Note: Dr. Anton is partner in charge of the Accounting Concepts
Department at Haskins & Sells and is a member of the Accounting
Standards Executive Committee of the AICPA.
Dr. Anton has taught economics and accounting at
several major universities including the University of Minnesota,
the University of Washington and the University of California in
Los Angeles and Berkeley. While at Berkeley, he was chairman of
the accounting faculty for five years and Associate Dean of the
Graduate School of Business Administration from 1969 to 1972.
He has also been a visiting professor at Victoria
University, New Zealand, the University of Chicago, Fulbright Professor
in Finland, and held a Wallenberg Foundation grant at the Stockholm
School of Economics. He is the author of several books and his published
articles have appeared in leading accounting journals both here
Dr. Anton earned his bachelors and masters degrees
at UCLA and his PhD in economics from Minnesota.
The topic of the lecture comes in response to
the continuing search for more acceptable accounting models, particularly
in the wake of a growing perceived dissatisfaction with the historical
cost model. A plethora of models have emerged. The more narrowly
drawn models lack applicability while those which are more general
lose reliability. This and other observations lead to a conclusion
that no information model can ever be developed that would be totally
responsive. Conven-tions must be agreed upon from among a family
of uses and purposes. Hopefully these will serve for a significant
period of time.]
I am pleased and honored to be here with you tonight honoring our
good friend, Emanuel Saxe. Actually, Mannie is the one that is distinguished,
and not the lectures, at least not my own. Mannie, my very best wishes
to you. I also feel a certain kinship with this school, one that I
discovered rather lately -- the school and I were launched in the
same year. And, despite the fact that we have encountered some rough
seas, we're still sailing along. I'm also glad to be back here in
an academic atmosphere where it is well understood that whatever I
have to say here is my own personal view and does not necessarily
reflect any position of Haskins & Sells.
In the search for a conceptual framework, the construction of models
is not sufficient, let alone models that are bastardized by incorporating
elements from a number of theories. If one is going to construct a
model, one should construct a nice neat model that stands on its own.
Then only the premises can be attacked. What then is my thesis? We
have been in search of a cohesive coordinated body of theory for a
long time, but that search has been accentuated within recent years
by widespread dissatisfaction with the historical cost model. That
dissatisfaction, as you well know, is widely shared by academics,
financial executives, independent accountants, economists, Congressmen
and others. Obviously, dissatisfaction with the historical cost model
has been fueled by world wide inflation, investor consumerism and
an accelerated rate of change.
As a result we have seen a great plethora of models -- one of which
was given at the last Saxe lecture. Each model has strong proponents;
and each model, frankly, is in search of a measurement system that
makes sense. However, each model is also flawed (any model is flawed
for that matter) because of a dilemma. That is: The narrower the basic
premises and objectives, the tighter and more reliable that model
will be; but then it suffers from narrowness in applicability. Those
of you who were here at the last lecture will attest that Dr. Chambers
has a very nice model. Frankly, given his premises, there is nothing
you could attack. However, his premises were flawed in that the only
things he was interested in, essentially, were cash flow and current
liquidity, thus his concentration on current exit prices. His model,
therefore, avoided or negated all other kinds of considerations, including
investment possibilities. Well, if one broadens the premises and objectives,
then a loss in general applicability will result and the model becomes
less useful in making individual decisions. Therefore, it is not surprising
that most of the questions addressed to Dr. Chambers had to do with
the fact that the audience didn't like his premises, or didn't like
his view of what the output from that decision model would be. Dr.
Chambers, of course, found fault with the conventional historical
cost model because it did not give him certain things that he wanted
-- such as being able to predict that no cash could be obtained from
a half-built work-in process inventory.
This general observation should not really surprise us, because
thorough study of measurement and information theory leads to the
conclusion that technically, let alone semantically or at the influence
level, no information model can ever be developed that would be totally
responsive. That's it, we're just not going to get there! Therefore,
conventions (compromises if you wish) must be agreed upon from among
a family of uses and purposes. Hopefully, these will serve satisfactorily
at least for a significant period of time -- a time during which those
measurement standards can be held constant, before enough people get
dissatisfied with it and then want to go on to something else. Therefore,
what we need is, frankly, arbitrary. No academic likes to hear that
word, but we do have to have arbitrary conventions about users, about
their utilities, about their decision functions, and about the actions
users need to make. Based on these compromises certain information
is required of the accounting model, then we have to make decisions
as to how we can use the information, and so forth.
