absorption depicted in Fig. 3, we see that, to a point, greater gains in population negatively correlate with gross absorption. Past that threshold, Population Gain positively correlates with Gross Absorption. Los Angeles, CA and Chicago, IL seemingly have their own relationship to population
growth that is consistently negative. In the subsequent analysis, these areas are defined as “major markets” and are identified as such by an indicator variable. These insights will be important to developing model specifications that appropriately capture these relationships. 2.3. Methodology The methodology used in this study is a regression-based analysis of the hypothesized functional relationship described in the “Research questions” subsection. The econometric model that we use in the analysis is the linear model for longitudinal data (Wooldridge, 2002). A model for longitudinal data is employed because the data consists of a number of cross-sectional entities observed over time. Mathematically, the model is specified as
yit ¼ α þ βX it þ εit
ð1Þ
where i refers to the cross-sectional units (metropolitan markets), t refers to the time period (years), α is the intercept, β is (the transpose of) a vector of parameter estimates, and X is a vector of explanatory variables corresponding to each metropolitan market i at time t. By not indexing the β term, we have assumed parameter homogeneity. That is, a covariate has the same influence regardless of the metropolitan area. This reflects our interest in a model that can be used in markets that may not have been included in the initial estimation sample. The α term is not indexed either. It captures the collective effect of all markets included in the study. 3. Findings & discussion In this section of the paper, we formulate and test different model specifications, present the results, and draw some conclusions from the results of the analysis. 3.1. Model specifications The strategies driving our selection of covariates were to (1) test variables that intuitively and theoretically correlate with industrial space consumption and to (2) test variables across geographic levels. Variables based on employment-, population-, and economic-based measures are expected to strongly correlate with consumption. Regarding the geographic level at which variables are measured, it is important to consider that local markets are not insulated from their surroundings and the general economic environment. Economic conditions at local and national levels affect where a firm decides to locate its facilities and subsequently consume industrial space. Because of this, we include in our models variables across geographic levels. Two model specifications are included in the analysis. The models are essentially variations on a base model with the chosen freight flow
Table 1 Variable definitions. Variable
Description (units)
Mean
Std. dev.
Data source
Gross absorption Distribution employment Population gain Import-export growth Freight flow measures Total tons
Total occupied space (thousands of sq. ft.) Number of persons employed in the distribution sector (thousands) Number of persons gained or lost in a metropolitan market from the preceding year (thousands) Percent change in value of goods imported into and exported out of the U.S. from the preceding year (%)
337,743 133 54,812 5.005
171,533 73.9 51,685 6.05
CBRE CBRE BEA BEA
Tons of truck and rail freight originating in (i.e. freight movements that start in the study area) and destined for (i.e. freight movements that terminate in the study area) a metropolitan market (millions) Tons of truck and rail freight originating in (i.e. freight movements that start in the study area) and destined for (i.e. freight movements that terminate in the study area) a metropolitan market (millions)
255,685
144,468
CFS
56,299
30,959
CFS
Total ton-miles
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
101
Fig. 1. Scatter plots of gross absorption and the independent variables.
measure as the alternating variable. Eq. (2). is the base model and is specified as follows: Gross Absorptionit ¼ α þ β1 Distribution Employmentit þ β2 ImportExport Growtht þ β3 Population Gainit þ 2 β4 Major Market i Population Gain it þ β5 Major Market i þ β6 Port i þ β7 Freight Flowit þ ε it :
potentially significant. The generic variable Freight Flow alternates between Total Tons and Total Ton-Miles.
3.2. Model results and discussion ð2Þ
All variables, except Major Market and Port, are as defined in Table 1. Major Market is a dummy variable that indicates whether or not a metropolitan market exhibits gross absorption levels above 700,000 for the given time frame. Similarly, Port is a dummy variable that identifies markets that contain a maritime port. A market is considered to be a port market if a port is located within the geographic boundary of the study area. It is important to control for these two effects in relation to the other variables because they are
The results of the model specifications given in Eq. (2) are presented in Table 2. Both Models 1 and 2 exhibit a substantial overall statistical fit with R-Squared values of 0.729 and 0.695, respectively. From the results, we have some insight into the determinants of gross absorption as well as their magnitude. As expected, parameter estimates associated with the control variables — the port and Major Market dummy variables — are statistically significant. Also, in both models Import–Export Growth, a national-level economic measure, positively correlates with increased gross absorption. Among the most highly statistically significant results are those that most closely relate to the freight and logistics industry, freight flow (as measured by Total Tons
Fig. 2. Box plots of gross absorption and import–export growth and the port indicator.
102
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104 Table 3 Model elasticities. Variable
Distribution employment Import-export growth Population gain Major market indicator (non-major)∗population gain 2 Major market indicator (major)∗population gain 2 Freight flow measures Total tons Total ton-miles
Fig. 3. Scatter plot of gross absorption and population gain.
