temperature are noted by equal volumes of expansion with an arbitrary zero point

agreed on for the scale. The temperature differential is divided between freezing and
boiling into 100 degrees, with the freezing point arbitrarily set at zero degrees. If the
temperature of two different rooms is measured with a Celsius thermometer and gives
readings of 22 degrees and 30 degrees, we can say not only that the second room is
hotter, but also that it is 8 degrees higher in temperature. The differences between the
numbers can be translated directly to represent the differences in the characteristic of
the objects.
The weakness of the interval scale is that the zero point is arbitrarily established.
For example, suppose we were to measure the height of a group of men on an interval
scale and assign a number to each according to his height with respect to the average
of the group. The average represents the zero point on the scale. If A is 3 centimetres
above the average, then we would assign him the number +3; and if B is 5 centimetres
below the average, then we would assign him the number -5. On this scale, we do not
!mow how tall A or B is in an absolute (actual) sense. B may be the shortest man in the
group, but the group may consist of tall basketball players.
Mattessich mentions standard cost accounting as one example where the interval scale
is used in accounting. 8 The standard may be based on theoretical, average, practical or
normal performance. Because the choice is more or less arbitrary, the calculation of
standards and variances generates an interval scale. If the variance is zero, this signifies
neutrality, but this point is arbitrarily selected.

Ratio scale
A ratio scale is one where:
• the rank order of the objects or events with respect to a given property is known
• the intervals between the objects are equal and are known
0 a unique origin, a natural zero point, exists where the distance from it for at least one

object is known.
The ratio scale conveys the most information.
The measurement of length is a good example of a ratio scale. When A is 10 metres
long and B is 20 metres, we can say not only that B is 10 metres longer but also that it
is twice as long as A. The ratios of the numbers are also directly interpretable as ratios
of the quantities of the properly measured. Thus, it makes sense to say that A is half as
long as B or B is twice as long as A, whereas we cannot say that 40 degrees Celsius is
twice as hot as 20 degrees Celsius.
An example of the ratio scale in accounting is the use of dollars to represent cost and
value. If asset A cost $10 000 and asset B cost $20 000, we can state that B cost twice as
much as A. A natural zero point exists, because O denotes absence of cost or value, just
as O for length means no length at all.

f!l&-i PERMISSIBLE OPERATIONS OF SCALES

One reason for discussing scales is that certain mathematical applications are permissible
only for different types of scales. The ratio scale allows for all the fundamental
arithmetical operations of addition, subtraction, multiplication and division, and
also algebra, analytic geometry, calculus and statistical methods. A ratio scale remains
invariant (fixed) over all transformations when multiplied by a constant. 9 For example,
consider the following:
X'=cX

136 PART 2 Theory and accounting practice

-.
r
If X represents all the points on a given scale, and each point is multiplied by a con-
stant c, the resulting scale X' will also be a ratio scale. The reason is that the structure of
the scale is left invariant, that is:
• the rank order of the points is unchanged
• the ratios of the points are unchanged
• the zero point is unchanged.
This means that if we measured the length of a room and found it to be 400 centimetres
and then converted the 400 centimetres to 4 metres by multiplying by the constant 1/100,
we can be assured that the length of the room is unchanged, even though the number
representing length has changed. This is the same point we make in chapter 6 regarding
the conversion of historical cost of, say, $100 000 of equipment under the nominal dollar
scale to the purchasing power of the dollar scale by applying a constant, say, 120/100, to
derive $120 000. The $120 000 is still historical cost.
The invariance of a scale permits us to know the extent to which a theory or rule
remains basically the same, even though the scale is expressed in different units, such
as from centimetres to metres or from nominal dollars to constant dollars. An invariant
transformation of a ratio scale will leave intact the same general form of the relationship
of the variables.
Without invariance, it is possible to find that X is twice as long as Y when measured
in centimetres, but three times as long when measured in metres. In accounting, the
scale for current cost is variant from that of historical cost, because the attributes to
be measured are different. When machine A is measured under historical cost it may
be $90000, but when measured under current cost it may be $110000. The unit of
measure, the dollar, is used in both cases but the scales are different: they are variant.
But changing from the nominal dollar scale to the purchasing power of the dollar scale
for the same attribute (historical cost or current cost) leaves the structure invariant.
With an interval scale, not all arithmetical operations are permissible. Addition and
subtraction can be used with respect to the particular numbers on the scale as well as
the intervals. However, multiplication and division cannot be used with reference to
the particular numbers, only to the intervals. 10 The reasons are due to the conditions of
invariance. An interval scale is invariant under any linear transformation of the form:
X'=cX+b
The transformation of one interval scale for measuring a specific property to another
interval scale for measuring the same property is made by multiplying each point of the
first scale X by a constant c and adding to it a constant b. The reason for b is that there
is no absolute zero point on an interval scale. Par example, to transform from Celsius
temperature to Fahrenheit temperature, we multiply each degree by 9/5 and add 32.
The 9/5 is used because the Celsius scale has 100 degrees as opposed to 180 degrees for
Fahrenheit, and 32 is added because that is the freezing point for the latter scale.
The conditions of invariance show that we can multiply and divide with respect to the
intervals, but these arithmetical operations cannot be used for the particular numbers
on the scale. To illustrate, consider the following transformation:
X'=X+ 10
Consider the objects on points 3 and 6 on scale X. Transforming to scale X', we now
have 13 and 16. The ratio of 13 to 16 is not the same as the ratio of 3 to 6 because
of the addition of the constant. Multiplication and division (i.e. ratios) are therefore
not permissible for the particular numbers. Thus, if Robyn receives 90 points on her
accounting exam and Maria receives 45 points, we cannot say that Robyn knows twice
as much as Maria regarding the subject matter of the exam. The reason is that there is

