U.S. Lawn And Garden Tractor And Equipment Market. Analysis And Forecast to 2020
Raymond Lift Truck Report
1. Estimation of Lift Truck Industry Sales
in U.S. Counties Using Economic Census
Data
Report Prepared By:
Cristina DeFilippis
Xue Gong
Yi Liu
M.P.S. Students, Department of Statistics
Cornell University: Ithaca NY, 14850
April 30, 2015
3. Estimation of Lift Truck Industry Sales in U.S.
Counties using Economic Census Data
Cristina DeFilippis
Xue Gong
Yi Li
Abstract
It is in the interest of The Raymond Corporation to understand as much as
possible about local markets (counties). Ideally, they would like to know how
many trucks were sold to each industry in each county, and what class/type of
trucks they are. The Industrial Truck Association (ITA) only provides sales data
at the county level by truck class, not by type or industry. Our goal was to bridge
this gap in the ITA data. We correlated U.S. census data (number of employees)
by industry with ITA Class 2 (C2) truck sales by industry and found four major
industries that were highly correlated: Manufacturing, Wholesale Trade, Retail
Trade, and Transportation and Warehousing. This result enabled estimation of
the number of C2 truck sales in these four industries from the Census Bureau
data for every county. This left only sales by truck type, which is identified by a
Lift Code, within C2 to be determined. It was assumed that the breakdown of
type of C2 trucks in an industry is independent of location. We found that the
Raymond Corporation sales for of each type of C2 truck are highly correlated
with the ITA sales for of each type of Class 2 truck. This result enabled the use
of the Raymond percentage of C2, Type XX truck in each industry to estimate
the number of C2, Type XX trucks sold to each of the four industries in each
county in the U.S.
1 Introduction
1.1 The Raymond Corporation
The Raymond Corporation is a leading provider of lift trucks, forklifts and
material handling solutions in North America. One of Raymonds business goals
is to increase their North American market share in all three classes of electric
lift trucks (Class 1 Counterbalanced trucks; Class 2 Narrow Aisle trucks; Class
3 Walkies). Their major production is in Class 2 lift trucks, for which their
sales are representative of the market. Growth in market share is validation
that they are meeting their customers needs.
2
4. 1.2 The Industrial Truck Association (ITA)
ITA member companies include industry manufacturers of lift trucks, tow trac-
tors, and other industrial truck vehicles throughout the U.S., Canada and Mex-
ico. The ITA comprises 90 percent of the forklift manufacturers in the United
States and Canada. They collect sales data from member companies and dis-
seminate the totals, allowing member companies to calculate market share using
their internal information and the industry totals. However, member companies
do not want to disclose any data specific enough to allow competitors to target
their accounts. Therefore, data from the ITA is deliberately provided in a segre-
gated fashion to avoid specificity. For example, ITA provides information about
end user industries, sales by county, and types of trucks sold- but separately,
with no direct link between datasets.
1.3 Local Markets
The Raymond Corporation and their authorized dealers would like to under-
stand as much as possible about local markets. Localized market knowledge
will help establish more relevant sales expectations and focus the companys ef-
forts in pursuing prospective customers. By correlating ITA information with
publicly available Census Bureau information on county business demographics,
characteristics of local markets can be estimated for their sales network.
1.4 Objective
Our objective is to estimate of the type of lift trucks purchased by various
industries within counties in the United States. The key question is How many
Class X Code Y trucks were sold to industry I in county Z in year T?
We will use Raymond Corporation data along with Census Bureau infor-
mation in order to bridge the gaps between the ITA datasets. The Raymond
Corporation contains a majority market share in Class 2 trucks. By correlating
Raymond Class 2 data with ITA Class 2 data by industry and by type of truck,
we can obtain market information about the types of Class 2 trucks in each
industry (i.e., how many Class 2 Code Y trucks were sold to industry I ). The
only piece missing then, is information about what industries are present in
each county (i.e. industry I in county Z), which will be obtained using Census
data. This will enable us to obtain estimates of the types of Class 2 trucks sold
in each industry for each county.