These conventional agreements, of course, will involve choices from
different theories because general acceptance involves people who
may agree about some things only if others compromise their positions.
Some of these compromises, incidentally, would have to be political.
Also, a few nonaccountants have been watching the current scene, with
the result that some politics have entered into decisions about accounting
rules and standards. Those choices, however, have to have certain
good qualities. One essential quality, and one that we ought to be
striving for constantly, is that the output of the accounting model
is capable of being transformed by the individual users into information
useful to them in unique circumstances. That's my thesis, in brief.
However, I want to support these assertions by reference to some of
my previous work as well as some of the work that others have done.
While I caution against the optimistic view that any model or conceptual
frame-work will be a panacea and solve our problems "forever,"
I do want to offer a systematic way of analyzing accounting information
models in such a way that we can evaluate these models' strength and
weaknesses. But, more importantly, regardless of the model, that we
can determine what kinds of assumptions about externalities, that
is factors external to the model, need to be made in order to make
a model work, in order to make the model viable, and if we can, to
determine what the cost of such assumptions might be. In order to
do so, and I want to be just a bit tight in the formulation, I will
draw on certain aspects of other related disciplines -- measurement
theory, communication theory, language, information economics and
behavioral theory. In brief, I don't want to stop at any arbitrary
"accounting" boundary, conventional or otherwise. I'm also
going to make assertions that can't be supported here, today, but
which are supportable either by my own writings or others'. What I
want to do here is simply to sketch a blueprint of the kind of relevant
questions that can be raised in model building; and perhaps give some
direction to our study of the conceptual framework of accounting.
First, a brief word about measurement and five points to keep in
mind for later use: One is that we measure only because we want to
be precise about something in a specific use. My favorite example
is of a yard. For my general purposes half my arm's span is within
half an inch of a yard. Generally, that is enough precision for what
I need. However, a standard is set under conventionally specified
but rigid conditions, measured in such a manner that the standard
will be useful in some other totally different circumstance or environment
from that in which the standard is set. The standard yard, believe
it or not, is a unique metal rod held under specific temperature and
atmospheric control at the Bureau of National Standards, but we need
to use the idea of the length of the specific metal rod in order to
measure under other conditions.
Any measurement is an agreed standard. It's "manufactured"
if you wish. The need for the precision, however, is set by how we
use that particular measure.
Secondly, we have to set the state for the measure, that is
we have to set the environment under which that measurement would
take place. Obviously, if I'm talking about a yard I can't go and
get that standard rod out of its vacuum case, but I do have certain
transformations or adaptation rules that enable me to make a measurement
in a particular case. In other words, we need first a standard, then
a specific environment that we want to measure, and thirdly, we need
to develop some adjustment rules. Those adjustment rules are important
because they are devised to make the standard applicable in a wide
variety of circumstances. We can then use the measure, not only here
and now, but tomorrow or years after, and still be able to bridge
both time and space in making the measure.
The fourth step is one that generally everybody considers to be the
measure but is only one step in the whole process: A number or other
symbol is applied. This is called numeration. Finally, we have to
use the symbol in a specific action. Some information theorists, among
them Henri Theil, go so far as to say that no information exists unless
that information influences a desired action. Keep those conditions
of measurement in mind as I develop my thesis.