CHAPTER 5 Measurement theory 137

-.
 1
no natural zero poinl for the exam for 'no knowledge'. Even if a student receives 0' on
the exam, we cannot say that he or she has no knowledge of the subject matter. In this
example, what we can say is that Robyn passed the exam and Maria failed the exam, but
we cannot infer comparatively the amount of knowledge to the numbers. Likewise, if
the quantity variance is $5000 favourable, as opposed to the previous month's variance
of $10 000 favourable, we cannot say that the use of materials this month is only half
as efficient as in the preceding month.
With ordinal scales, none of the arithmetical operations can be used. We cannot add,
subtract, multiply or divide the numbers or the intervals on the scale. Ordinal scales,
therefore, convey limited information.

(mlJ-, TYPES OF MEASUREMENT

As we said earlier, the measurement process is similar to the scientific approach of
theory construction and testing. Our discussion of scales relates to the question of the
construction and implementation of theory. There must be a rule to assign numbers
before there can be measurement. This rule is usually a set of operations which must be
devised for the given task. The formulation of the rule gives rise to a scale. Measurement
can be made only on a scale.
The question of testing of a theory relates to the question of the different kinds of
measurement. Campbell mentioned two kinds: fundamental and derived." Recall
that in Campbell's definition of measurement he stated that the numbers are assigned
according to the 'laws' governing the property. For Campbell, measurements can
take place only when there are confirmed empirical theories (laws) to support the
measurements. A further type of measurement, fiat measurement, was mentioned by
Torgerson as being additional to the fundamental and derived measurements discussed
by Campbell. All three measurements are discussed on the following pages.

Fundamental measurements
A fundamental measurement is one where the numbers can be assigned to the property
by reference to natural laws and which does not depend on the measurement of any
other variable. Properties such as length, electrical resistance, number and volume are
fundamentally measurable. A ratio scale can be formulated for ead1 of these properties
on the basis of!aws relating the different measures (quantities) of the given property.
The interpretation of the numbers depends on the confirmed empirical theory that
governs the measurement operation.
As it turns out, the fundamental properties are additive. Because of this, it is simple
to find physical parallels to the operations of arithmetic. 12 For example, adding the
length of object X to the length of object Y is paralleled by the actual operation of
placing two straight rods end to end, with one rod having the same length as X and
the other the same length as Y. We can physically determine the total length of X and
Y. Because of these physical parallels, the scientist can simply perform the mechanical
operations of mathematics without having to conduct an experiment for length.

Derived measurements
According to Campbell, a derived measurement is one that depends on the measurement
of two or more other quantities. The measurement of density is an example. It depends
on the measurement of both mass and volume. Derived measurement operations
depend on known relationships to fundamental properties. They are based on a
confirmed empirical theory relating the given property to other properties. Mathematical

138 PART 2 Theory and accounting practice

-.