View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

FEDERAL RESERVE BANK OF RICHMOND

MONTHLY
REVIEW




A SEASONALLY ADJUSTED WORLD
THE CENSUS SEASONAL ADJUSTMENT TECHNIQUE
One of those anonymous seasoned observers of the
economic scene allegedly said that, seasonally ad­
justed, the Great Lakes never freeze. Another com ­
monplace is that a camel, seasonally adjusted, might
well be a horse. In any event, the U. S. Department
of Labor reported that the unemployment rate in
January was 4.2% , but seasonally adjusted it was
only 3.9% . It later reported that the Consumer
Price Index increased at an annual rate of 7.2% in
April, but that the seasonally adjusted rate was only
6 % . It is probably of precious little consolation to
the consumer when he purchases groceries to know
that part of the rising prices is only seasonal in
nature, but the knowledge should help him to plan
his future purchases. Th e business conditions
analyst, on the other hand, finds it essential to know
whether a given rise in prices is seasonal in nature.
The seasonality of price changes even affects his fore­
cast of economic policies, for if prices are likely to
fall later in the year, policy makers would not be ex­
pected to impose any new anti-inflationary measures.
A s a general rule, policy making is greatly in­
fluenced by seasonal adjustment procedures. A
policy maker or business analyst needs to know
whether an increase or decrease in a series of busi­
ness data is seasonal in nature or whether it might
indicate a longer-term tendency of the economy. A
business strategy designed to cope with long-term
tendencies might well be different from one designed
to remedy seasonality. For example, one would ex­
pect dog sled sales in July to be less than in January.
A policy prescription to remove some of the sea­
sonality in those sales might be aimed toward find­
ing some summer use for a dog sled. On the other
hand, if sales are falling below what would be ex­
pected after allowing for seasonal influences, a dif­
ferent set of forces are probably responsible for the
decline. This set of forces could include such things
as declining consumer income, relative price and cost
changes, changes in consumer tastes (away from
dogs toward snowmobiles), etc.
In recent months some problems associated with
seasonal adjustment have received attention in the
financial press. In particular, the accuracy of the
seasonally adjustments for automobiles and retail
sales series was questioned. Last fall, the earlier in­
troduction of new model automobiles necessitated a
reexamination and revision of the seasonal adjust­

2


ment factors for automobile sales. Lrntil new factors
were published, it was difficult to find much meaning­
ful information in the data. The retail sales series
seems to be subject to chronic seasonal adjustment
problems. Since retail sales are influenced by unusual
weather, holidays, and other special events, it is often
difficult to isolate the regular seasonal influences in
the data. In any event, the recent problems in in­
terpreting economic data because of seasonal dif­
ficulties have pointed out how important it is to be
aware of what is behind the statistical technique that
is called seasonal adjustment.
Most persons have a general idea of the meaning
of seasonal adjustment. However, it is easy to
forget the many assumptions and technical considera­
tions that lie behind the statistical processes that con­
vert raw data into seasonally adjusted data. In an
effort to make this information more readily avail­
able, this article will describe some general prin­
ciples of seasonal adjustment. Then its focus will
turn to a technique of seasonal adjustment which is
probably used more often on economic data than any
other— the U. S. Department of Commerce, Bureau
of the Census’ X - l l Variant of Census Method II.
The name “ X - l l Variant of Census Method I I ”
actually refers to a computer program which is used
by the Census Bureau to adjust economic data for
seasonal influences.
The X - l l program is also
available to other users and is widely used through­
out the country.
The discussion of X - l l relies in no small measure
on the technical papers published by the Bureau of
the Census which are listed at the end of the article.
SEASONALLY ADJUSTING A TIME SERIES
A time series is the name given to any series of
data regularly spaced over time. The variation in
such a series is conventionally assumed to reflect
trend, cyclical behavior, seasonal behavior, and ir­
regular influences. The trend and cycle components
reflect the longer-term tendencies of the series,
whereas the seasonal and irregular components re­
flect the shorter-term variation.
Suppose that the time series to be adjusted is
daily ice cream sales by the neighborhood vendor.
An examination of his daily sales statistics over
several years should show changes in sales attri­
butable to long-term influences like increasing popula­