Next, we have to know something about language. Language in some
ways is a kind of a measure, but in other ways it is different. Language
is a general code that enables us to communicate adequately from one
person to another -- again bridging time and space. For example, even
though I read it poorly, I can read Chaucer -- who in turn never heard
of me or America. And, the other day I met a Japanese professor who
studied flow of funds from my book 15 years ago. Language bridges
time and space, and obviously, that is also what is needed in the
measurement process. The aim of language is generality, again to improve
the chances of utilization. We want a language broad enough so that
we can communicate with many people and, at different times -- which
also is one of the aims of accounting. We want to communicate broadly,
but as the language or code becomes more general the dilemma arises:
It loses its ability for fine distinction. For example, all of us
who have studied economics know that even though economics texts appear
to be written in English, that those economics books are really written
in Economese. We also know that the man on the street does not make
any sense of them. (Perhaps they do not make any sense to economists
either, but that's another story.) At any rate, as the code becomes
more general, it loses the characteristic of being useful in precise
situations. As special cases recur, specialists need to be more precise,
and before long a metalanguage develops.
To a certain extent, all professionals and scientists use metalanguages
to communicate better with each other (and sometimes, it is claimed,
to keep the rest of the population ignorant). If metalanguages develop,
then the general audience loses the capacity to understand what they
are saying. That is a problem that constantly has to be faced and
will continue ad infinitum. Society keeps developing various
sets of languages and metalanguages for special purposes. In order
to understand the various languages we build redundancy into the system.
A dictionary, for example, helps by defining some words that are unknown
in terms (words) of others that are known. Using a dictionary permits
one to be more precise, but it takes time. At the same time, of course,
some learning takes place and if enough learning is achieved, the
person is no longer in the "general" audience but has become
part of the "precise" audience.
The important thing to stress here is that many people who read economics
or accounting or other special books will look at the English words
and believe they know what they are reading, but without actually
knowing. That extends also to such things as financial statements,
and that extension requires us to guard very carefully against superstitions.
A superstition is simply perceiving something falsely. In the situation
above, what is written may not be what one perceives in the reading.
So, guard against superstitions both in measurement and in language.
The next step is tougher because it introduces a little-known concept,
entropy -- that is the measure of information content in mathematical
communication theory. That basic premise is used to measure the capacity
of the information source, the communication channel, or the communication
receiver. Simply stated, experts in communication theory, information
scientists if you wish, have been able to quantify information such
that they are able to make general decisions about the information
system. That is information is capable of being measured in a technical
sense; we have the means for measuring the amount of information in
the source, the capacity of any communication channel to transmit
that information, and also the capacity of the receiver to receive
and perceive such information. To parallel the accounting system,
the information source would be the business environment, the information
communication channel would be the accounting and reporting system
and the receiver would be the user of the reports.
In physical and statistical terms, entropy is measurable. Information
theory is well developed and has many applications in a great variety
of areas. What we need to know here is that entropy is a concept that
is measurable, and that any message in any information system (such
as accounting) has limits on what can technically be achieved. In
other words, there are physical limitations in these areas. The capacity
of the channel (that is, the accounting system) and the capacity of
the receiver (the user) set limits, when considered in relation to
the magnitude of the information source, that cannot be overcome --
they just simply cannot be overcome.
Unfortunately, those capacity relationships result not merely in
selective loss of information, but also in increased confusion. Three
examples that one can relate to are: (1) If a TV channel gets too
strong a signal, or one signal too many for its capacity, one does
not get part of the picture, most of the time the result is what we
call "TV snow." I know you all have experienced that. The TV snow
that you see on your TV receiver is the result of too much signal
for its capacity. Secondly, and I'm sorry it didn't snow today, you
have heard of "snow jobs " -- a use totally unrelated to
TV snow, but representing the same phenomenon. A snow job results
when somebody feeds you too much information, deliberately. In that
case, if you don't understand a thing of what's going on, and if you
want to appear that you do, then you use the "superstition"
and "get snowed." A third sample, is this particular part
of the lecture.
The mathematical theory of communication also defines conditions
as to what in, and what goes out, of the system. Encoding is needed
to input information into a system, and decoding is needed to output
information. Unless one knows what code was used in encoding, one
cannot decode. Unless the conditions or measures of the input are
known (I'm going to talk about standards a little later), there's
no way that one can know (cognitively) what comes out. If one does
use that output without knowing the code, again one will be using
it as a superstition rather than as a fact. There is ample literature
in this area. Other aspects of information are also relevant to what
I have to say, such as costs of information, the economics of information,
the relationship of the cost of information to outcomes, etc., but
these aspects are not central here.
