i
Linking a San Pedro Bay Forecast to Regional Labor Markets
Final Report
Metrans Project 09- 23
January 2010
Kristen Monaco
Guy Yamashiro
and
SungMinh Shin
Department of Economics
California State University Long Beach
Long Beach, CA 90840
ii
DISCLAIMER
The contents of this report reflect the views of the authors, who are responsible for the
facts and the accuracy of the information presented herein. This document is
disseminated under the sponsorship of the Department of Transportation, University
Transportation Centers Program, and California Department of Transportation in the
interest of information exchange. The U. S. Government and California Department of
Transportation assume no liability for the contents or use thereof. The contents do not
necessarily reflect the official views or policies of the State of California or the
Department of Transportation. This report does not constitute a standard, specification,
or regulation.
iii
ABSTRACT
Using time series forecasting techniques, we develop multiple forecasts of inbound
container traffic through the San Pedro Bay Ports. These forecasts, combined with panel
data labor market models, allow us to forecast the impact of declining freight volumes on
total employment, transportation employment, and transportation payroll in Los Angeles
and Orange Counties and the Inland Empire. We find that the decline in traffic is
associated with a decline of nearly 330,000 jobs in 2009 and 147,000 jobs in 2010 in the
4- county region. Transportation employment is estimated to have declined by nearly
14,000 jobs in 2009 due to declining port activity and forecast to decline by another
5,000 in 2010.
iv
TABLE OF CONTENTS
1. Introduction 1
2. Trends in Trade through the San Pedro Bay Ports 2
3. Port Forecast 5
4. Linking Regional Employment to Inbound Port Traffic 11
5. Conclusions and Recommendations 20
6. Implementation 21
7. References 22
v
LIST OF TABLES AND FIGURES
Figure 1: Trends in Loaded Inbound Containers at the San Pedro Bay Ports
Figure 2: Theory Forecast
Figure 3: Correlation Forecast
Figure 4: Control Forecast
Table 1: Trends in Imports by HTC; Top 15 by Weight
Table 2: Trends in Imports by HTC; Top 15 by Value
Table 3: Forecast Models
Table 4: Forecast Model Diagnostics
Table 5a: Trends in Employment and Payroll, Los Angeles County
Table 5b: Trends in Employment and Payroll, Orange County
Table 6b: Trends in Employment and Payroll, Inland Empire
Table 6: Random Effects Estimation Results, Total County Employment
Table 7: Panel Estimation Results, County Transportation Employment
Table 8: Random Effects Estimation Results, County Transportation Payroll
Table 9: Forecast Container Counts
Table 10: Forecasted Impact on 4- County Labor Market
Table 11: Forecasts of Labor Market Outcomes by County
vi
DISCLOSURE
This project was funded in entirety under this contract to California Department of
Transportation.
1
1. INTRODUCTION
The volume of trade entering the U. S. through the Ports of Long Beach and Los
Angeles ( the San Pedro Bay ports) has risen considerably over the past two decades. The
Ports are seen as engines of economic growth, as jobs related to goods movement have
increased considerably in the Southern California region. This rise in trade- dependent
jobs, concurrent with the decline in manufacturing employment, has led to increasing
interest in facilitating goods movement in the region, through investment in public and
private infrastructure.
The importance of goods movement to the region is typically measured
quantitatively through input- output ( I- O) modeling, which uses established multipliers to
estimate the impacts of port activity on the local, regional, and national economy. For
example, a 2007 study by BST Associates attributes 3 million jobs to port activity in
2005 ( BST, 2007).
The limitations of I- O studies is that they use multipliers established in prior
periods ( which can be quite dated) and lack visible links of the channels by which port
activity results in jobs. In this study we employ a different quantitative approach to
linking port activity to the regional labor market. First, we develop and evaluate
forecasts of inbound port traffic using three techniques of time series econometrics.
Second, we use panel data estimation techniques to model local labor markets, to
generate an estimate of the impact of changes in inbound container volumes on
employment and payroll. Finally, we combine the forecast of port activity with the labor
market estimation results to predict the short- term job market impacts of the decline in
imports on the regional economy.