After carrying out these analyses, we were able to obtain appropriate esti-
mates for 40% of the counties which had greater than average sales across the
years studied.
3
5. 2 Data Sets
2.1 ITA
There are three ITA data sets that contain data from all member companies.
The datasets contain the number of units (trucks) sold per month for the years
1990-2014. The industries are identified by a four-digit Standard Industrial
Classification (SIC) code.
• Retail Orders by Class, County
• Retail Orders by Class, Industry
• Factory Orders by Class, Lift Code
2.2 Raymond Corporation
There is one Raymond Corporation data set that contains the number of trucks
sold per year for the years 2005-2014. The data contains information about
Class, Lift Code, Industry, and County. Like the ITA data, industries are
identified by a four-digit SIC code.
2.3 Census Bureau
We obtained Census data online on County Business Patterns for the years
2005-2012. It consists of the demographic variables Number of Establishments,
Number of Paid Employees, and Annual Payroll (in thousands of dollars) for
each industry and each county. The industries are identified by a two-digit
North American Industry Classification (NAIC) code. We converted SIC codes
to their two-digit NAICs code equivalents in order to compare industry data
from Raymond and the ITA with the census industry data.
3 Methods and Results
3.1 Census-ITA Correlation by County
We extracted the Census business pattern data in year 2005 to 2012 from Amer-
ican Factfinder website. We aggregated employee numbers for the four major
industries yearly (see Section 3.2 ). We selected ITA sale units by county in year
2005 to 2012, and filtered out classes other than Class 2. We dropped all missing
values from the ITA data to ensure that all counties we used for the correlation
test had 8 data points. Then we found the correlations between em-ployee
numbers and truck sales in year 2005 to 2012 of counties. After carrying out these
analyses we were able to obtain suitable estimates for 179 of the 724 total
counties that have had lift truck sales in 2005 to 2012 (roughly 25%). Of the 166
counties that had greater than average sales across years 2005 to 2012, we
obtained suitable estimates for 64 of them (roughly 40%). See Tables 3 and 4 in
the Appendix for more details.
4
6. 3.2 Census-ITA Correlation by Industry
The first step in the analysis was to establish a correlation between ITA truck
sales by industry with Census Bureau data by industry at the national level. To
do this, the ITA industry codes were converted from SICs to NAICs. National
totals for each industry for each year 2005-2012 were used. Separate correla-
tion analyses were done using several census industry attributes: Total Sales,
Number of Employees, Number of Establishments, and Annual Payroll. The
analyses were done on the raw numbers of the census attributes and ITA sales
as well as for the growth rates of each from year to year. In total 8 correla-
tion analyses were done, using the four census attributes and in raw/growth
rate form. The results were compared and the census attribute/form that was
most highly correlated with the ITA industry sales was selected to use in the
estimation of ITA industry sales by county.
The strongest overall correlation with ITA sales was obtained by using the
census attribute Number of Employees. Using this variable, four major indus-
tries were found to be highly correlated with ITA: Manufacturing, Wholesale
Trade, Retail Trade, and Transportation and Warehousing. Table 1 shows the
summary of the correlation between the number of employees and ITA sales for
each industry. A limitation of this analysis is the low number of data points
(n=8), but we feel that the correlation is strong enough to obtain results with
some confidence for these four industries. Figure 1 below contains the scatter
plot of ITA vs. Census with the fitted regression line for each industry. It
is clear that there is a linear relationship between the two variables for each
industry. Larger-scale plots of this relationship for each of the four industries
individually can be found in appendix Figure 6.
Moving forward we only felt comfortable calculating the estimates of class
and lift code in these four industries for different geographic regions because
they are the only ones with acceptable correlations. Note that a few other
industries had high correlations: mining, information, waste management but
they each only represent 0.3%, 0.4%, and 0.2% of all industries, respectively.
Conversely, in 2014, these four industries selected accounted for 85% of total
unit sales for ITA.