Let's see if I can summarize the elements. In the simplest terms,
first I have talked about the existence of some state of the world
-- the environment (source) as it relates to the activities of the
business unit. Secondly, in order to make use of that state of the
world, and that is really what accounting is all about, we want to
transform that real state of the world into abstract symbols -- the
data that, can be put into an information system and which later can
be output. In order to do that, a decision must be made as to which
of the activities and states of the environment are to be put into
the particular system. Immediately, we need selection standards. In
a sense the perfect accounting system, of course, is the whole world.
(There are some models, incidentally, that cannot be modeled -- that
is the only model possible is the real thing itself.) Third, the selected
items must be encoded into a suitable language that is susceptible
to measurement (data bits are used in information theory); recall,
assignment of a number in the discussion on measurement. These, then,
must be structured in such a way that the channel, here the accounting
system, can be consistent with the capacity requirements of both the
environment (source) and the user. In accounting we use a simple two-bit
basic code -- zero or one, debit or credit, in or out, negative or
positive -- but those two little bits have been structured in such
a way that significant parts of the whole world are encompassed. You
know, that's really a marvelous thing.
Processing and other data manipulations take place within the system.
Here those who want to study information seriously will study cost
of information and information economics. There are costs of summation,
there are costs of mathematical processing -- even arithmetic, there
are costs of classification, there are cost of selection, and there
are costs of aggregation. These manipulations are done for a variety
of reasons, but mostly to enable more efficient output from the system.
A partial state of the world is now encompassed in a little black
box, but it is useless unless something can be retrieved from that
blackbox. What is needed? Another policy decision, this time selection
criteria are required to guide us as to what should be retrieved,
and how and when. That selection results in some form of output which,
in turn, must be decoded in order to make the abstract data relate
to real state of the world considerations. Finally, given that an
action relative to that information must be taken in order to affect
the environment, the cycle is complete and another starts anew.
We can now clearly focus on many problems. If, as many information
theorists claim, no information exists unless it can affect a decision,
and if, in order to construct a system, the actions that need to be
influenced must be known (as well as who's going to take the action,
and when and how), these output considerations must be built into
the input considerations. However, general systems must be built because
we do not know who the decision makers are going to be, or at what
time they are going to take the action or under what circumstances.
A general system obviously cannot accomplish the task perfectly, so
certain kinds of compromises are needed. One thing is clear, actions
need to be known at the input state.
Secondly, even one individual will have many activity decisions,
not one but many, and individual input and output selection standards
may very well conflict. There has been some work done on conflicting
objectives and conflicting standards (by Balderston, for one), but
we're not very far along in that area. At best, a system tailored
to one individual may be designed, let's say, by asking the individual
to state his preferences and behavioral assumptions and working on
them. But getting an individual to state preferences frequently results
in erroneous signals because individuals do not communicate very well.
Therefore, inferences from observed actions are frequently drawn as
to what the preferences really are, etc. Even if we can determine
individual preferences at a given time, we know that any individual
can change them easily; sometimes without rationality or apparent
rationality or explainable circumstances.
Thirdly, to have generality many users must be reached. Combinations
and permutations of desired actions then become very large. If one
individual has many actions that he can take over his life-time, you
can imagine the number that many individuals will have. Therefore,
to achieve any kind of rationality we must reach agreement by convention.
(No academic likes to say that because it sounds too much like the
opprobrius "real world" kind of thing -- the "other
world.") Agreement by convention must be achieved so as to serve
potential users and their potential actions as they arise. That agreement
has a side characteristic: If one agrees on a set of actions or a
set of standards, automatically all others are excluded. That's kind
of tough. You can't have a nice, narrow, precise model and come out
with nice, narrow, precise answers without excluding and ignoring
all the other things that happen. The potential users, then, really
have to be considered "standard users" in the true sense
of the word standard. The actions that they take, or the decision
functions ascribed to them, will have to be standard actions or decisions.