2
The advantage of this approach is that the data is readily available and the
estimations provide results that clearly estimate the linkages between port activity and the
regional labor market outcomes ( as opposed to the " black box" approach of I- O models).
These models are easily updated as more data becomes available and are easily replicated
( all data and programs are available from the authors upon request).
2. TRENDS IN TRADE THROUGH THE SAN PEDRO BAY PORTS
The economic slowdown has caused substantial declines in U. S. imports. Figure
1 shows the trends in loaded inbound containers into the Ports of Los Angeles and Long
Beach through the second quarter of 2009.
Figure 1: Trends in Loaded Inbound Containers at the San Pedro Bay Ports
3
While is clear from figure 1 that loaded imports ( which generate the most economic
activity in the region) decreased substantially at the end of 2007, it is not clear which
imports were most affected and whether some of this decline may have been caused by
diversion to other U. S. ports.
Using data from the U. S. Census we examine the changes in the largest imports
( by weight and by value) by type of commodity ( classified by harmonized tariff code).
Tables 1 presents the 2008 weight of the top 15 import classifications through the Ports
of Los Angeles and Long Beach, as well as the percentage change from 2006 to 2008,
and the percentage change over the same period for imports into all U. S. ports. Table 2
presents the same data by value of imports ( measured in 2008 dollars).
Table 1: Trends in Imports by HTC; Top 15 by Weight
Commodity San Pedro Bay 2008
SWT Imports
Percentage
Change
( 2006‐
2008)
US
Percentage
Change
( 2006‐
2008)
27 Mineral Fuel, Oil Etc. 19,356,645,954 ‐ 11.02% ‐ 8.00%
84 Nuclear Reactors, Boilers, Machinery Etc.;
Parts
4,368,019,988 ‐ 8.65% ‐ 9.66%
94 Furniture; Bedding Etc; 3,996,629,057 ‐ 21.31% ‐ 12.22%
85 Electric Machinery Etc; 3,858,134,008 ‐ 5.68% ‐ 2.71%
73 Articles Of Iron Or Steel 3,384,478,722 ‐ 14.17% 10.40%
87 Vehicles, Except Railway Or Tramway, And
Parts Etc
3,032,612,815 ‐ 8.91% ‐ 11.91%
72 Iron And Steel 2,679,606,532 ‐ 39.88% ‐ 42.25%
39 Plastics And Articles Thereof 2,476,092,713 ‐ 15.60% ‐ 10.25%
95 Toys, Games & Sport Equipment; 1,876,229,693 ‐ 9.75% ‐ 6.94%
68 Art Of Stone, Plaster, Cement, Asbestos,
Mica Etc.
1,645,476,461 ‐ 8.49% ‐ 17.38%
40 Rubber And Articles Thereof 1,635,684,953 ‐ 3.75% ‐ 2.27%
25 Salt; Sulfur; Earth & Stone; Lime & Cement
Plaster
1,392,029,787 ‐ 72.16% ‐ 28.50%
48 Paper & Paperboard & Articles 1,314,180,426 ‐ 7.70% ‐ 15.80%
69 Ceramic Products 1,219,027,178 ‐ 26.70% ‐ 33.94%
4
44 Wood And Articles Of Wood 1,105,042,724 ‐ 32.99% ‐ 50.73%
Table 2: Trends in Imports by HTC; Top 15 by Value
Commodity San Pedro Bay Value
of Imports, 2008
Percentage
Change
( 2006‐ 8)
US
Percentage
Change
( 2006‐ 8)
85 Electric Machinery Etc; $ 46,923,836,355 8.12% 11.04%
84 Nuclear Reactors, Boilers, Machinery Etc.; $ 43,740,013,011 ‐ 11.84% ‐ 4.42%
87 Vehicles And Parts Etc $ 24,486,108,688 ‐ 17.03% ‐ 12.01%
95 Toys, Games & Sport Equipment $ 13,709,854,928 11.27% 23.90%
27 Mineral Fuel, Oil Etc.; Bitumin Subst;
Mineral Wax
$ 13,140,865,790 23.20% 35.69%
61 Apparel Articles And Accessories, Knit Or
Crochet
$ 12,903,616,878 12.17% 4.80%
94 Furniture; Bedding Etc; Lamps $ 11,688,172,007 ‐ 16.58% ‐ 7.79%
62 Apparel Articles And Accessories, Not Knit
Etc.