3.3 Industry Sales Estimates by County
Using census variable Number of Employees, four major industries were found to
be highly correlated with ITA: Manufacturing, Wholesale Trade, Retail Trade,
and Transportation and Warehousing. Thus, ITA C2 sales were calculated for
each of these four industries for all counties and all years 2005-2012. The ITA
C2 industry sales in a particular county for a particular year were estimated
using simple percentages as follows:
Let:
xA = total number of C2 trucks sold in County A
pA,i = % of employees in County A in Industry i
5
7. y = 7.2068x -4938
R² = 0.7643y = 34.358x -15355
R² = 0.3757
y = 18.329x -5926.3
R² = 0.5531
y = 24.024x -31944
R² = 0.5941
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
350 450 550 650 750 850 950 1050 1150 1250 1350 1450 1550
NumberofEmployees(tensofthous.)
ITA Sale Units
Manufacturing Wholesale Trade Retail Trade Transportation and Warehousing
Figure 1: Census Number of Employees vs. ITA Sale Units
6
8. Number of Employees (tens of thousands) ITA Class 2 Unit Sales Correlation
NAIC
Code
NAIC Label 2005 2006 2007 2008 2009 2010 2011 2012 2005 2006 2007 2008 2009 2010 2011 2012 2005-2012
11 Agriculture,Forestry,
Fishing and Hunting
17 17 17 17 15 16 16 16 188 134 190 228 153 158 109 199 0.553
21 Mining, Quarrying, and
Oil and Gas Extraction
50 55 70 63 60 58 65 73 26 32 40 35 21 37 35 54 0.746
22 Utilities 63 61 62 64 64 64 64 64 73 64 72 82 74 41 52 53 -0.170
23 Construction 678 734 727 704 597 539 519 526 113 99 92 63 30 52 48 89 0.539
31-33 Manufacturing 1367 1363 1332 1310 1163 1086 1098 1119 4900 4967 4839 4402 2539 2501 3399 3855 0.874
42 Wholesale Trade 597 603 596 617 583 560 563 578 5718 5910 6020 4910 3150 3301 4518 4976 0.613
44-45 Retail Trade 1534 1577 1576 1561 1480 1450 1470 1481 6112 6708 5188 4495 2448 2112 3988 4771 0.771
48-49 Transportation and
Warehousing
417 431 440 444 416 401 411 423 2149 2238 2075 2026 1431 1257 1616 1785 0.744
51 Information 340 340 340 343 329 312 312 314 129 177 138 117 142 81 53 73 0.825
52 Finance and Insurance 643 665 655 651 617 593 589 598 45 25 37 13 6 24 12 31 0.316
53 Real Estate, Rental, and
Leasing
214 222 222 220 204 195 192 194 83 130 68 189 37 172 139 201 -0.326
54 Professional, Scientific,
and Technical Services
769 805 818 803 784 782 793 802 140 126 175 166 83 67 125 162 0.602
55 Management of Compa-
nies and Enterprises
286 292 312 289 285 283 292 304 0 4 0 3 0 1 1 5 0.193
56 Administrative, Support,
Waste Mgmt, and Remedi-
ation Services
928 1000 998 1022 906 898 939 987 45 41 69 54 16 10 30 38 0.804
61 Educational Services 288 298 304 314 320 327 339 348 33 33 39 31 40 40 28 26 -0.390
62 Health Care and Social
Assistance
1603 1645 1680 1722 1753 1779 1806 1838 38 77 59 67 38 35 46 48 -0.331
71 Arts, Entertainment, and
Recreation
194 197 201 207 201 200 200 206 16 15 28 32 13 15 13 13 0.401
72 Accommodation and Food
Services
1103 1138 1156 1193 1144 1131 1156 1199 26 38 44 39 20 20 23 22 0.189
81 Other Services (except
Public Administration)
539 546 552 545 526 520 518 526 192 178 151 208 92 122 158 150 0.560
Table 1: Correlation between the census data, Number of Employees (tens of
thousands), and the ITA data, Class 2 Unit Sales, for each industry across the
years 2005-2012. Positive correlations are highlighted in green,while negative
correlations are highlighted in red. The four industries that were selected to
obtain estimates for are bolded.