Superstitions must be avoided, otherwise, again we're lost. Following
measurement theory, if we have standard users, standard activities,
standard functional relations, etc., then we must have some means
of following the rules of measurement. That is, we must be able to
adapt from the standard case to the unique case in the unique environmental
situation. Frankly, we have been weak in pursuing that kind of information
both as academics and as professional people.
Given conventional agreement as to users and actions, standards for
processing within the accounting system, and standards for input selection
and measurement will also require widespread agreement. As we all
know people don't agree for very long.
Conventions are transitory. It is no real surprise that we may be
unhappy with the historical cost model. That model, although it appears
as if it has been with us forever, has really not been dominant except
since the 1930s. Prior to that, we had current cost models, single
entry, and mixtures of different things. When conditions change again,
we'll undoubtedly become dissatisfied with whatever model is agreed
upon now. As the state of the world changes, as economic payoffs change,
the relative power of different users changes, and then there will
be pressure to change the input and the output measurement standards
-- and especially to change the focus as to what should represent
standard users and standard actions.
Well, I've been rather abstract, and now to pin it down. Let's see
if I can make it a little more relevant to today's conditions. First
and easiest, there has been some talk of changing the basic structure
-- that is double entry bookkeeping. Double entry has been known for
over 600 years; Paciolo wrote in 1492 but by that time it had been
used sporadically for 150 years. However, double entry did not come
into prominence until the second half of the last century. In fact,
there are some classic 19th century books, debating the issue -- I
remember one by Cronhelm, with the interesting title Single Entry
by Double. At any rate, since the argument in the mid-1800s as
to whether or not we needed double entry, there has not been too much
talk about doing away with double entry. There has been some talk
about adopting quadruple entry; German scholars have advocated a form
of quadruple entry, and others, especially computer experts, have
been talking about n tuple entry for some time. No one has
done much about those speculations, so abandoning double entry is
not too serious an issue. Therefore, we cannot expect channel capacity
to increase very much from that direction. More serious attention
has been given recently, and you might look for that kind of thing
in the conceptual framework, as to whether or not there should be
articulation between the income statement and the balance sheet. As
viewed by some, it could take the form of a step back toward single
entry. However, the problems that are posed by articulation can probably
be solved through other means.
Secondly, the state of the world certainly suggests extended pressure
on the present conventionally-agreed upon model which has been paramount
since the 1930s. Two major influences are inflation and volatility
of foreign exchange rates, and heavier taxation with its consequent
impact on capital formation. Frankly, it is too early to tell whether
any agreement can be reached on any one alternative to the present
model. The FASB's general purchasing power approach and the SEC experiment
with replacement cost information have been more cursed than blessed
and seem to have created more problems than solutions. On top of that,
no general standards are available for either one. Other models posed
by individuals have not had much general support. In brief, they either
narrow the objective too much as in Dr. Chambers, or are essentially
nonmeasurable as in Edwards and Bell. Given the institutional penetration
of the present model, that is historical cost permeates practically
everything we do in the business world today, one wonders whether
a change from that model will ever be economically feasible short
of runaway inflation.
Thirdly, analysis of users (and recall that current focus is on users
not on input, although past models have been built on input characteristics,
i.e., what could be recorded, rather than on output characteristics,
i.e., what information is needed in decision-making) indicates there
are some marked changes taking place relative to the users themselves.
There is less personal control by individual shareholders. We are
faced to a considerable extent by, not even absentee ownership, but
by stockholders that don't behave as owners. To a considerable extent
stockholders have been atomized. The major concern of stockholders
appears to be quite changed from what it was 20 or 25 years ago. There
has been a trend towards being in and out of the market, towards looking
at one's portfolio, and away from either loyalty to a company or keeping
tabs on one single company over a long time. Secondly, the impact
of the institutional investors has not yet been comprehended. The
influence of these institutional investors, such as trusts, pension
funds and ESOP's, etc is growing. As some of you may have read recently,
Peter Drucker is turning his attention to that area and will probably
popularize it. Drucker forsees some drastic social changes as a result
of that trend, and obviously, he will indicate paths for other students
to study in greater detail.