and Total Ton-Miles in Models 1 and 2, respectively) and Distribution Employment. In fact, distribution sector employment exhibits the highest statistical significance at a level far greater than 1% in both model specifications. Regarding Population Gain, the model specification captures its highly nonlinear and disjointed effect through the interaction of its squared value with the Major Market dummy variable. The results confirm what is observed in the scatter plot. Overall, Population Gain is negatively correlated with Gross Absorption. However, for non-major markets Population Gain is only negatively correlated to a point, after which the relationship reverses — increased population is associated with higher gross absorption levels. The likely reason for this observation is that small to moderate gains in population do not greatly benefit a region's industrial sector. In fact, they may hurt it as new residents increase population density creating for the industrial sector a more adverse business environment due to greater congestion, competition for land, and nuisance complaints. However, large gains indicate a thriving regional economy with more customers to serve and a larger labor pool from which to draw employees. The level of influence of the independent variables on Gross Absorption is also reflected in their elasticity measures (see Table 3). Using the sample means of the dependent and independent variables the elasticities are calculated. Distribution sector employment and the
Model 1
Model 2
Elasticity
Elasticity
0.688 0.00546 −0.0259 0.0110 −0.0228
0.728 0.00401 −0.0368 0.0122 −0.0246
– 0.144
0.108 –
freight flow measures are among the most influential factors in both model specifications. The results suggest that a 1% increase in distribution sector employment corresponds to an approximately 69% or 73% increase in gross absorption according to Models 1 and 2 specifications, respectively. Likewise, a 1% increase in the freight flow measures correspond to approximately 14% and 11% increases in gross absorption, respectively. Plots of the residuals, shown in Fig. 4, suggest that for both model specifications the regression model assumptions are not violated and that the results are sensible. The residuals do not exhibit any pattern that would suggest that there is some correlation in the error term that is unaccounted for in the model. Furthermore, the residuals help to identify which markets are outliers based on the analytical results. If the top and bottom 5th percentiles of the residuals are delineated, the data reveals that Atlanta, Chicago and Riverside are outliers according to both cut-off points when using Model 1 results. If Model 2 results are instead used, the same pattern emerges. These areas rank among the largest industrial real estate markets and each have significant amounts of freight infrastructure. A possible reason for their large residual values is that they all have experienced significant growth over the observation period relative to the other markets included in the sample. From these results we can conclude that distribution sector employment and freight flows, particularly Total Ton-Miles, are good indicators of gross absorption at the metropolitan level. In fact, those two variables alone explain much of the variation in the gross absorption data. This conclusion and these results form the foundation of the methodology that we propose for scoring and ranking potential investment markets. 3.3. Ranking metropolitan markets for investment A potential managerial application of the methodological framework outlined in this work is to use the analytical results to formulate an indicator that allows us to score and rank potential markets for
Table 2 Model results. Variable
Model 1
Model 2
Coefficient
t-Value
Coefficient
t-Value
Intercept Distribution employment Import–export growth Population gain Major market indicator (non-major)∗population gain 2 Major market indicator (major)∗population gain 2 Major market indicator (major) Port (port present)
6.26e + 04 1.75e + 03 3.69e + 02 −1.59e-01 6.94e-07 −9.21e-07 1.83e + 05 −5.702e + 04
2.68 11.9 2.85 −2.26 1.69 −1.64 3.31 −1.80
6.56e + 04 1.85e + 03 2.704e + 02 −2.27e-01 7.67e-07 −9.904e-07 1.79e + 05 −5.66e + 04
2.78 12.1 1.94 −3.00 1.73 −1.63 3.24 −1.81
Freight flow measures Total tons Total ton-miles
– 8.64e-01
– 8.20
1.42e-01 –
5.83 –
Model statistics N R-squared Adj. R-squared
0.729 0.699
220 0.695 0.666
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
103
Fig. 4. Plots of the residuals.
investment. Recall that the dependent variable in the regression analysis is gross absorption, the total amount of industrial space consumed in a metropolitan market at a given time. A score for each market could be formulated from the computation of the estimated version of Eq. (2) and taking the summation over a forecast horizon. In other words, if ̂ we take â and b to be the estimated parameters of α and β in Eq. (2), respectively, j = 1, …, J to be the forecast period, and continue to let i denote the metropolitan market, then J X ˆþ bˆ X ij : Scorei ¼ a
ð3Þ
j¼1
The score is basically an indicator that suggests which markets stand to make the greatest gains over the forecast horizon. Given the parameter estimates and reasonable forecasts of the macroeconomic, demographic, and freight flow variables in a potential investment market, the suggested indicator is easily computed. As an example of how the score could be used, consider three fictitious markets — A, B, and C. Using historical demographic, macroeconomic, and freight flow data for three actual markets we use simulation to create a five-year forecast for our fictitious markets (see Fig. 5). Based on that forecast, we then score each market over the forecast horizon using our metric. (Note that the point of this exercise is simply to demonstrate how the score may be used in an investment scenario using realistic data, not to create accurate forecasts.)
Using Eq. (3) and Model 1 results, the score for each market is calculated. Model 1 results were used because considering that we are interested in logistics hubs, Total Ton-Miles accounts for both the amount of goods shipped to a market as well as the distance traveled. In this manner, it is more indicative of a “hub” market than Total Tons. The score for each market is calculated to be 951,149, 2,292,170, and 1,249,364 for markets A, B, and C, respectively. Based on this criterion, market B is clearly the preferred market due to its dominant score. The score is made more intuitive by making it relative to the dominant market (i.e., the market with the highest score). This is achieved by dividing each individual market's score by the dominant market's score and multiplying by 100. Relative Scorei ¼
Scorei
. maxðScorei Þ
100:
ð4Þ
The relative score is calculated to be 42, 100, and 55 for markets A, B, and C, respectively. This score, along with other criteria, could then be used by managers to make comparisons between potential investment markets and to weigh their options. 4. Research implications Using rigorous econometric and statistical modeling techniques, this study has explored the relationship between freight flows, macroeconomic and demographic variables, and industrial space consumption. In doing so, it has made significant contributions to the freight and
Fig. 5. Simulated market forecasts.