$ 11,561,048,474 ‐ 6.90% ‐ 7.52%
64 Footwear, Gaiters Etc. And Parts Thereof $ 8,980,625,285 ‐ 4.19% ‐ 3.40%
73 Articles Of Iron Or Steel $ 7,526,858,174 7.77% 33.70%
39 Plastics And Articles Thereof $ 7,318,969,640 ‐ 4.15% 1.29%
40 Rubber And Articles Thereof $ 5,811,993,814 5.61% 12.78%
90 Optic, Photo Etc, Medic Or Surgical
Instrments Etc
$ 4,725,283,931 7.75% 13.98%
42 Leather Art; Saddlery Etc; Handbags Etc;
Gut Art
$ 4,050,243,382 ‐ 11.28% 1.01%
29 Organic Chemicals $ 3,662,003,370 65.84% 24.81%
From Table 1 it is clear that the SPB ports saw a drop in total weight shipped for each of
the top 15 import categories over the 2006- 8 period. What is notable is that some
commodities, such as articles of iron or steel, increased for the U. S. as a whole over the
period, indicating that shipments were being diverted from the SPB ports to other U. S.
ports. A similar trend is seen with iron and steel and salt, sulfur and stone, where the
decline in volume through the SPB ports far exceeds the decline for all U. S. ports.
Trends in freight values shipped through the two ports allows a comparison of
freight that incorporates both weight and price. Vehicles, a major high- value import for
5
the SPB ports, declined substantially between 2006 and 2008, mirrored in the declines for
the US as a whole, indicating that most of the decline was due to economic reasons, not
due to vehicles being imported through other U. S. ports. The value of toys, mineral fuel
and oil, rubber, and apparel increased through the SPB ports, but increased more for the
U. S. as a whole, suggesting diversion of this freight to other U. S. ports. The SPB ports
did not lag behind US ports in all of the top commodities; the value of organic chemicals
imported through SPB increased 66%, substantially higher than the increase in the U. S.
as a whole. A similar trend is seen for knit apparel.
Given the substantial changes in imported goods over the 2006- 8 period, it is clear
that a forecast of imports would be a useful tool. The goal is to construct and evaluate a
number of forecasts to determine if there is a model that can reliably forecast port traffic
with a parsimonious specification, making it easy to update and replicate.
3. PORT FORECAST
3A. Models of Inbound Loaded Containers
To forecast quarterly1 inbound container traffic to the San Pedro Bay Ports, we
consider three different models. The first is developed from international trade theory, the
second is correlation- based, and the third is a control model. The theory model generates
forecasts of loaded inbound container traffic, while the correlation and control models
generate forecasts of the percentage change in of loaded inbound container traffic. 2
1 We focus on quarterly projections, as opposed to monthly, primarily because of data availability.
2 We use percentages ( log changes) because of strong evidence that the loaded inbound series displays a
unit root ( as well as the GDP and exchange rate data). It is not necessary to correct for this in the VECM as
this is accounted for in the estimation procedure.
6
The theory model is based on a standard imperfect substitutes trade model. 3 The
model is given by:
( 1)
Where imp is real imports, y is real income, and r is the real value of the dollar. Of
course, in our case we are interested in port activity ( in), not imports, so we simply swap
out imp for in. Additionally, we can convert this contemporaneous model to a predictive
model by simply lagging the variables on the right- hand- side. Allowing there to be a long
run relation between the levels of port activity, the value of the dollar ( measured by the
real effective exchange rate), and income results in the baseline model:
( 2)
This baseline model can be expanded by including additional lagged changes ( we
include two lags in our final model), or, more interestingly additional variables. Other
potential important variables, in terms of forecasting, include U. S. real household net
worth, changes in business inventories ( motivated by the view that big import changes
are associated with inventory investment or disinvestment due to just- in- time supply
management), or U. S. credit standards ( in an attempt to capture the effects of trade
financing on imports). We tried various combinations of these variables, but in terms of
performance measures such as information criteria, correlogram analysis, root mean
square forecast error, etc, the baseline model performed best.