The "transient" investor has tended to focus on current
results rather than on long term averages. The emphasis on current
results which is reflected in FASB statements, obviously, has been
accentuated by litigation and by significant settlements. You can't
have it both ways, you can't have emphasis on cash flow to the current
stockholder and also look at trends that erase abberations over time.
You can't both have current emphasis on depiction of the economic
impact of foreign exchange fluctuations and volatility such as we
have had recently, and at the same time say, "Let's look currently
at the trend of the earnings and not on the impact of those fluctuations."
That's the whole problem with FASB Statement No. 8: some people are
still looking at the financial statements having one objective in
mind, while the financial statements are based on a premise going
in other direction.
Recent academic interest, especially from the finance field, has
focused on efficient markets and portfolio theory. The strength of
such influence, and its impact on conventional agreements about our
accounting models, is not yet clear. In terms of the above analysis,
in order for efficient markets hypothesis to be important, users would
need to desire changes in market prices as the action. That's what
those models are measuring, impact on changes in market prices as
opposed to having information to help determine whether to hold, buy
or sell, or indeed whether owners would be better off or not in the
long run. Market behavior would have to be conventionally accepted
as an accounting goal.
Similarly, George Benston's hypothesis that SEC control is ineffective,
and perhaps not needed, is based on a premise (perhaps warranted,
perhaps not) of a desired market result -- that is, the impact of
the SEC on market prices. A predictive market view of accounting could
be constructed if such information were required, much as SEC control
is required whether or not the market follows. So again the question
is whether conventional agreement may be reached, and whether these
relatively narrow views could or should be paramount.
Just a few words about the Objectives Study, and what's likely to
be adopted. Most of us have seen the FASB's draft on the Tentative
Conclusions. We should look at those objectives to see whether they
are likely to lead to measurement standards that, though general,
are susceptible to adaptation to individual unique use. We can focus
on two objectives that are likely to be adopted by the FASB. One is
that information is useful to present or potential investors and creditors
in making rational investment decisions. Here the FASB is implicitly
addressing a general set of investors and creditors. The narrowing
down to a set of "standard" investors, including a set of
"standard" creditors, and also some sub-set of "standard"
users that may make rational investment decisions. Adopting that objective
may lead us in paths that will help us develop models -- asserting
the objective by itself does not.
Secondly, it's clear that the FASB will assert that these users need
information to help them assess the prospect of receiving cash from
dividends, or interest, and from the proceeds of sale or redemption
of the security. Primarily, they propose to do that by evaluating
the enterprise's ability to obtain cash through operations and financing
activities. In turn, the prospects of the user receiving such cash
flows will also be affected by the perceptions of investors and creditors
(somehow these will add-up!). Investors must determine how the enterprise's
ability to obtain (net) cash could affect market prices of the enterprise
relative to those of other enterprises. So while we have an objective
to present accounting information that would be useful in making decisions
rational decisions, we don't know what that information is.
Further, it is proposed that those rational decisions will be forthcoming
from the prospects of the enterprise being able to pay cash dividends;
and also, of course, from the stock market prices upon the redemption
or the sale of the security. Notice that a transformation function
has been introduced here that is not part of accounting: it
is a transformation function that is based on the perception of investors
in judging the enterprise's ability to affect market prices of its
securities. We know next to nothing about that transformation at this
time, and there is no way accounting can comprehend it, but at least
the tentative objectives focus on it. Again output information comes
from the accounting system, which in turn comes from the real world,
representations of which must be encoded in a standard way that must
be known, so that when the output is used it cannot be thought to
represent something else.
The FASB is apparently not convinced that financial accounting can
measure an entire enterprise, nor are they convinced, on the other
hand, that cash receipts and payments during a short period such as
a year would adequately indicate whether performance is satisfactory.
To buttress the objectives the FASB has proposed the use of certain
qualitative characteristics which include relevancy and materiality,
substance over form, freedom from bias, (everybody has bias, every
measure has bias, the best we accomplish is freedom from bias given
a standard use or a standard measure) comparability, consistency,
The objectives and characteristics are still thought of in the abstract,
the test will come when attempts are made to make them concrete, and
when these characteristics and objectives conflict. There are conflicts,
frankly, between the perception of a creditor or the perception of
an investor, and both are central to the Tentative Objectives. There
is nothing in the Tentative Objectives, for example, that would help
us to determine whether the present model or any one of the alternatively
proposed current value models would be a better one over the others.