tion in the area or the changing makup of the popula­
tion, or changes in disposable income. The ice cream
sales data also will show regular fluctuations at
given times of the year. M ore ice cream will ob­
viously be sold during the warmer months. These
regular fluctuations make up the seasonal component
of the time series. Short, erratic upturns or down­
turns in sales can be caused by any number of events
from flat tires to parades. Such short and erratic
fluctuations make up the “ irregular” component of
the time series.
The task for any seasonal adjustment technique is
to isolate the seasonal variation by removing the
trend, cycle, and irregular components from the
original series. The resulting set of seasonal factors
or seasonal index numbers reflects the seasonal
variation in the data and may be used to remove
the seasonal influences from the original raw data.
Almost every seasonal adjustment technique in
regular usage employs some kind of moving average.
A moving average is simply an average that moves
forward one period at a time, dropping one term and
adding another. This technique has the property of
smoothing the fluctuations in the data. For example,
to smooth the irregular jumps in a series like do­
mestic automobile sales, which is published for tenday intervals, a three-term moving average of the
ten-day sales periods might be utilized. The average
of the first three periods would be computed, and the
resulting figure would represent the automobile sales
for the second ten-day period. Then, sales in the
second, third, and fourth time periods would be
averaged and the average used to represent sales in
the third period. This simple process is repeated,
moving forward one period at a time, as long as
data are left to average.
In the event that the period of the moving average
consists of an even number of terms, the average is
not centered on a particular term.
It must be
centered by again averaging each pair of averages.
This procedure is illustrated in the accompanying
table.
After the “ placement” or “ centering” problem is
solved and the moving average is satisfactorily de­
termined, each average term is divided into the
original number for its corresponding time period.
Since each term of a 12-month (or 4-quarter) moving
average includes data from an entire year, it is un­
affected by seasonal and most irregular influences
and can be used to trace the longer-term movements
in the series. W hen the original data are divided by
the corresponding moving average value, the re­
sulting ratio, therefore, reflects the seasonal and ir­
regular variation in the data. The long-term varia­



tion is occasionally removed by subtracting the
moving average value from the original data, but
usually the components of the time series are as­
sumed to be related multiplicatively so the long-term
variation is most often removed by division. This
technique of dividing by the moving average is called
a ratio-to-moving average method. T o convert the
resulting ratios into a series of index numbers, they
are multiplied by 100.
T h e table illustrates these techniques using
quarterly data on auto sales. The first column shows
the original data; the second shows the results for
the four-term moving average placed between the
quarters; the third column shows the centered moving
average (a two-term average of the second column) ;
the fourth column shows the seasonal-irregular fac­
tors (or ratios) ; and the last column shows stable
seasonal factors. These were derived by averaging
the seasonal-irregular (S I ) ratios by quarter in order
to remove the irregular variation. For the data in
Table I, the seasonal factor for the first quarter
would thus be calculated to be 0.975, the average of
0.936 and 1.005.
Suppose, for example, that six years of monthly
data are to be used to determine the seasonal
changes in a series of economic data. After calculat­
ing the moving average and the ratio-to-moving
average series, a series of five years of monthly
seasonal-irregular ratios will result. This is true
because data are lost for the first half of the first
year and the last half of the last year when the 12
month moving average is calculated. The adjustor
can then average the five SI values for each month
to remove any irregular (random ) jumps in the
series. Normally, the median is chosen in this pro­
cedure to remove the sensitivity to extreme irregular
values. The result is a series of 12 seasonal ratios,
one for each month.
This method of seasonal adjustment is also based
on the assumption that the seasonal pattern in the
data remains stable over the time period. In the
real world seasonals are rarely stable, and most pro­
cedures in actual use are designed to allow for
changing seasonality. Thus, those techniques would
supply 60 instead of 12 seasonal ratios for the 5
years of monthly data.
Since one of the principles of seasonal adjustment
is that strictly seasonal patterns should average out
over the course of a year, monthly seasonal ratios
should add to 12 (to average one). In other words,
seasonal adjustment should not affect the annual
total. If the year’s ratios do not average to one,
they are adjusted so that they will, usually by mul­
tiplying each ratio by a constant adjustment factor.
3