As an alternative, we also construct a model based on correlations. The goal with
this model was to capture the correlations between world trade and port activity.
3 For more information on imperfect substitutes trade models, see Krugman and Obstfeld ( 2008)
7
Specifically, we assume that the historical relationship between world trade and San
Pedro Bay port traffic can predict future inbound traffic. The model is then:
( 3)
Where in is port activity and imp is actual and OECD forecasted U. S. imports. Other
variables we considered ( in addition and in isolation) included forecasts of U. S. real
GDP, G8 imports, and G8 economic activity. As with the first model, however, the most
parsimonious model proved best.
The final, control model, is a univariate model with no basis in economic theory.
The model was constructed using Box- Jenkins- type methodology. The model is specified
as follows:
( 4)
The three forecast models are summarized below: 4
Table 3: Forecast Models
Model Type Variables Estimation Procedure
Theory Loaded inbound, US real
exchange rate, US real GDP
Vector error correction
( VEC)
Correlation Percent change in loaded
inbound, OECD forecast of
US imports
Ordinary least squares
( OLS)
Control Percent change in loaded
inbound
Ordinary least squares
( OLS)
4 All models also contain a constant, quarterly dummy variables, a 2002: 4 dummy ( lockout), and a 2004: 3
dummy ( congestion). All variables, with the exception of the dummy variables, are logged. Lag lengths are
chosen using Schwartz Information Criterion.
8
The data begin in the first quarter of 1995 ( 1995: 1) and run through the second
quarter of 2009 ( 2009: 2). Forecasts are generated through the second quarter of 2010.
The data for the forecasting models are from several sources. Loaded inbound containers
for the Ports of Los Angeles and Long Beach are obtained by the websites of the two
ports, OECD forecasts are obtained from the OECD Economic Outlook No. 85, U. S. real
GDP data is from the FRED ® ( Federal Reserve Economic Data) database
( http:// research. stlouisfed. org/ fred2/), and the U. S. real effective exchange rate is from
the Bank for International Settlements ( http:// www. bis. org/ statistics/ eer/ index. htm)
3B. Evaluating the forecasts
The primary method of forecast evaluation within each of the three categories of
models was pseudo- out- of- sample forecasting. The models were estimated using the
1995: 1- 2008: 2 sample, with forecasts generated for 2008: 3- 2010: 4. Using the actual
realizations and comparing them to the forecasts we were able to calculate the root mean
square forecast errors of the various models. As mentioned above, we supplemented this
“ out- of- sample” procedure with “ in- sample” procedures such as comparison of
information criteria and correlogram analysis ( where appropriate). While the final
models were estimated in both levels and first differences, the forecasts below are the
results from the differenced models ( which were more accurate) transformed into levels
( to make the charts more accessible). The standard errors of the forecast are shown in
dashed lines.
9
Figure 2: Theory Forecast
Figure 3: Correlation Forecast
10
Figure 4: Control Forecast
Diagnostics for the three models are presented below.
Table 4: Forecast Model Diagnostics
Model Type Root Mean Squared Error Adjusted R- squared
Theory 0.1931 0.7262
Correlation 0.0948 0.4127
Control 0.2268 0.6991
Comparing across the three models, the Control/ Box- Jenkins- type models seem to
perform better in the pseudo- out- of- sample forecasting exercises. The argument can be
made that they should perform better because they are parsimonious and simple ( in terms
11
of estimation). On the other hand, the Theory/ VECM looks to be more accurate when it
comes to the 2009- 10 forecasts.