104
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
logistics field and has revealed important managerial implications for firms that invest in and develop logistics hubs. These two aspects of the study will now be further discussed. 4.1. Implications for managerial practice The primary managerial implications of this work are the confirmation that investment strategies that target regional logistics hubs are well-founded and the development of a scoring and ranking methodology for metropolitan markets. The analytical results further confirm that freight hub markets, primarily as indicated by freight flows and distribution sector employment, represent better investment opportunities over other markets. In both model specifications, these are the most statistically significant variables and are positively correlated with gross absorption. Based on this result, we were able to develop our scoring methodology. The scoring methodology is simple, easily understood and applied, and only requires the acquisition of forecasts for a few macroeconomic, demographic, and freight flow variables. Though acquiring this forecast data may present a small challenge, it is worth the extra effort. By using the forecasts of the determinants, the score maintains the objectivity of the estimated relationships and their associated insights. This is not the case when using a time-series forecast or moving average of gross absorption alone. With the proposed indicator, managers have another tool to aid them in making market investment decisions. The authors do not suggest, and would not expect, managers to use this indicator alone when evaluating investment markets. Instead, managers would incorporate the indicator into their existing investment evaluation process. Ideally, our indicator captures a market dynamic not currently being properly considered — the logistics “hub” effect — and provides additional useful information. 4.2. Contribution to scholarly knowledge The primary scholarly contribution of the research is the empirical insight into the relationship between industrial space consumption and macroeconomic, demographic, and freight flow variables. An interesting relationship between gross absorption and changes in population was revealed. Overall, the relationship is negative with gains in population corresponding to decreases in gross absorption levels. However, for non-major markets the relationship reverses beyond a threshold after which increased population is associated with higher gross absorption levels. Major markets exhibit an altogether unique relationship that is consistently negative. Though our model is not causal, it does establish clear associations and demonstrates the practicality of the information derived from these relationships. In future work, a reformulation of our models as reduced form equations with supply-side proxies as right-hand side controls would greatly improve the analysis. It would be the first step toward revealing the underlying causality and its direction. In addition,
it would be useful to extend the present work to consider the underlying firm-level choice process that results in a business choosing to consume space in one location versus another. This would require observation of the location decisions made by many firms over a large geographic area. Lastly, the study could be improved with a more disaggregate measure of freight flows, though the Commodity Flow Survey is the only publicly available source in the U.S.
Acknowledgment The paper has benefited from the helpful comments and suggestions of the Editor and two anonymous referees.
References Bowen, J. T. (2008). Moving places: The geography of warehousing in the U.S. Journal of Transport Geography, 16(6), 379–387. Cidell, J. (2010). Concentration and decentralization: The new geography of freight distribution in the U.S. Journal of Transport Geography, 18(3), 363–371. Hesse, M., & Rodrigue, J. -P. (2004). The transport geography of logistics and freight distribution. Journal of Transport Geography, 12(3), 171–184. Holmes, T. J. (2011). The diffusion of Wal-Mart and economies of density. Econometrica, 79(1), 253–302. Lindsey, C., Mahmassani, H. S., Mullarkey, M., Nash, T., & Rothberg, S. (2014). Industrial space demand and freight activity: Exploring the connection. Journal of Transport Geography (Forthcoming). Marshall, A. (1920). Principles of economics. London: Macmillan. National Association of Industrial and Office Properties Research Foundation (NAIOP) (2005). NAIOP terms and definitions: U.S. Office and Industrial Market. Nuzum, P. (2006). Moving freight today — How shippers are creating greater capacity, reliability, and rate stability. Available from. http://www.prologis.com/docs/research/ supply_chain/Moving_Freight_-_August._2006.pdf (Downloaded September 16, 2013) Sheffi, Y. (2013). Logistics clusters: Delivering value and driving growth. Cambridge, MA: MIT Press. Sivitanidou, R. (1996). Warehouse and distribution facilities and community attributes: An empirical study. Environment and Planning A, 28(7), 1261–1278. Targa, F., Clifton, K., & Mahmassani, H. S. (2005). Economic activity and transportation access: An econometric analysis of business spatial patterns. Transportation Research Record, 1932, 61–71. Targa, F., Clifton, K., & Mahmassani, H. S. (2006). Influence of transportation access on individual firm location decisions. Transportation Research Record, 1977, 179–189. U.S. Army Corp of Engineers (2014). U.S. Waterborne Container Traffic by Port/Waterway in 2012. http://www.navigationdatacenter.us/wcsc/by_porttons12.html (Downloaded April 13, 2014) U.S. Bureau of Economic Analysis (BEA) (2011a). Table 1.1.1: Percent change from preceding period in real gross domestic product. Available from. http://www.bea.gov (Downloaded April 20, 2011) U.S. Bureau of Economic Analysis (BEA) (2011b). Table SA04: State income and employment summary. Available from. http://www.bea.gov (Downloaded November 14, 2011) U.S. Bureau of Transportation Statistics (BTS) (2000). Metropolitan Areas: 1997 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 14, 2011) U.S. Bureau of Transportation Statistics (BTS) (2004). Metropolitan Areas: 2002 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 13, 2011) U.S. Bureau of Transportation Statistics (BTS) (2011). Metropolitan Areas: 2007 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 13, 2011) Wooldridge, J. (2002). Econometric analysis of cross section and panel data. Cambridge: MIT Press.