AUTOMOBILE SALES
M illions of Autos at A nnual Rates Not Seaso n ally A djusted*
1
Autom obile
Sales1

2

3

Four-Term
A verage

4

Centered
A verage

Season alIrreg ular
Factors2

5
Seasonal
Factors3

1967

First Q uarter
Second Q uarter
Third Q uarter
Fourth Q uarter

6.17
894
6.77
7.58

7 37
_

7.61

7.93

0.890

0.894

7.90

0.959

1.023

1968

First Q uarter
Second Q uarter
Third Q uarter
Fourth Q uarter

8.09

8

1.005

0.971

8.40
8.63

1.106
0.897

1.101
0.894

*

9.40

8.05
‘
*

9.29
7.74

8.65

1.087

1.023

' .

8.66
8.56

0.936
1.096

0.971
1.101

8.66

1969

First Q uarter
Second Q uarter
Third Q uarter
Fourth Q uarter

8.11
9.38
7.70

8.46

* Computed using monthly figures.
1 N ew car sales by U. S. fran chise dealers.
2 Column 4 equals Column 1 divided by Colum n 3.

8.22
3 Column 5 show s a v e ra g e of each quarter's ratio in Column 4.
Note:
Figures derived from data published in W ard's Autom otive
Reports.

THE X -l 1 VARIANT OF THE CENSUS METHOD II
SEASONAL ADJUSTMENT PROGRAM
The procedure outlined above is a simple seasonal
adjustment procedure. The actual procedures used
by those who have access to large-scale computer
facilities are considerably more complex. The Bureau
of the Census’ computer program, the X - l l Variant
of the Census Method II Seasonal Adjustment P ro­
gram, is used quite often to adjust series of monthly
economic data. In fact, the technique is used to
seasonally adjust almost all monthly economic data
published by the U. S. Government and is in wide­
spread use throughout the world for adjusting ag­
gregate and even company level data. The method
is more complex, mainly because of all the practical
problems which arise in moving from the rarified
atmosphere of the theoretical into the real world
of practical application.

A m on g these practical

problem s a r e :
1.

2.

Differences in the number of “ trading” or “ work­
ing” days in various months. For example, one
month might contain five Sundays and four
Saturdays; another might contain only four of
each.
Special influences attributable to holidays coming


4


3.

4.

at different dates in different years, e.g., the ef­
fect of Easter on retail sales.
Irregular changes, such as a different introduc­
tion date for a new model automobile, unseason­
able weather, and strikes.
Changing seasonal patterns over time.

The Prior A dju stm ent Factors T h e X - l l p ro ­
gram allows the user to adjust for trading day dif­
ferences by supplying a series of prior weights (or
ratios) for the series before seasonally adjusting it.
These weights can be calculated directly if daily
figures are available or indirectly if they are not.
They may be calculated from daily data by dividing
each number by some measure to take out influences
other than trading day variation. For example, one
might use a seven-day moving average to get a series
of daily adjustment factors, then remove the irregular
jumps in the series either by averaging like days
(Mondays, Tuesdays, etc.) or some other smoothing
technique. Then a series of daily weights would be
obtained which would be used to weight the monthly
data. For example, if the day of the week weights
for a particular month totaled 29.1 and since an
average month has 30.44 days, the data for the month
in question would be adjusted for trading day varia­