Based on the RMSE, the correlation model appears to be the best estimator,
followed by the theory model and the control model. The adjusted R- squared can only be
used to compare the correlation and control models ( as the theory model has a different
dependent variable) and, contrary to the RMSE, implies the control model outperforms
the correlation model.
Clearly, none of the models really predicted the severity of the fall in port activity
in 2008. We do not find this surprising. No forecasting models that we are aware of
matched the decline in imports over the past year. As the fall in traffic over the 2008- 9
period was unprecedented in the data, it is difficult to accurately forecast such a decline.
As more periods of increasing and decreasing activity will be evidenced in the coming
years, a continued effort to generate quarterly forecasts should lead to more reliable
forecasts in the future.
4. LINKING REGIONAL EMPLOYMENT TO INBOUND PORT
TRAFFIC
Both Ports boast of the number of local and regional jobs linked to port activity.
In this section, we use panel regression analysis to measure the link between loaded
imports and regional employment, transportation employment, and transportation payroll.
All data run from 1995- 2008. The counties included are Los Angeles, Orange, and
Riverside and San Bernardino. The latter two are combined into an Inland Empire
designation due to some data limitations in gathering data on the two counties separately.
12
Trends in total employment, employment in the transportation industry, and
transportation payroll are presented in Tables 5a- 5c.
Table 5a: Trends in Employment and Payroll, Los Angeles County
year Total
Employment
Transportation
Employment
Transportation
Payroll ( in
millions)
1990 4,149,500 143,200 3873
1991 3,992,600 142,700 4197
1992 3,813,600 136,800 4342
1993 3,716,800 134,000 4357
1994 3,710,400 134,900 4477
1995 3,754,500 139,500 4646
1996 3,795,700 142,400 4763
1997 3,872,000 147,300 5190
1998 3,951,200 154,200 5455
1999 4,010,200 159,300 5763
2000 4,079,800 162,200 6259
2001 4,082,000 163,500 6369
2002 4,034,600 155,400 6233
2003 3,990,800 149,200 6140
2004 4,004,100 148,500 6271
2005 4,031,600 149,100 6385
2006 4,100,100 152,300 6788
2007 4,129,600 152,300 6951
2008 4,076,200 148,500 6764
Table 5b: Trends in Employment and Payroll, Orange County
year Total
Employment
Transportation
Employment
Transportation
Payroll ( in
millions)
1990 1,179,000 21,200 455
1991 1,150,800 23,000 459
1992 1,133,200 23,200 580
1993 1,122,700 24,700 581
13
1994 1,133,800 27,500 624
1995 1,158,000 27,800 673
1996 1,191,000 27,100 669
1997 1,240,700 27,100 715
1998 1,305,700 25,900 756
1999 1,352,200 26,200 823
2000 1,396,500 26,900 882
2001 1,420,800 27,000 853
2002 1,411,000 25,100 849
2003 1,436,200 25,500 889
2004 1,463,400 25,700 997
2005 1,496,500 25,200 986
2006 1,524,300 24,700 1030
2007 1,520,500 25,100 1137
2008 1,489,300 25,400 1129
Table 5c: Trends in Employment and Payroll, Riverside and San Bernardino Counties
year Total
Employment
Transportation
Employment
Transportation
Payroll ( in
millions)
1990 735,100 24,300 426
1991 741,600 27,100 492
1992 751,500 27,900 543
1993 755,800 30,400 609
1994 772,800 32,700 673
1995 801,700 35,900 736
1996 824,800 36,100 792
1997 863,200 37,800 837
1998 903,800 42,000 960
1999 960,300 44,800 1123
2000 1,010,100 46,300 1196
2001 1,050,700 45,700 1191
2002 1,084,800 46,800 1221
2003 1,119,500 50,100 1309
2004 1,178,700 55,500 1721
2005 1,240,200 60,200 1880
2006 1,285,000 63,800 2077
2007 1,285,500 66,800 2290
2008 1,222,508 64,450 2364
14
While transportation employment remained relatively stable in LA County and grew
moderately in Orange County, the Inland Empire experienced substantial growth in
transportation employment over the period ( outstripping the general growth in
employment in this region). The development of the Inland Empire as a region that
supports considerable transportation employment is linked to its location close to the
ports and to rail routes that leave the Southern California area, making it a desirable area
for warehouses and distribution centers handling international freight entering the SPB
ports and destined for areas outside of the Southern California Region.