Contents lists available at ScienceDirect
Research in Transportation Business & Management
Regional logistics hubs, freight activity and industrial space demand: Econometric analysis Christopher Lindsey a, Hani S. Mahmassani a,⁎, Matt Mullarkey b, Terry Nash b, Steven Rothberg c a b c
Transportation Center, Northwestern University, 600 Foster Street, Evanston, IL 60208, United States CenterPoint Properties, 1808 Swift Drive, Oak Brook, IL 60523, United States Mercator International, LLC, 620 Kirkland Way, Suite 206, Kirkland, WA 98033, United States
a r t i c l e
i n f o
Article history: Received 4 November 2013 Received in revised form 24 May 2014 Accepted 6 June 2014 Available online 20 June 2014 Keywords: Logistics hubs Freight flows Industrial space Economic development
a b s t r a c t There has been a continuing interest among transportation researchers and the logistics industry in the relationship between the consumption of industrial space and freight transportation activity. With the growing importance of logistics and supply chain economics to global industries, firms organizing their industrial activities and locating their warehousing and operational centers must increasingly consider the availability, quality and cost of a range of transportation services. Accordingly, the development of logistics facilities in conjunction with regional freight transportation hubs has become an important element of the overall industrial economy, predicated on the notion that robust freight activity is a good indicator of the consumption of industrial space. In this study, we conduct an econometric analysis of a longitudinal data set consisting of twenty metropolitan markets observed annually from 1997 to 2007. From those results, we develop a methodology to score and rank metropolitan markets according to their potential for industrial space consumption based on macroeconomic, demographic, and freight flow variables. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction The relationship between regional consumption of industrial space and freight flows has been of increasing interest to urban geographers, transportation researchers, and logistics and supply chain managers. With growing importance of logistics and supply chain economics to global industries, firms organizing their supply chains (through the location of warehouses and operational centers) increasingly consider the availability, quality, and cost of a range of transportation services as major decision factors. This makes regional freight hubs attractive to developers of major logistics facilities, predicated on the notion that robust freight activity is a good indicator of demand for industrial space. This approach to the development of logistics facilities — i.e., concentrating on regional freight hubs — has become an important, but rarely studied, element of the overall industrial economy. Generally, the geography of freight and logistics distribution, and its associated locational dimensions, has seen limited systematic investigation and scholarly research (Hesse & Rodrigue, 2004). Motivated by the above considerations, this research explores the relationship between the consumption of industrial space and freight flows in U.S. metropolitan markets, and uses this relationship to help assess the relative potential demand for industrial space in these markets. ⁎ Corresponding author. Tel.: +1 847 491 7287. E-mail address: [email protected] (H.S. Mahmassani).
http://dx.doi.org/10.1016/j.rtbm.2014.06.002 2210-5395/© 2014 Elsevier Ltd. All rights reserved.
The interrelation between freight flows and industrial land consumption is considered in conjunction with the macroeconomic and demographic factors that influence consumption. Accordingly, the study objectives are to (a) identify determinants of industrial space consumption in metropolitan markets, and (b) develop an indicator of relative metropolitan market potential that would help predict likely development opportunities. This research contributes to transportation business and management by providing some insight into industrial space consumption using relatively simple indicators of economic activity within a rigorous statistical framework. It provides a methodology for assessing development opportunities provided by different markets as indicated by macroeconomic, demographic, and freight flow variables. The results of this work have implications for (a) management professionals formulating development investment strategies, (b) logistics planners considering location opportunities, and (c) planners in metropolitan and state agencies interested in predicting likely development patterns or in formulating economic development strategies in a spatially competitive environment. The remainder of the paper is organized as follows: First, some brief background information about the problem is given. Next, the research methodology is outlined along with a discussion of the data used for the study. This is followed by a discussion of the findings of the research. Lastly, the managerial implications and research contributions are discussed.
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
1.1. Background The theorized attractiveness of hub markets to firms making location decisions is firmly rooted in the field of regional science (Marshall, 1920) and transport geography (Bowen, 2008; Cidell, 2010; Sivitanidou, 1996). Marshall (1920) observed that the economic advantages gained from geographic proximity encourage firms to cluster together. Sheffi (2013) argued that logistics clusters are a unique type of cluster characterized by a collection of firms with logistics-intensive operations. Though they share many of the same characteristics that generally make industrial clusters attractive, logistics clusters particularly rely on the operational advantages gained from transportation and asset-sharing. Access to transportation infrastructure plays an influential role in the location decisions of businesses across a wide range of industries (Targa, Clifton, & Mahmassani, 2005, 2006). For instance, in an empirical analysis of Wal-Mart store locations, Holmes (2011) found that a dense network of stores allowed Wal-Mart to achieve significant distribution costs. However, for businesses in the industrial sectors, accessibility to goods' movement infrastructure is particularly important (Sivitanidou, 1996). Both Bowen (2008) and Cidell (2010) observed that the distribution of firms' logistics facilities is heavily correlated with accessibility. Lindsey, Mahmassani, Mullarkey, Nash, and Rothberg (2014) showed that there is a positive and statistically significant relationship between the demand for industrial space and freight flows. The results suggested that as freight flows increase, so too does demand. As centers of logistics activity and knowledge, hub markets allow firms to achieve lower transportation costs and greater supply chain flexibility (Nuzum, 2006). In hub markets, shippers have better opportunities to generate full truckload shipments (as opposed to less-thantruckload shipments and partial loads) and fewer empty miles, both of which allow carriers to offer better rates. Regarding supply chain flexibility, multimodal freight hubs allow customers' access to all of the major modes of freight transportation (rail, trucking, and air), which mitigates the risks associated with service disruptions in a single mode. The contribution of this work is a methodology that identifies the metropolitan markets that are most desirable for industrial and logistics real-estate investment. The basic idea is that those areas that are regional logistics hubs are prime investment markets. This is supported with a methodology that uses a few relatively simple and easily observed variables to develop a single indicator. Developers of logistics facilities desire indicators because they help to distinguish better investment opportunities from inferior ones. We develop our indicator using the parameter estimates of an econometric analysis of real estate, demographic, macroeconomic, and freight flow data. 2. Research questions, methods, & data This section discusses the specific research question we are examining, the methodology by which we will conduct this examination, and the data to be used in the study. 2.1. Research questions This study hypothesizes that the total consumption of industrial space in a metropolitan market can be formulated as a function of a few relatively simple demographic-, macroeconomic-, and freight flow-based barometers of consumption; this relationship can then be used to formulate a single indicator that investors can then use to help identify and compare prime investment markets. Industrial space consumption is measured by gross absorption, the total amount of consumed space in a market (NAIOP, 2005). Gross absorption can be considered a temporary equilibrium between supply and demand. It reflects the supply of land suitable for developing industrial and logistics facilities and the demand for those facilities. It is an appropriate measure for this research because it is
99
more useful from a developer's perspective than the year-to-year fluctuations in the difference between occupied and vacated space — net absorption (NAIOP, 2005). Gross absorption accurately reflects market size and can better inform investment strategies predicated on capturing market share of large and burgeoning markets; that is the goal of this research. 2.2. Data The study is conducted using data on the real estate, demographic, macroeconomic, and transportation characteristics of several metropolitan markets. The study markets are comprised of regionally and nationally significant metropolitan areas. They can be characterized as either having significant levels of freight activity, large consumer bases, or both. The markets include: ∙ Atlanta, GA ∙ Boston, MA ∙ Chicago, IL ∙ Cincinnati, OH ∙ Cleveland, OH ∙ Dallas, TX ∙ Detroit, MI ∙ Edison, NJ ∙ Houston, TX ∙ Indianapolis, IN
∙ Los Angeles, CA ∙ Minneapolis, MN ∙ New York, NY ∙ Oakland, CA ∙ Orange County, CA ∙ Philadelphia, PA ∙ Phoenix, AZ ∙ Riverside, CA ∙ St. Louis, MO ∙ Seattle, WA.