tion by multiplying it by a standardizing ratio, 1.046,
which is derived by dividing 29.1 into 30.44. The
data for the remaining months of the series would
be adjusted in like manner.
Also, if daily data are available, the special in­
fluences of holidays may be calculated and incor­
porated into the monthly weights. The retail sales
series around the Easter holiday provides a prime
example of the holiday date problem since Easter
comes on different dates in different years and since
it has an important impact upon retail sales. In
order to adjust a series of retail sales figures for
Easter, one might get data on sales for the several
weeks before and after Easter for a number of years.
A series of Easter factors could then be developed
which would be used to weight the raw data for the
month or months involved to standardize the re­
sulting series for Easter. Similar procedures could
be used for other holidays.
The X - l l program allows the user to supply prior
adjustment factors which can standardize for both
trading day and holiday influences. Also, if a par­
ticular series of data includes irregularly recurring
events, such as different model introduction dates,
periods of unseasonable weather, strikes, etc., the
prior adjustment factors can be designed to account
for those disturbances in the series.
Factors to adjust for different model introduction
dates can be estimated in a manner similar to the
adjustment for Easter mentioned above. The other
types of adjustment for the irregular events are more
difficult and require an intimate knowledge of the
series to be adjusted. It might be added that in
addition to a rather detailed grounding in the data,
the user also needs excellent judgment and no small
measure of luck to weight properly the effects of
these events.

If an event such as a strike or bad

growing season is nonrecurring but has an important
impact upon the data series, the X - l l program can
recognize the problem and ignore data from that time
period in computing the seasonal factors.

series by dividing each SI ratio by its correspond­
ing smoothed SI (o r seasonal) ratio. The resulting
set of irregular factors reflects the random or un­
predictable influences in the data and also includes
the regular variation in the data attributable to dif­
ferences in the number of trading or working
days per month.
T o find the properties of the series which are
regular and caused by trading day differences, each
“ irregular” factor is classified in one of 21 different
categories depending upon the day of the week with
which the month started and whether the month had
29, 30, or 31 days. A twenty-second category also
exists to include 28-day months, but since each 28-day
month has exactly four weeks, all such months are
included in the same category regardless of the
starting day.
After putting each ratio in its appropriate cate­
gory, the “ extreme” irregular values (those way out
of line with the other data) are ignored. A least
squares regression technique is then used on the
remaining irregular ratios to estimate the factor
which would properly weight each of the 22 types of
months for trading day differences. The weights are
then statistically tested to find whether it is likely
that they are due solely to chance. If not, they are
used in the subsequent seasonal adjustment process.
The X - l l technique performs these operations
automatically. After incorporating the prior weights
into the original series, the computer operates on the
weighted series to remove long-term and seasonal
fluctuations, adjusts the resulting irregular series for
extreme values, and then generates a set of weights
for the irregular component. Since irregular series
are mostly random, the standard deviation provides
a measure of the likelihood that a particular value
or range of values will occur. Using this concept, the
irregular ratios are examined and those ratios which
are not likely to be explained by chance alone are
weighted, depending upon how unlikely they are, on a
graduated basis from zero to the full weight.

The

extreme irregulars, therefore, are ignored in subse­
Internally Generated A djustm ents
It is also
possible to have the X - l l program calculate its own
weights for monthly trading day differences. The
calculations are based upon statistical properties of
the data which allow the trading day differences to
be inferred from them. This inferential process
begins by removing long-term influences from the
data by a ratio-to-moving average method to get a
series o f seasonal-irregular ratios. Prelim inary
seasonal ratios are then estimated by smoothing the
irregular components out of the SI series. The
seasonal influences are then removed from the SI



quent calculations because of their zero weight. The
less infrequent irregulars are weighted more heavily,
and all values which are probably not infrequent are
given the full weight of one.
Actually, the X - l l program is designed to perform
these operations even if prior adjustment factors are
supplied.

In that case, these final weights for the

irregular component provide an internally generated
means of removing extreme random influences which
are not adequately covered by the prior adjustment
continued on page 8

5

FEDERAL OUTL/

4

Federal G overnm ent outlays (expenditures and
net lending) more than doubled in the sixties.
N early one-half of the $93 billion increase took
place between 1966 and 1968.
Increasing out­
lays for national defense contributed heavily to
the expansion. Spending for defense leveled off
after 1968 w hile hum an resource program s con­
tinued to expan d rapid ly.
This change in Federal
spending priorities is expected to continue this
fiscal y e a r as hum an resource spending outstrips
national defense for the first time since the 1930's.
Expenditures for physical resource program s more
than doubled in the sixties but rem ained low at
about 12% of the total budget. The sam e is true
of other outlays, prim arily interest on the national
debt, which accounted for about 10% of total out­
lays throughout the period.