To formally model the link between port traffic and employment and payroll in
the four county region, we develop three models: total county employment, county
transportation employment, and county transportation payroll. As the data spans counties
and time, we use panel data estimation techniques for these models. All models are
specified in log- log functional form, so the coefficients can be expressed as elasticities.
4A. Total Employment Model
The model of total employment has the lag of county employment, California
total employment, education, inbound containers, county unemployment rate and time as
explanatory variables. Using the lag of the dependent variable as an explanatory variable
is sensible as employment in one period tends to be most dependent on the employment
level in the prior period ( thus we expect a positive sign on this coefficient). We also
expect positive coefficients on California total employment and education ( measured as
the percent of adults with a high school education). We expect a negative coefficient on
the unemployment variable, as employment and unemployment are inversely related by
15
definition. For the purpose of this study, our focus is on the sign and significance of the
coefficient on loaded inbound containers. We expect the sign of the coefficient to be
positive ( more port activity should increase regional employment). Table 6 presents the
estimation results. The choice of county fixed effects versus random effects was based
on the results of a Hausman test, which indicated that random effects is the correct
specification.
Table 6: Random Effects Estimation Results, Total County Employment
Coef. Std. Err. z P>| z|
lagged total
employment
0.968931 0.004013 241.46 0.00
California Total
Employment
- 0.04365 0.127726 - 0.34 0.73
Education - 0.19891 0.047086 - 4.22 0.00
Inbound Containers 0.164946 0.054964 3.00 0.00
Unemployment Rate - 0.00516 0.001692 - 3.05 0.00
Time - 0.01694 0.004292 - 3.95 0.00
Overall R- squared 0.9997
Wald Chi- squared 119324.3
P- value of Wald 0
The coefficient on inbound containers is positive and significant, as expected.
The magnitude of the coefficient suggests that a one percent increase in loaded inbound
containers through the SPB ports will increase county- level employment by 0.16% in the
four county area.
4B. Transportation Employment Model
16
We next measure the impact of port traffic on employment in the transportation
industry, which should be the industry with the most direct dependence on port activity.
The specification of the model is largely the same as that of total employment, however
the dependent variable is county transportation employment and California transportation
employment is used as an explanatory variable, replacing California total employment
used in the prior model. The Hausman test rejects random effects at the 5% level, but not
at the 10% level, so both the random effects and fixed effects estimation results are
presented in Table 7. As expected, loaded inbound containers have a larger impact on
transportation employment than in the prior model of total employment, but the
coefficient is only statistically significant in the random effects model. The coefficient
on loaded inbound containers in the random effects model suggests that a 1% increase in
loaded inbound containers through the SPB ports will increase county- level
transportation employment by 0.31%.
Table 7: Panel Estimation Results, County Transportation Employment
Random Effects Model
Coef. Std. Err. z P>
lagged transportation emp 0.941369 0.018754 50.2 0
California trans. Emp 0.162234 0.208548 0.78 0.437
education - 0.77362 0.219783 - 3.52 0
loaded inbound 0.30566 0.162449 1.88 0.06
time - 0.03007 0.015571 - 1.93 0.053
Overall R- squared 0.9979
Wald Chi- squared 15715.97
P- value of Wald 0.0000
Fixed Effects Model
Coef. Std. Err. t P>
lagged transportation emp 0.926577 0.060404 15.34 0
17
California trans. Emp 0.084108 0.190169 0.44 0.661
education - 0.38518 0.238721 - 1.61 0.117
loaded inbound 0.149598 0.152234 0.98 0.333
time - 0.01447 0.014496 - 1 0.326
Overall R- squared 0.9977
F- statistic 101.72
P- value of F- stat 0.0000
4C. Transportation Payroll Model
The last labor market model estimated is that of county level transportation
payroll. Again using a panel data approach, the explanatory variables include the lagged
dependent variable ( real transportation payroll), California transportation gross state
product, education, and loaded inbound containers. The results of the Hausman test
suggest that the county effects should be measured as random effects and the estimation
results are presented in Table 8.