Such a choice would still be dependent on evaluating the transformation
function between reported enterprise data (Objective 3) and the individual
stockholder's ability to adapt from that to their ability to predict
receipt of cash (Objective 2), or to adapt such information to market
prices (Objective 2b). The activity model which I presented to you
in broad outlines clearly brings out those relationships.
In summary, technical constraints limit our ability to use ideal
models. Instead conventional agreements have to be reached as to standard
input, standard systems, standard output, standard users and standard
uses. A major task (which I hope all of your PhD's will attack vigorously)
is to develop and provide adaptive rules that will make accounting
results usable in individual circumstance. Any conventional system,
remember conventional means agreed upon, that does not provide standard
output that is susceptible to such adaptation would not have conventional
support for very long. The conceptual framework would be very helpful,
and I hope that it will be helpful for at least 12 years.
SELECTED QUESTIONS AND ANSWERS
Could you elaborate a little bit on the models that you are describing,
particularly the more abstract ones I missed? The reference, if there
was one, to the kind of feedback control that you would hope to find
in the information models?
I'll give you a couple of references: Jerry Feltham's book, Joe Demski's
book, O.K.? Very obviously, any model of this sort constitutes a feedback
loop; a great number of loops as a matter of fact. Although I didn't
get into such things as a channel with memory, or the structuring
of the channel and so forth, the feedback itself can help structure
the information. Obviously, these models are almost organic in the
way they behave. The mere fact that we can't get the kind of information
from the conventional accounting model that some people, not all people
but some people, want is itself fed back into the structure of the
system itself. The mere fact that we take investment decisions of
some sort and they prove wrong is also fed back to improve the system.
As to the needs of users, and user interaction, next year is the
year to look at related parties, joint ventures, Addendum to APB Opinion
No. 2, and off balance sheet financing. Perhaps 1978 will the year.
In the meantime, there are all kinds of work going on, though not
really on a concerted basis. I think we probably know far less about
the needs of the users of municipalities' statements than we know
about the need of users of the corporate information, which is little.
But it's coming. I don't know if that was responsive to your question
In terms of your comment that decision making is a political process,
as opposed to a logical process, wouldn't that be another reason why
on the agenda of Objectives the nonprofit institutions would have
to be lower on the scale? Remember back when Mayor Lindsay was mayor,
one of the rating agencies cut New York City's rating; they went up
to a Senate Committee and said, "We have to start investigating
these rating agencies." If, a rating agency cuts a rating of
a profit--making corporation, they wouldn't try to do that. They wouldn't
have any chance of succeeding. So that's one of the reasons why they
have to be on a lower scale.
Thank you. I hope that you don't go away with the impression, however,
that I said this was not a logical process. Rather, it is a political
process, but certain words in the English language have certain nuances
that I don't like, so I'm only trying to stress that we cannot construct
models that can be used simply in practice. We've got to have conventional
agreements, but these conventional agreements, I hope, will be based
on logic. Though I may not have accomplished it, I thought I had a
very logical structure: from the actions that need to be taken, to
the user, his perception, his transformation functions, back to the
output, from the output back to standard rules, to the channel, from
the channel through all machinations back to the selection process
from the real world. That is a very logical process, and I would hope
in future developments that we don't become irrational.
I use the word transformation to get a better relationship to the
standard, and adaptability rules, and so forth. Feedback is implicit
in these models, as in all information models. Some very elaborate
theorems have been developed as to how this feedback interacts, not
only in the selection process, but also in the very basic input process.
How do we select what we input into a system?
Dr. Anton, I'm concerned about your comments with respect to a nonprofit
organization, or the current emphasis being placed on municipalities
and the need for measurement and disclosures there, Would you care
to comment about the applicability of models to these nonprofit or
municipality type accounting situations?
Let me just be general and abstract to start with. I kept shifting
back between accounting models and information models, etc. because
the whole process is the same. Now the process of arriving at conventional
agreement about the different models may very well change in the future.