NATIONAL DEFENSE
SH ARE O F TOTAL O U TLA YS

Per Cent

M ilita ry Person nel
O p e ra tio n a n d M a in te n a n c e

§5 P rocu rem en t
U O th e r

50

40

30

20

The largest component of total outlays during
the sixties w a s national defense. W hile expen d i­
tures for national defense increased $35.3 billion
during the sixties, the relative im portance of d e ­
fense outlays declined from n early 50% of the total
budget at the beginning of the decade to only
44% in 1969; it is estimated that this figure w ill
fall to 37% this fiscal y e a r.


6


10

1960

61

62

63

64

65

66

F isca l Y e a rs
* F e b ru a r y
S o u rce :

estim a te s.

B u re au o f the B u d get.

67

68

69

70*

71*

IN THE SIXTIES
O u tlays for hum an resource program s showed
the most spectacular growth of the past decade,
e xpan ding by 150% and increasing from 28% of
total outlays in 1960 to 34% in 1969.

The largest

component of this category is income security,
prim arily social security paym ents, which rem ained
at about 20% of total outlays throughout the sixties.
This figure is expected to climb to 25% this fiscal
ye ar as a result of more beneficiaries and higher
benefits.
Health and education expenditures in ­
creased substantially during the period w hile
veterans' benefits declined in relative im portance.
The most rapid increase w a s in health program s
associated with M edicare and M edicaid.

PHYSICAL RESOURCES
SH ARE O F TOTAL O U TLA Y S

20

10

g
■
□

C o m m u n ity D evelo p m en t
C o m m erce a n d
T ra n sp o rta tio n

-

A g ricu ltu re
Sp a ce
N a tu ra l Reso urces

The relative im portance of physical resource pro­
gram s increased in the first half of the decade and
jm then declined to about its original position. During
%-s the sixties net gains w ere m ade by community d e­

-

velopm ent, space, and natural resource program s
w hile net reductions w ere registered in the a re a s
of commerce and transportation an d agriculture.
I9 6 0

61

62

63

64

65

66

67

68

69

70*

7V

W ynnelle Wilson

Fiscal Years
*February estimates.
Source:

Bureau of the Budget.




7

A SEASONALLY ADJUSTED WORLD
continued from page 5

factors. The more usual situation encountered in
actual practice, however, is one in which no prior
adjustment factors are supplied. Extreme irregulars
in the series would then be handled solely by these
internal adjustment methods.
After the final weights are calculated, the X - l l
program adjusts the original series modified by prior
adjustment factors for extreme irregular variation.
T h e resulting data (the series m ight be labeled
M O X I for modified original with extreme irregulars
modified) are used to prepare a set of final seasonal
adjustment factors.
Developing the Final Seasonal Index A centered
12-month moving average is taken of the M O X I
series, and the ratio-to-moving average method is
used to derive a set of seasonal-irregular ratios.
After smoothing the ratios to take out the irregular
variation, the resulting seasonal factors are used to
make a preliminary seasonal adjustment of the
M O X I series.
The Census program then uses a rather sophisti­
cated moving average type technique to smooth the
preliminary seasonally adjusted M O X I series. The
result is a smooth curve which represents trendcycle or long-term fluctuations. This smooth curve
is called the final trend-cycle component and is used
to begin the final adjustment of the M O X I series.
The final trend-cycle variations are removed from
the M O X I series by dividing it by the final trendcycle component, then the remaining seasonal-ir­
regular values are smoothed with a weighted seventerm moving average to remove the irregular in­
fluences. The result is a series of seasonal factors.
Perhaps it might be emphasized that the X - l l
technique, as opposed to less sophisticated tech­
niques, generates a seasonal factor for almost every
time period for which data are supplied, i.e., each
August could have a slightly different seasonal factor.
The seasonal factors are not identical for a given
month unless there is a stable seasonal pattern in the
series over the years.