Then the estimate C2, Industry i truck sales in County A is:1
yA,i = xA × pA,i
3.4 Raymond-ITA Correlation by Industry
We assumed that Raymonds sales are representative of ITA sales. Therefore, by
applying the fractions of ITA C2 lift code sales by industry for Raymond, we can
get the estimated ITA lift code breakdown in different industries for Class 2. To
validate our assumption, we first checked if we could estimate ITA C2 industry
sales using Raymond data. We started by calculating the Raymond fractions
of different industries in Class 2. Then, we applied the fraction of Raymond
C2 sales in each industry to the aggregated C2 ITA sales and got the estimated
1Note that we actually obtained three estimates of yA,i for each county: an average, upper,
and lower estimate. This is because in the census data, some counties contain industries with
a letter listed instead of a numerical value for the number of employees. Each letter represents
a range of the number of employees for that industry. A table listing these letter ranges can
be found in Appendix Table 8. Since the totals of Number of Employees are listed in the
table regardless of whether any industry has a letter, we used the lower and upper bound of
the range to get the average, upper and lower Number of Employees in these industries.
7
9. y = 0.8771x + 65.961
R² = 0.9168
N = 670
p < .0001
0
5000
10000
15000
20000
25000
30000
35000
0 10000 20000 30000 40000
ActualUnitSales
Estimated Unit Sales
Figure 2: ITA Class 2 Industry Sales vs. ITA Estimates from Raymond Frac-
tions
ITA sales in each industry. We then compared estimates of industry breakdown
based on Raymond fractions with the actual ITA sales in each industry.
In order to see how well the Raymond data could estimate the ITA data, we
performed a regression analysis on actual ITA Class 2 industry sales vs. their
estimates (obtained from Raymond fractions). As shown in Figure 2, the model
fits well; 91.7% of the variability in the actual sales is accounted for by the
variability in the estimated sales (R2
=0.9168). There is strong evidence to con-
clude that there is a linear relationship between actual unit sales and estimated
unit sales of different industries, and the correlation is 0.958 (p-value<0.0001).
Descriptive statistics of the actual and estimated industry sales can be found in
Appendix Table 7.
3.5 Raymond-ITA Correlation by Lift Code
The second step in validating our assumption that Raymonds sales are repre-
sentative of the ITA sales was to check if we can estimate ITA C2 lift code sales
using Raymond data. We followed a similar procedure as in section 3.4 and
started by calculating Raymond fractions of different lift codes in Class 2. We
applied the fraction of each lift code in Class 2 from Raymond to the aggregated
C2 ITA sales and got the estimated ITA sales units for each lift code. We then
compared estimates of lift code breakdown based on Raymond fractions with
the actual ITA sales.
In order to see how well the Raymond data could estimate the ITA data,
we performed a regression analysis on actual ITA Class 2 industry sales vs.
their estimates (obtained from Raymond fractions). As shown in Figure 3, the
model fits well; 99.7% of the variability in the actual sales is accounted for by
the variability in the estimated sales (R2
=0.9969). There is strong evidence
8
10. y = 0.9756x + 2094.8
R² = 0.9969
N = 5
p < .0001
0
50000
100000
150000
200000
250000
0 50000 100000 150000 200000 250000
ActualUnitSlaes
Estimated Unit Sales
Figure 3: ITA Class 2 Lift Code vs. Estimates from Raymond Fractions
to conclude that there is a linear relationship between actual unit sales and
estimated unit sales by lift code, and the correlation is 0.998 (p-value<0.0001).
Descriptive statistics of the actual and estimated lift code sales can be found in
Appendix Table 6.
3.6 Result of Raymond-ITA Correlations
By correlating Raymond Class 2 data with ITA Class 2 data by industry and
lift code, we were able to validate our assumption that Raymond data is rep-
resentative of ITA data in these aspects. Therefore, we can use the Raymond
fractions of each lift code in each industry to estimate the ITAs Class 2 lift code
breakdown in each industry for individual counties.