Many of the included markets are historical freight hubs at both the regional and national levels. Los Angeles and New York are gateway markets for much of the U.S.'s imports and exports. The ports of New York/New Jersey and Los Angeles/Long Beach are the nation's busiest in terms of container traffic. In 2012, they accounted for approximately 45% of U.S. waterborne container traffic (U.S. Army Corp of Engineers, 2014). Chicago, IL has long been considered the hub of the U.S. freight rail system. Six of the seven Class I railroads converge on the metropolitan area. Additionally, the availability of intermodal services has spurred in the region the development of large logistics clusters that offer to shippers other essential freight services such as warehousing and last-mile deliveries. Examples include the Burlington Northern Santa Fe and Union Pacific facilities located in Joliet and Rochelle, IL, respectively. Metropolitan areas such as Atlanta, Dallas and Seattle can be characterized as regional freight hubs, especially in regard to rail and intermodal services. Most have two Class I railroads operating within their areas and a few others have relatively large ports (e.g. Houston and Seattle). These markets are the economic and distribution centers of their respective regions. Smaller markets such as Riverside and Edison contain a predominant amount of the freight infrastructure used to service very large and populous regions (i.e. New York and Los Angeles). All of the markets, with the exceptions of Edison and Riverside, represent relatively large consumer bases as well. Thus, the selection of markets was driven a number of criteria. Because there was a desire to have broad geographic representation, metropolitan areas from across the U.S. were targeted. Markets that are widely considered freight hubs, especially in regard to rail services, were included in the sample alongside large, economically significant metropolitan areas. Also selected were regional economic and freight hubs. Additionally, markets with large amounts of freight activity relative to their size were included. However, the sample size is limited by the availability and quality of the data, especially considering that every market is observed several years. 2.2.1. Summary of the data The primary source of U.S. multimodal freight flow data is the Commodity Flow Survey (CFS). CFS data largely focuses on the most economically significant metropolitan areas in each state. Industrial
100
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
real estate data is primarily collected by private firms and tends to focus on those areas with significant manufacturing and distribution sector activity. Real estate data was acquired from CB Richard Ellis (CBRE). The geographic scale of the real estate data is synonymous, though not always equivalent, to metropolitan statistical areas. Macroeconomic and demographic data come from the U.S. Bureau of Economic Analysis (BEA) (BEA, 2011a and 2011b). Data for the freight flow variables was obtained from the Bureau of Transportation Statistics (BTS) which regularly publishes the Commodity Flow Survey (CFS) (BTS, 2000, 2004, and 2011). The CFS estimates the amount and value of commodities flowing through states and metropolitan areas. The most recent surveys are from the years 1997, 2002 and 2007. This study uses CFS data to estimate freight flows as measured by tons and ton-miles through the study areas included in the analysis. Data for years in between surveys (i.e. 1998–2001 and 2003–2006) were interpolated using simple growth rates. Additionally, for areas that are confounded with others, because the CFS aggregates data at the MSA and CBSA levels, freight flow is allocated among entities based on their percentage of the greater area's overall industrial stock. Overall industrial stock is included in the real estate data. Gross absorption is used as the dependent variable in this analysis because the magnitude of a metropolitan area's industrial space market is an important investment consideration. For instance, it may be more desirable to an investor to capture a small amount of market share in Chicago, IL as opposed to a large amount of market share in Indianapolis, IN. Gross absorption measures are directly derived from the overall industrial stock and vacancy rates included in the data set from CBRE. The definitions and summary statistics of the variables included in the analysis are given in Table 1. Simple scatter plots of the independent variables along the x-axis and the dependent variable, gross absorption, along the y-axis reveal some useful information (see Fig. 1). Firstly, Gross Absorption is strongly positively correlated with both freight flow measures, Total Tons and Total Ton-Miles, and distribution-sector employment. Secondly, there appears to be a robust relationship between gross absorption and Import–Export Growth as well as the presence of a maritime port in a metropolitan market (see Fig. 2). Lastly, in relation to many of the independent variables there are clearly two regimes of gross absorption with two markets — Los Angeles, CA and Chicago, IL — exhibiting levels much higher than all other markets. Though this is evident in the scatter plots of many of the independent variables, it is perhaps the most interesting in the case of Population Gain. There is already a significant nonlinear relationship between Population Gain and Gross Absorption when those two outlier markets are ignored (see Fig. 3). Their presence in the data set introduces an additional level of complexity due to a relationship with Gross Absorption that seemingly bucks the overall trend. Focusing on the relationship between Population Gain and Gross Absorption depicted in Fig. 3, we see that, to a point, greater gains in population negatively correlate with gross absorption. Past that threshold, Population Gain positively correlates with Gross Absorption. Los Angeles, CA and Chicago, IL seemingly have their own relationship to population