If, as is usually the case, the

seasonality of the data gradually changes over time,

those differences will be reflected in the set of seasonal
factors generated.
Some data are lost because of the moving average
techniques employed, but factors for the last months
of the data span are estimated by extending the data
into the future using averages of the last four sea­
sonal-irregular ratios available for each month as new
input for the moving average.
The ratios are then adjusted so that they will total
12 for the year and the final seasonal factors are
finally determined. A s soon as the factors are ready,
they are divided into the original raw data and,
voila, a final seasonally adjusted series results. Seasonals are also estimated for a year ahead by adding
half the change in a particular month’s factor from
the preceding year to the current factor.
Some Commentary A s is evident from the ou t­
line of the seasonal adjustment procedure given
above, the adjustment of a time series for seasonal
variation is not a simple task. The supplying of
accurate prior adjustment factors requires a great
deal of sophistication and familiarity with the data.
Often, past data are simply not available to indicate
how the figures behaved when they were affected
by a strike, a tornado, or any of a myriad of non­
recurring events. Even if historical information is
available, one might always question whether the
data is affected in the same way now as then. The
effect of a strike on production of automobiles, for
example, depends upon such factors as how well and
for what length of time the strike was anticipated
by the industry.
For these reasons, the X - l l program normally
isolates data affected by events such as these and
ignores such data in subsequent calculations.
It
should be made clear that in so doing X - l l usually
provides quite serviceable seasonally adjusted data.
Analysts have come to rely on these data to help
them in their business and economic research. A s
time passes and more basic data of better quality
become available, the seasonal adjusted data can be
expected to give ever more accurate portrayals of the
basic nonseasonal tendencies and patterns in eco­
nomic and business statistics.
William E. Cullison

REFERENCES
Shiskin, Julius, Introductory Comments, Seasonal A d ju stm en t Sem inar, N ational A ssocia tion o f B usiness E con om ists,
M a rch 10, 1969.
Som er, M o rto n , The X - l l Variant of the Census Method I I Seasonal Adjustment Program— Description of the Standard
Output, N ational A s s o cia tio n o f Business E c o n o m is ts ’ Sem inar on Seasonal A d ju stm en ts, P h iladelphia, M a rch 10, 1969.
U . S. B ureau o f the Census. The X - l l Variant of the Census Method II Seasonal Adjustment Program, Technical
Paper No. 15, (1967 revision), U . S. G overn m en t P rin tin g O ffice , W a sh in g ton , D . C., 1967.


8


CONTAINER SHIPPING
The scramble into container shipping over the past
five years is spawning broad changes in sea trans­
portation with effects that reach beyond the docks
to the railroad and trucking industries. Containerization is the shipping industry’s version of mass pro­
duction. In sharp contrast to conventional shipping
techniques under which cargo has to be repacked
for each different carrier along its route, a packed
container can be shuttled back and forth between
carriers with no handling of its interior cargo.
Moreover, containers can be maneuvered quickly and
efficiently by machines, avoiding the labor intensive
methods of moving break-bulk cargo.
Halting Beginnings Container shipping in the
United States is approximately 14 years old. The
Pan-Atlantic Steamship Company began carrying
semi-trailers to Atlantic Coast ports between New
Y ork and Puerto Rico in 1956 on modified T-2
tankers.
A t that time trailers were also being
shipped along the Pacific Coast from California to
Alaska and from California to Hawaii. The United
States Arm y used containers to ship equipment to
Korea in the 1950’s and currently uses them in
shipping supplies to Viet Nam. Container shipping
thrived in U. S. coastal trade for a decade before
entering international trade. The transatlantic ship­
ping lines knew the merits of containerization but
hesitated to sink large amounts of capital in new
ships and containers until goaded by the press of
competition.
The fillip came in 1965 when Sea-Land Services,
formerly the Pan-Atlantic Steamship Company, an­
nounced that it would initiate container shipping in
the lucrative North Atlantic trade which includes
a high proportion of containerizable cargo. Unwill­
ing to lose part of their market, the major Atlantic
shipping lines rapidly began to plan for containerized
operations.1
Container Characteristics T he container itself is
a large rectangular box made of steel, aluminum, or
plywood. It must be strong enough to withstand
heavy weather at sea and rough handling on land
while bearing heavy loads. Containers are usually
eight feet high, eight feet wide, and come in lengths
of multiples of ten feet. Those most commonly used
are either 20 or 40 feet in length. The 40-foot con­
tainer is particularly popular with truckers because