Table 8: Random Effects Estimation Results, County Transportation Payroll
Coef. Std. Err. z P>
lagged trans payroll 0.946666 0.016871 56.11 0
California transport
GSP
1.031879 0.328596 3.14 0.002
education - 0.58069 0.229147 - 2.53 0.011
loaded inbound 0.259425 0.191507 1.35 0.176
time - 0.03894 0.017768 - 2.19 0.028
Overall R- squared 0.9975
Wald Chi- squared 12941.55
P- value of Wald 0.0000
The coefficient on loaded inbound containers suggests that a 1% increase in
loaded containers through the SPB ports results in a 0.26% increase in county- level
18
transportation payroll, however, this coefficient is only significant in a 10% one- tailed
test.
It should be noted that the signs on the education variable are opposite of what is
expected in all three models. We anticipate that this is due to the lack of precision in
what the education variable measures, which is only the percent of residents with a high
school degree. We would have preferred a more detailed measure of the degrees earned
by the residents of the four counties, however, this data was not available for the most
recent years. As more detailed data becomes available, the models can be re- estimated.
4D. Combining the Forecast and Labor Market Results
To estimate the impact of the declines in inbound containers through the Ports of
Los Angeles and Long Beach on the regional economy, we combine the forecasts ( from
section 3) with the estimation results above.
Table 9 presents the actual loaded inbound container counts for 2008 as well as
the forecasts for 2009 ( last two quarters forecast) and 2010 for the Theory, Correlation,
and Control forecasts.
Table 9: Forecast Container Counts
Theory Correlation Control
2008 7,246,382 7,246,382 7,246,382
2009 5,894,153 5,802,757 5,659,733
2010 5,378,085 5,892,978 4,663,847
percent change 2008- 9 - 18.66% - 19.92% - 21.90%
percent change 2009- 10 - 8.76% 1.55% - 17.60%
19
Recall that the Theory and Correlation models appeared to perform best when evaluating
the forecasts. These two forecasts ( and not the Control model) will be used to evaluate
the potential impact of declining port traffic on the regional labor market.
Table 10 presents the projected declines in total employment, transportation
employment, and transportation payroll associated with the forecasted declines in
inbound containers.
Table 10: Forecasted Impact on 4- County Labor Market
Forecasted Percentage
Change 2008- 9
Forecasted Percentage
Change 2009- 10
Model Coefficient Theory Correlation Theory Correlation
Total Employment 0.26 - 4.85% - 5.18% - 2.28% 0.40%
Transportation
Employment
0.31 - 5.78% - 6.18% - 2.71% 0.48%
Transportation Payroll 0.16 - 2.99% - 3.19% - 1.40% 0.25%
The impact on total employment in the 4 County area from declining container
counts in 2009 is estimated to range from - 4.9% to - 5.2%. As expected, the hit to the
transportation employment is estimated to be of higher magnitude, - 5.8% to - 6.2%,
however the decline in transportation payroll is estimated to be approximately 3%.
While the Theory forecast estimates continued declines in employment and
payroll in 2010, the Correlation forecast estimates no substantial impact on the regional
labor market ( due to the fact that the correlation forecast actually predicts inbound
containers increasing in 2010 slightly over 2009 levels).
It is perhaps more useful to translate the percentage changes into actual numbers.
Table 11 presents the actual employment and payroll along with the forecasted levels of
employment and payroll associated with the Theory forecast ( the worst case scenario).