growth that is consistently negative. In the subsequent analysis, these areas are defined as “major markets” and are identified as such by an indicator variable. These insights will be important to developing model specifications that appropriately capture these relationships. 2.3. Methodology The methodology used in this study is a regression-based analysis of the hypothesized functional relationship described in the “Research questions” subsection. The econometric model that we use in the analysis is the linear model for longitudinal data (Wooldridge, 2002). A model for longitudinal data is employed because the data consists of a number of cross-sectional entities observed over time. Mathematically, the model is specified as
yit ¼ α þ βX it þ εit
ð1Þ
where i refers to the cross-sectional units (metropolitan markets), t refers to the time period (years), α is the intercept, β is (the transpose of) a vector of parameter estimates, and X is a vector of explanatory variables corresponding to each metropolitan market i at time t. By not indexing the β term, we have assumed parameter homogeneity. That is, a covariate has the same influence regardless of the metropolitan area. This reflects our interest in a model that can be used in markets that may not have been included in the initial estimation sample. The α term is not indexed either. It captures the collective effect of all markets included in the study. 3. Findings & discussion In this section of the paper, we formulate and test different model specifications, present the results, and draw some conclusions from the results of the analysis. 3.1. Model specifications The strategies driving our selection of covariates were to (1) test variables that intuitively and theoretically correlate with industrial space consumption and to (2) test variables across geographic levels. Variables based on employment-, population-, and economic-based measures are expected to strongly correlate with consumption. Regarding the geographic level at which variables are measured, it is important to consider that local markets are not insulated from their surroundings and the general economic environment. Economic conditions at local and national levels affect where a firm decides to locate its facilities and subsequently consume industrial space. Because of this, we include in our models variables across geographic levels. Two model specifications are included in the analysis. The models are essentially variations on a base model with the chosen freight flow
Table 1 Variable definitions. Variable
Description (units)
Mean
Std. dev.
Data source
Gross absorption Distribution employment Population gain Import-export growth Freight flow measures Total tons
Total occupied space (thousands of sq. ft.) Number of persons employed in the distribution sector (thousands) Number of persons gained or lost in a metropolitan market from the preceding year (thousands) Percent change in value of goods imported into and exported out of the U.S. from the preceding year (%)
337,743 133 54,812 5.005
171,533 73.9 51,685 6.05
CBRE CBRE BEA BEA
Tons of truck and rail freight originating in (i.e. freight movements that start in the study area) and destined for (i.e. freight movements that terminate in the study area) a metropolitan market (millions) Tons of truck and rail freight originating in (i.e. freight movements that start in the study area) and destined for (i.e. freight movements that terminate in the study area) a metropolitan market (millions)
255,685
144,468
CFS
56,299
30,959
CFS
Total ton-miles
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
101
Fig. 1. Scatter plots of gross absorption and the independent variables.
measure as the alternating variable. Eq. (2). is the base model and is specified as follows: Gross Absorptionit ¼ α þ β1 Distribution Employmentit þ β2 ImportExport Growtht þ β3 Population Gainit þ 2 β4 Major Market i Population Gain it þ β5 Major Market i þ β6 Port i þ β7 Freight Flowit þ ε it :
potentially significant. The generic variable Freight Flow alternates between Total Tons and Total Ton-Miles.
3.2. Model results and discussion ð2Þ
All variables, except Major Market and Port, are as defined in Table 1. Major Market is a dummy variable that indicates whether or not a metropolitan market exhibits gross absorption levels above 700,000 for the given time frame. Similarly, Port is a dummy variable that identifies markets that contain a maritime port. A market is considered to be a port market if a port is located within the geographic boundary of the study area. It is important to control for these two effects in relation to the other variables because they are
The results of the model specifications given in Eq. (2) are presented in Table 2. Both Models 1 and 2 exhibit a substantial overall statistical fit with R-Squared values of 0.729 and 0.695, respectively. From the results, we have some insight into the determinants of gross absorption as well as their magnitude. As expected, parameter estimates associated with the control variables — the port and Major Market dummy variables — are statistically significant. Also, in both models Import–Export Growth, a national-level economic measure, positively correlates with increased gross absorption. Among the most highly statistically significant results are those that most closely relate to the freight and logistics industry, freight flow (as measured by Total Tons
Fig. 2. Box plots of gross absorption and import–export growth and the port indicator.