1 This article is indebted to the Journal o f Commerce for information
contained in its excellent series on containerization.




it enables them to take maximum advantage of their
carrying capacity.
Aluminum is popular in container construction
because it is approximately 20 per cent lighter than
other container materials. For truckers who must
observe over-the-road weight limits, this reduction
in non-productive weight is an important source of
profits.
Many ingenious methods have been devised to get
cargoes into containers and stabilize them for heaving
sea voyages. Some containers open at the end, some
at the side, and others from the top. They are de­
signed for a variety of cargoes: refrigerated perish­
ables, bulk liquid, dry bulk, pressure tanks, and dry
general cargo.
The container shipment is a better security risk
than break-bulk cargo from the insurance under­
writer’s viewpoint. Once it is packed and locked at
the point of departure, the container is not opened
until it reaches its final destination, making it at­
tractive for shipping high value cargo. However, in
the event one exporter is unable to fill a container, a
freight forwarder can consolidate his shipment with
some other cargo to take full advantage of the space.
Port Facilities The challenge of containerization
was no less urgent for Europeans and American
ports that depended on the Atlantic commerce. They
had to build modern terminals in preparation for the
container traffic and to avoid losing business to com ­
petitors. The special crane which hoists containers to
and from the ship costs $1 million. The specialized,
berth, including installation of equipment, pile driv­
ing, and paving costs approximately $6 million. In
addition, a large backup area adjacent to the berth
is needed for the temporary storage and sorting of
containers in transit.
The Port of New Y ork was well ahead of other
East Coast ports when the race began in 1965. In
that year one container terminal on Newark Bay had
been completed and others came into service soon
thereafter.
For the small port with light traffic, the question
of developing for container freight poses a serious
dilemma. Should it forego the development and risk
losing part of its business to com petitors; or should
it invest and risk not being able to provide enough
cargo to lure the containerships ? Many small ports
are hedging the bet by only adding to their breakbulk capacity one container berth, or a larger crane.
W ith this strategy they can continue to handle
9

break-bulk shipments and also be prepared for oc­
casional container traffic.

commodate additional containers stacked six high A
second type of containership which is popular is the

Terminal facilities for handling containers in the

roll-on roll-off ship, which has ramps so that trailers

Fifth Federal Reserve District are located in Balti­

can be driven onto the ship in the same way a

more, Hampton Roads, Morehead City, Wilmington,

ferry is loaded.

and Charleston.

The Dundalk Marine Terminal in

Baltimore Harbor has two container berths in opera­
tion with five more to be in service by 1976. Hamp­
ton Roads has four container berths serviced by an
equal number of rail-based cranes and is the first
port where trains can come directly to the terminal
to transfer containers. Large gantry cranes are used
to move containers on the open docks at Morehead
City and W ilm ington; Charleston will have a new
container terminal in operation by February 1971.
Impact on Ship Design

Impact on Labor Demand

T h e job

o f lon g­

shoreman has traditionally been to sort, load, un­
load, and store the cargo either in the hold or on
dock.

Approximately 100 men must work for a

week both to unload and to load the conventional
cargo ship.

In contrast, 40 men can do the same

job in only 24 hours using a large container crane.
One crane can move a 35-ton container from ship to
shore m 2 l 2 minutes— a job that would require 20
/
man hours if handled as break-bulk.2

The shipping com pany

The changes in methods of cargo handling brought

that undertakes container operations can remodel a

about by containerized operations have caused fric­

conventional cargo ship or buy a newly designed con-

tion with longshoremen’s unions whose jobs are

tainership.

threatened.

Either approach is costly.

The U. S.

Instead of one gang per hatch on a 5-7

Lines’s American Legion, for example, cost $18 mil­

hatch ship, only two gangs are needed. Furthermore,

lion.

the packing of cargo for the ocean voyage which

It is 700 feet long, can carry over 1,000 con­

tainers, and has a service speed of 24 knots.

Such

longshoremen have traditionally performed on the

ships are constructed with large hatches through

docks can now be done hundreds of miles inland from

which containers, guided by rails, can be lowered into
the hold. The rails also secure the containers during
the voyage.