20
Table 11: Forecasts of Labor Market Outcomes by County
Total Employment Transportation Employment Transportation Payroll ( in
millions)
2008 2009 2010 2008 2009 2010 2008 2009 2010
LA 4,076,200 3,878,431 3,790,141 148,500 139,910 136,725 6,764 6,562 6,470
OC 1,489,300 1,417,042 1,384,784 25,400 23,931 23,386 1,129 1,095 1,095
IE 1,222,508 1,163,194 1,136,715 64,450 60,722 59,339 2,364 2,293 2,293
5. CONCLUSIONS AND RECOMMENDATIONS
Using time series forecasting techniques, we developed multiple forecasts of inbound
container traffic through the SPB Ports. These forecasts, combined with panel data labor
market models, allow us to forecast the impact of declining freight volumes on total
employment, transportation employment, and transportation payroll in Los Angeles and
Orange Counties and the Inland Empire. We find that the decline in traffic is associated with
a decline of nearly 330,000 jobs in 2009 and 147,000 jobs in 2010 in the 4- county region.
Transportation employment is estimated to have declined by nearly 14,000 jobs in 2009 due
to declining port activity and forecast to decline by another 5,000 in 2010.
These are preliminary results that will benefit from additional estimations as more
data becomes available. While two of the forecasts were deemed acceptable using the
appropriate diagnostic tools, it should be noted that neither of these forecasts predicted the
sharp decline in traffic experienced in the first half of 2009. As more data become available,
the forecasts became more accurate. This is not surprising, given that the fall in inbound port
21
traffic was unprecedented. This suggests that continued data collection as time passes will
allow us to extend the current forecasting models and make them more reliable.
The same recommendation applies to the regional labor market models. These may
be extended with additional years of data and perhaps to include more counties. Both of
these efforts are relatively low cost and will allow the development of valuable economic
forecasts that are kept current and constantly re- evaluated to test the models' performance as
more data become available.
6. IMPLEMENTATION
The data and models will be uploaded to a webpage accessible to area researchers
and available through the Department of Economics at California State University Long
Beach. The data on this page will be regularly updated.
22
7. REFERENCES
Bank For International Settlements, BIS Effective Exchange Rate Series,
http:// www. bis. org/ statistics/ eer/ index. htm.
BST Associates, “ Trade Impact Report,” prepared for the Port of Los Angeles, Port of
Long Beach, Alameda Corridor Transportation Authority, March, 2007,
http:// www. portoflosangeles. org/ DOC/ REPORT_ ACTA_ Trade_ Impact_ Study. pdf.
Diebold, Francis X. Elements of Forecasting, 2006, South- Western College Publishing.
Employment Development Department, Industry Employment,
http:// www. labormarketinfo. edd. ca. gov/? pageid= 1014.
Employment Development Department, California Regional Economics Employment
Series, http:// www. labormarketinfo. edd. ca. gov/? pageid= 173.
Federal Reserve Bank of St. Louis, Economic Research, GDP and Components,
http:// research. stlouisfed. org/ fred2/ categories/ 18.
Krugman, Paul and Maurice Obstfeld, 2008, International Economics: Theory and
Policy, 8th ed., Addison- Wesley.
Organisation for Economic Development and Cooperation, OECD Outlook No. 85, June
2009, http:// stats. oecd. org/ Index. aspx? DataSetCode= EO85_ MAIN.
Port of Long Beach, TEUs Archive Since 1995,
http:// www. polb. com/ economics/ stats/ teus_ archive. asp.
Port of Los Angeles, Historical TEU Statistics,
http:// www. portoflosangeles. org/ maritime/ stats. asp.
RAND California, Business and Economic Statistics, California Employment and
Unemployment, http:// ca. rand. org/ stats/ economics/ employmentNAICS. html.
RAND California, Business and Economic Statistics, Gross State Product,
http:// ca. rand. org/ stats/ economics/ gspnaics. html.
RAND California, Education Statistics, http:// ca. rand. org/ stats/ education/ education. html.
U. S. Census Bureau, Foreign Trade Division, USA Trade Online,
http:// www. usatradeonline. gov/.
U. S. Department of Labor, Bureau of Labor Statistics, Consumer Price Index,
http:// www. bls. gov/ CPI/.