102
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104 Table 3 Model elasticities. Variable
Distribution employment Import-export growth Population gain Major market indicator (non-major)∗population gain 2 Major market indicator (major)∗population gain 2 Freight flow measures Total tons Total ton-miles
Fig. 3. Scatter plot of gross absorption and population gain.
and Total Ton-Miles in Models 1 and 2, respectively) and Distribution Employment. In fact, distribution sector employment exhibits the highest statistical significance at a level far greater than 1% in both model specifications. Regarding Population Gain, the model specification captures its highly nonlinear and disjointed effect through the interaction of its squared value with the Major Market dummy variable. The results confirm what is observed in the scatter plot. Overall, Population Gain is negatively correlated with Gross Absorption. However, for non-major markets Population Gain is only negatively correlated to a point, after which the relationship reverses — increased population is associated with higher gross absorption levels. The likely reason for this observation is that small to moderate gains in population do not greatly benefit a region's industrial sector. In fact, they may hurt it as new residents increase population density creating for the industrial sector a more adverse business environment due to greater congestion, competition for land, and nuisance complaints. However, large gains indicate a thriving regional economy with more customers to serve and a larger labor pool from which to draw employees. The level of influence of the independent variables on Gross Absorption is also reflected in their elasticity measures (see Table 3). Using the sample means of the dependent and independent variables the elasticities are calculated. Distribution sector employment and the
Model 1
Model 2
Elasticity
Elasticity
0.688 0.00546 −0.0259 0.0110 −0.0228
0.728 0.00401 −0.0368 0.0122 −0.0246
– 0.144
0.108 –
freight flow measures are among the most influential factors in both model specifications. The results suggest that a 1% increase in distribution sector employment corresponds to an approximately 69% or 73% increase in gross absorption according to Models 1 and 2 specifications, respectively. Likewise, a 1% increase in the freight flow measures correspond to approximately 14% and 11% increases in gross absorption, respectively. Plots of the residuals, shown in Fig. 4, suggest that for both model specifications the regression model assumptions are not violated and that the results are sensible. The residuals do not exhibit any pattern that would suggest that there is some correlation in the error term that is unaccounted for in the model. Furthermore, the residuals help to identify which markets are outliers based on the analytical results. If the top and bottom 5th percentiles of the residuals are delineated, the data reveals that Atlanta, Chicago and Riverside are outliers according to both cut-off points when using Model 1 results. If Model 2 results are instead used, the same pattern emerges. These areas rank among the largest industrial real estate markets and each have significant amounts of freight infrastructure. A possible reason for their large residual values is that they all have experienced significant growth over the observation period relative to the other markets included in the sample. From these results we can conclude that distribution sector employment and freight flows, particularly Total Ton-Miles, are good indicators of gross absorption at the metropolitan level. In fact, those two variables alone explain much of the variation in the gross absorption data. This conclusion and these results form the foundation of the methodology that we propose for scoring and ranking potential investment markets. 3.3. Ranking metropolitan markets for investment A potential managerial application of the methodological framework outlined in this work is to use the analytical results to formulate an indicator that allows us to score and rank potential markets for
Table 2 Model results. Variable
Model 1
Model 2
Coefficient
t-Value
Coefficient
t-Value
Intercept Distribution employment Import–export growth Population gain Major market indicator (non-major)∗population gain 2 Major market indicator (major)∗population gain 2 Major market indicator (major) Port (port present)
6.26e + 04 1.75e + 03 3.69e + 02 −1.59e-01 6.94e-07 −9.21e-07 1.83e + 05 −5.702e + 04
2.68 11.9 2.85 −2.26 1.69 −1.64 3.31 −1.80
6.56e + 04 1.85e + 03 2.704e + 02 −2.27e-01 7.67e-07 −9.904e-07 1.79e + 05 −5.66e + 04
2.78 12.1 1.94 −3.00 1.73 −1.63 3.24 −1.81
Freight flow measures Total tons Total ton-miles
– 8.64e-01
– 8.20
1.42e-01 –
5.83 –
Model statistics N R-squared Adj. R-squared
0.729 0.699
220 0.695 0.666
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
103
Fig. 4. Plots of the residuals.
investment. Recall that the dependent variable in the regression analysis is gross absorption, the total amount of industrial space consumed in a metropolitan market at a given time. A score for each market could be formulated from the computation of the estimated version of Eq. (2) and taking the summation over a forecast horizon. In other words, if ̂ we take â and b to be the estimated parameters of α and β in Eq. (2), respectively, j = 1, …, J to be the forecast period, and continue to let i denote the metropolitan market, then J X ˆþ bˆ X ij : Scorei ¼ a
ð3Þ
j¼1
The score is basically an indicator that suggests which markets stand to make the greatest gains over the forecast horizon. Given the parameter estimates and reasonable forecasts of the macroeconomic, demographic, and freight flow variables in a potential investment market, the suggested indicator is easily computed. As an example of how the score could be used, consider three fictitious markets — A, B, and C. Using historical demographic, macroeconomic, and freight flow data for three actual markets we use simulation to create a five-year forecast for our fictitious markets (see Fig. 5). Based on that forecast, we then score each market over the forecast horizon using our metric. (Note that the point of this exercise is simply to demonstrate how the score may be used in an investment scenario using realistic data, not to create accurate forecasts.)
Using Eq. (3) and Model 1 results, the score for each market is calculated. Model 1 results were used because considering that we are interested in logistics hubs, Total Ton-Miles accounts for both the amount of goods shipped to a market as well as the distance traveled. In this manner, it is more indicative of a “hub” market than Total Tons. The score for each market is calculated to be 951,149, 2,292,170, and 1,249,364 for markets A, B, and C, respectively. Based on this criterion, market B is clearly the preferred market due to its dominant score. The score is made more intuitive by making it relative to the dominant market (i.e., the market with the highest score). This is achieved by dividing each individual market's score by the dominant market's score and multiplying by 100. Relative Scorei ¼
Scorei
. maxðScorei Þ
100:
ð4Þ
The relative score is calculated to be 42, 100, and 55 for markets A, B, and C, respectively. This score, along with other criteria, could then be used by managers to make comparisons between potential investment markets and to weigh their options. 4. Research implications Using rigorous econometric and statistical modeling techniques, this study has explored the relationship between freight flows, macroeconomic and demographic variables, and industrial space consumption. In doing so, it has made significant contributions to the freight and
Fig. 5. Simulated market forecasts.