For simplification, we assume the breakdown of lift code in industry is in-
dependent of location (county level). This means the distribution of each lift
code in industry stays the same across all counties. Thus, we can simply apply
these fractions to the estimated ITA Class 2 industry sales for a given county
to obtain Class 2 lift code sales in each industry (see Section 3.7). The Ray-
mond fractions of each Class 2 lift code in each major industry can be found in
Appendix Table 5, and is also visualized in Figure 4.
3.7 Calculating Lift Truck Sales Estimates
Once the estimates of C2, Industry i truck sales were obtained for each county,
the final step was to further break this down by truck type (Lift Code). The
Raymond percentage of C2, Lift Code j truck in each industry was used to
estimate the number of C2, Lift Code j trucks sold to each of the four industries
in each county in the U.S. The ITA C2, Lift Code j truck sales for a particular
industry, county, and year were estimated using simple percentages as follows:
Let:
9
11. Figure 4: Raymond Lift Code Fractions of Major Industries
yA,i = xA × pA,i = estimate of C2, Industry i truck sales in County A 2
qi,j = % of Code j trucks sold in Industry i 3
Then the estimate of C2, Industry i, Code j truck sales in County A is:4
yA,ij = yA,i × qi,j
4 Discussion
4.1 Practical Use of the Excel Model
The excel model is designed to store the analysis results and present the es-
timation in a well-structured format. With the free selection of state, county,
years, and life codes, it enables the user to access the estimations from different
approaches.
There are two tabs that are the most relevant to users:
• Input Location- This tab asks for user selection of a specific county in
order to produce the corresponding estimates and graphs.
2Calculated in Section 3.3
3Calculated from Raymond data
4Note that there will again be three estimates calculated in our excel model here: average,
low, and high estimates of unit sales, for the same reasoning as before.
10
In the end, we were able to obtain estimates of the lift truck sales for the five lift
codes (1,2,3,4,6) within the four industries (Manufacturing, Wholesale, Retail, and
Transportation/Warehousing) across eight years (2005-2012) for 232 counties with
high confidence.
12. • Output Results- After making the selections in the Input Location tab,
this is where the resulted table and graphs will be displayed. Users will
be able to see the actual Raymond lift code sales in a particular county
and industry, and directly compare that to the market sales for the same
county and industry. This will enable the users to learn about the local
markets and identify potential sales opportunities.
For example, in order to get the estimate of ITA Order Picker trucks (lift
code 2 in Class 2) in Los Angeles, CA in 2011, the procedures would be: In the
Input Location tab, select California for state and Los Angeles for county. The
Geo Code is automatically generated. In the Output Results tab, select 2011
for year and check 2 for life code. The estimation table and plot by industry
would be generated. The sample input and output can be found in Figures 7
and 8 in the Appendix.
4.2 Limitations of Correlating Census Employee Numbers
with ITA Truck Sales
4.2.1 Different Sensitivity to Economic Conditions
We used number of employees instead of other Census variables in our national
level analysis because the number of employees had the highest correlation with
ITA sales. We can see in Figure 5 showing annual ITA sales that around the year
2009, the ITA truck sales dropped dramatically across all industries. However,
the decrease in employee numbers from Census data was not as great. ITA
truck sales are more sensitive to economic changes compared to the number of
employees. The 2008 economic crisis is a factor that may have impacted our
analysis. If we were able to obtain census data for more years, we might drop off
these extreme years from our analysis. Because we only had access to Census data
from 2005 to 2012, we did not drop the extreme period in our analysis.
4.2.2 Regular Life/Lease Cycles of Trucks
Different types of trucks vary in their life cycles and lease cycles. For example,
Class 2 l ift trucks usually have 5 to 7 years of l ease and 10 to 12 years of working
life on average. Therefore, correlating the sum of all new l ift truck sales i n a cycle
period with the sum of the Census data during that period may yield better
results; as the i nterim values may not be linear based on fleet replacement cycles.
It might be more reasonable to look at the aggregated data for several years, and
the estimations could also be made for a longer period, instead of a single year.