The decks are wide enough to ac­


10


2 Goldberg, Joseph P ., “ Containerization as Force for Change on the
W aterfron t,” M onthly Labor R eview , Vol. 91, January 1968, p. 8.

Photograph courtesy o f Virginia P ort A uthority.

port. The inland exporter can pack his container,
which is then brought via truck or train to the dock
and hoisted aboard with longshoremen never touch­
ing the interior cargo.
The shipping companies and longshoremen’s unions
worked out an agreement in the early sixties whereby
the companies would pay into a union fund part of
the savings they made by using containers. In re­
turn, the unions consented to reductions in the num­
ber of gangs per ship and agreed to move the con­
tainers. A further aspect of the agreement was a
penalty to be paid by the companies for any con­
tainers consolidated within a 50-mile radius of the
port by non-members of the longshoremen’s unions.
Turnaround Time T h e conventional cargo ship
is in port much of the time, where it earns nothing.
Revenues are lost by lengthy stays, and the effective­
ness of improved ship power plants and navigational
equipment are reduced.
By comparison, the containership has a fast turnaround time which makes
it more desirable from the ship owner’s point of view.
The containerships which move Army supplies to
Viet Nam are in and out of Da Nang harbor in less
than 24 hours, an increase in speed which has en­
abled the Department of

Defense to

reduce air

freight contracts and return some old freighter ships
to moth balls.

Another example of faster turn­

around is given in the freighter trip from Japan to
the W est Coast of the United States.

Twenty-five

days were required for a conventional cargo ship to
make the trip, including 12 days in port.

Container­

tainers from small ports and then transfer them to
a containership at the major port of call.
Cutting the Red Tape T h e advent of containerization has had considerable impact on the
paper work attending cargo shipment. Under tradi­
tional shipping procedures separate sets of papers
had to be prepared each time the cargo was received
by a different carrier. This has been a source of
considerable delay. The primary advantage of con­
tainerized shipments is rapid delivery, and to maxi­
mize this advantage advocates of containerization are
trying to reduce the paper work. Shipping rep­
resentatives from many countries have been meeting
in Geneva to develop a simplified piece of paper
which would replace the Bill of Lading in that it
would be written out when the container was packed
and would serve the needs of all through carriers, in­
surers, and bankers.
The new method of shipping presents challenges
to the Customs Bureaus in many countries with con­
tainer ports. In the past incoming shipments have
been inspected at the port of entry, a procedure
which will probably change where containers are not
opened until traveling many miles inland from port.
T o avoid hindering the rapid travel of containers,
the United States Customs Bureau is considering the
establishment of regional inspection stations near
heavy users of containers. Provisions are also being
made for the temporary duty-free entry of containers
which are to be re-exported within three months.
Freer movement of containers without inspection or
the posting of bonds has also been arranged.

ships make the journey in 15 days of which only

One

Summary D espite its slow start container ship­
ping has caught on in transoceanic commerce. Since
1965 ports and shipping companies on all major
ocean trade routes have been preparing for container
traffic. The facility with which containers can be
m oved b y m achines permits speedy delivery o f
cargoes and provides their major advantage over
traditional methods of handling cargoes. The rapid
transit of the container has meant increased pro­
ductivity for all carriers involved and less waiting
for consignees.

approach is to send a ship around to collect con­

R obert W . Chamberlin

three are spent in port.
The shipping lines would like to take advantage
of the rapid turnaround and increase the utilization
of their container vessels by reducing the number of
port calls.

Then while the conventional ships chug

in and out of ports looking for cargo, the containership would get a full load on one or two calls. W ith
this goal in mind, some containership operators are
waging a campaign to have inland traffic channeled
into a small number of advantageous ports.

IN S T R U M E N T S O F T H E M O N E Y M A R K E T
A revised edition of Instruments of the M oney Market is now available free of charge from
this Bank. This booklet describes the principal money market instruments and the markets in which
they are traded.




11


Federal Reserve Bank of St. Louis, One Federal Reserve Bank Plaza, St. Louis, MO 63102