104
C. Lindsey et al. / Research in Transportation Business & Management 11 (2014) 98–104
logistics field and has revealed important managerial implications for firms that invest in and develop logistics hubs. These two aspects of the study will now be further discussed. 4.1. Implications for managerial practice The primary managerial implications of this work are the confirmation that investment strategies that target regional logistics hubs are well-founded and the development of a scoring and ranking methodology for metropolitan markets. The analytical results further confirm that freight hub markets, primarily as indicated by freight flows and distribution sector employment, represent better investment opportunities over other markets. In both model specifications, these are the most statistically significant variables and are positively correlated with gross absorption. Based on this result, we were able to develop our scoring methodology. The scoring methodology is simple, easily understood and applied, and only requires the acquisition of forecasts for a few macroeconomic, demographic, and freight flow variables. Though acquiring this forecast data may present a small challenge, it is worth the extra effort. By using the forecasts of the determinants, the score maintains the objectivity of the estimated relationships and their associated insights. This is not the case when using a time-series forecast or moving average of gross absorption alone. With the proposed indicator, managers have another tool to aid them in making market investment decisions. The authors do not suggest, and would not expect, managers to use this indicator alone when evaluating investment markets. Instead, managers would incorporate the indicator into their existing investment evaluation process. Ideally, our indicator captures a market dynamic not currently being properly considered — the logistics “hub” effect — and provides additional useful information. 4.2. Contribution to scholarly knowledge The primary scholarly contribution of the research is the empirical insight into the relationship between industrial space consumption and macroeconomic, demographic, and freight flow variables. An interesting relationship between gross absorption and changes in population was revealed. Overall, the relationship is negative with gains in population corresponding to decreases in gross absorption levels. However, for non-major markets the relationship reverses beyond a threshold after which increased population is associated with higher gross absorption levels. Major markets exhibit an altogether unique relationship that is consistently negative. Though our model is not causal, it does establish clear associations and demonstrates the practicality of the information derived from these relationships. In future work, a reformulation of our models as reduced form equations with supply-side proxies as right-hand side controls would greatly improve the analysis. It would be the first step toward revealing the underlying causality and its direction. In addition,
it would be useful to extend the present work to consider the underlying firm-level choice process that results in a business choosing to consume space in one location versus another. This would require observation of the location decisions made by many firms over a large geographic area. Lastly, the study could be improved with a more disaggregate measure of freight flows, though the Commodity Flow Survey is the only publicly available source in the U.S.
Acknowledgment The paper has benefited from the helpful comments and suggestions of the Editor and two anonymous referees.
References Bowen, J. T. (2008). Moving places: The geography of warehousing in the U.S. Journal of Transport Geography, 16(6), 379–387. Cidell, J. (2010). Concentration and decentralization: The new geography of freight distribution in the U.S. Journal of Transport Geography, 18(3), 363–371. Hesse, M., & Rodrigue, J. -P. (2004). The transport geography of logistics and freight distribution. Journal of Transport Geography, 12(3), 171–184. Holmes, T. J. (2011). The diffusion of Wal-Mart and economies of density. Econometrica, 79(1), 253–302. Lindsey, C., Mahmassani, H. S., Mullarkey, M., Nash, T., & Rothberg, S. (2014). Industrial space demand and freight activity: Exploring the connection. Journal of Transport Geography (Forthcoming). Marshall, A. (1920). Principles of economics. London: Macmillan. National Association of Industrial and Office Properties Research Foundation (NAIOP) (2005). NAIOP terms and definitions: U.S. Office and Industrial Market. Nuzum, P. (2006). Moving freight today — How shippers are creating greater capacity, reliability, and rate stability. Available from. http://www.prologis.com/docs/research/ supply_chain/Moving_Freight_-_August._2006.pdf (Downloaded September 16, 2013) Sheffi, Y. (2013). Logistics clusters: Delivering value and driving growth. Cambridge, MA: MIT Press. Sivitanidou, R. (1996). Warehouse and distribution facilities and community attributes: An empirical study. Environment and Planning A, 28(7), 1261–1278. Targa, F., Clifton, K., & Mahmassani, H. S. (2005). Economic activity and transportation access: An econometric analysis of business spatial patterns. Transportation Research Record, 1932, 61–71. Targa, F., Clifton, K., & Mahmassani, H. S. (2006). Influence of transportation access on individual firm location decisions. Transportation Research Record, 1977, 179–189. U.S. Army Corp of Engineers (2014). U.S. Waterborne Container Traffic by Port/Waterway in 2012. http://www.navigationdatacenter.us/wcsc/by_porttons12.html (Downloaded April 13, 2014) U.S. Bureau of Economic Analysis (BEA) (2011a). Table 1.1.1: Percent change from preceding period in real gross domestic product. Available from. http://www.bea.gov (Downloaded April 20, 2011) U.S. Bureau of Economic Analysis (BEA) (2011b). Table SA04: State income and employment summary. Available from. http://www.bea.gov (Downloaded November 14, 2011) U.S. Bureau of Transportation Statistics (BTS) (2000). Metropolitan Areas: 1997 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 14, 2011) U.S. Bureau of Transportation Statistics (BTS) (2004). Metropolitan Areas: 2002 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 13, 2011) U.S. Bureau of Transportation Statistics (BTS) (2011). Metropolitan Areas: 2007 Commodity Flow Survey. Available from. http://www.bts.gov/publications/commodity_flow_ survey/ (Downloaded October 13, 2011) Wooldridge, J. (2002). Econometric analysis of cross section and panel data. Cambridge: MIT Press.