11
14. 5 Acknowledgements
This project consumed huge amount of work, research and dedication. Still,
implementation would not have been possible if we did not have the support
of our sponsor company and our academic adviser. Therefore we would like to
extend our sincere gratitude to them.
First of all, we are very thankful to everyone at the Raymond Corporation
for their support and guidance throughout the semester. Especially to Rebecca
Graham, Arlan Purdy, and Mark Day for the time they spent working with us
during weekly conference calls and through emails, which was invaluable to the
success of the project. Without their industry experience, the project would
never have gotten off the ground, and thus their support has been essential.
We are also very grateful to our project adviser Johannes Lederer for his
provision of expertise, encouragement, and academic support in the implemen-
tation.
6 References
1. Raymond Overview for BU 10 9 14. [Powerpoint slides]. Raymond Cor-
poration, Greene, NY.
2. U.S. Census Bureau. 2015. County Business Patterns 2005 - 2012; using
American FactFinder; ¡http://factfinder2.census.gov¿;(May 29, 2014).
3. U.S. Census Bureau. 2015. 1987 SIC to 2002 NAICS; generated by John
Murphy ¡https://www.census.gov/eos/www/naics/concordances¿;(July 14,
2008).
13
15. 7 Appendix
Missing Data Points High Positive Low Positive No or Negative Correlation
0-2
A
(179)
3-5
B
(53)
C
(99)
D
(393)
>5
N/A
(2496)
Table 2: Grading System County Correlation Scores and Counts of
Counties With Each Grade. ”High Positive” correlation is defined as having
a correlation coefficient greater than or equal to 0.4. Similarly, ”Low Positive”
is having a coefficient between (0.2, 0.39), and ”No or Negative Correlation” is
having a coefficient less than 0.2. ”N/A” means that there is not enough data
points (years) to perform a correlation analysis. Many counties may never have
had any truck sales sure this time period of 2005-2012.
14
17. Average Sales
per Year
High Positive Low Positive No or Negative Correlation Total
79 29 97 205
>10000
39% 14% 47% 100%
42 11 39 92
>25000
46% 12% 42% 100%
Table 3: ITA-Census Correlation Grouped by County Size (as mea-
sured by Number of Employees)
Average Sales
per Year
High Positive Low Positive No or Negative Correlation Total
64 22 80 205
>=13
39% 13% 48% 100%
33 10 36 79
>=30
42% 13% 46% 100%
Table 4: ITA-Census Correlation Grouped by County ITA Unit Sales
Figure 7: Example Excel Model Input for Los Angeles, CA
16
19. NAICS 2digits Lift Code Percentage
31-33
1 0.43%
2 27.45%
3 59.00%
4 10.61%
6 2.51%
42
1 0.01%
2 38.67%
3 54.25%
4 5.21%
6 1.85%
44-45
1 0.01%
2 36.22%
3 58.67%
4 3.11%
6 1.99%
48-49
1 0.24%
2 29.44%
3 64.33%
4 4.31%
6 1.68%
Grand Total 100%
Table 5: Raymond Lift Code Fractions in Each Major Industry
Variable N Mean Std Dev Sum Minimum Maximum
ITA Actual Sale 5 85797 104944 428986 1557 235884
Estimated Sale 5 85797 107401 428986 567.9167 244648
Table 6: Summary Table of Estimated Unit Sales and ITA Unit Sales
in Lift Code
Variable N Mean Std Dev Sum Minimum Maximum
ITA Actual Sale 670 593.0328 2239 397332 1 32078
Estimated Sale 670 60039192 2444 402616 -12.2532 244648
Table 7: Summary Table of Estimated Unit Sales and ITA Unit Sales
by Industry
18
20. Letter Range of employee numbers
a 0 to 19 employees
b 20 to 99 employees
c 100 to 249 employees
e 250 to 499 employees
f 500 to 999 employees
g 1,000 to 2,499 employees
h 2,500 to 4,999 employees
i 5,000 to 9,999 employees
j 10,000 to 24,999 employees
k 25,000 to 49,999 employees
l 50,000 to 99,999 employees
m 100,000 employees or more
Table 8: Census Number of Employees Letter Codes
19