Up to now we have tended to stay within the same general model for
all types of entities. The Trueblood Committee did propose 12 objectives,
one of which had to do with nonprofit and governmental units. As nearly
as I can tell at this point, the FASB is not doing anything relative
to that area. They are essentially saying, if you wish, that that
standard group is a little too far removed from the others, and that
perhaps it will have to be studied in more detail later. As I recall,
in my article relative to Objectives in last year's Journal of Accountancy,
I made the point that perhaps at this time the FASB would be better
off, as far as efficiency is concerned, by not trying to build the
same model for everything. At the same time, I fully understand that
information requirements for users of information from municipalities
and other governmental units are every bit as real as investors' need
for information about a corporation. This need is becoming more and
more important. To use Tom's feedback notion, as this need is posed,
more and more attention is going to be paid to it.
As to the political nuance, any agreement, any conventional agreement,
or the process of arriving at that agreement is a political process.
I did not mean to infer that this was something that only politicians
do, we're all in a political process of some sort when we try to arrive
at conventional agreements that in turn may become standards.
Dr. Anton, would you comment on the work of the Moss committee, especially
as it relates to the continued role of the FASB in the development
of accounting principles?
Well, never having spoken to Mr. Moss, although he's a fellow Californian,
I don't really understand, that's a good word, I'm ignorant of what
the particular process ... of what his particular objectives are --
whether it was an in-progress report or whether he had other aims
in mind. I do think, very honestly, that inspection of the whole process
is warranted. We're all doing it; we're all in one way or another
working with them.
Frankly, without having a lot more information about the Moss report,
what I could say probably would be a superstition. I do have great
faith in people, and I think that in the long run any investigation,
I guess that's what they call it since they're an oversight committee
for the SEC, is bound to look at the constructive things as well as
those that need mending. So I'm not pessimistic at all about those
Dr. Anton, just to support your point about the political process,
and to defend the concept of the political process against those who
think it in nonquantitative, I'd like to make two suggestions. One,
the theory of the Edgeworth-Box is a very simple illustration for
political process; a political process being one of negotiation between
two parties, both parties of which are under their own realistic constraints.
I discovered a more recent source in a recent journal of regional
science (the Regional Science Association which meets alternately
in Europe and the United States publishes these papers), I came across
the name of Peter Nijkamp of Holland. His concern was a multivariable
decision process concerning the building of polders of land salvaged
from the sea by building dykes, into which you have to work in economic
and political considerations and quantify them. I don't have the citations
at hand, but I'm going to be using it in my paper and I think that
those that are interested might look it up.
The fact that we talk about political processes does not mean that
we talk about arbitrary positions and arbitrary values. To emphasize
his point we're all building darns of one sort or another, and they
contain a lot of different facets. You know tile FASB process is a
political process; it is almost a legislative process in the way it
behaves. The end result of it, however, is that the board's decisions
will come out as standards, and then as they are usable and achieve
the desired ends they result in general agreement. The process
of the FASB is just a little bit more formalized than the APB's, but
in essence, there were "political pressures" on the APB
as well as there are political pressures on the FASB. Again, I think
it's good that we all have our input and our day in court, as it were.
I hope all of us are going to be constructive about it, and help those
fellows do the very rough job that they have.
In terms of your talk, I wonder whether you might want to comment
on two things that have occurred to me? One is whether the search
for the sort of perfect conceptual framework, or a better conceptual
framework as we saw it in the True-blood Committee is to develop a
framework that can be used to narrow arguments? If you wait until
you have the last word, you won't have the first word; we've now spent
three years doing this. And secondly, and related in terms of your
notion of encoding and decoding, I wonder whether standards building
is really what we ought to be doing at this point, or whether we should
really be concentrating our efforts on slaying superstitions? In other
words, because as you make the encoding more complicated, you make
the decoding more complicated, and more costly. I wonder, therefore,
whether you probably create an environment that will create more superstitions,
that is more wrong interpretations than what the data are? My perception
is that, presently, we have too many superstitions, and perhaps more
of the effort should be devoted toward slaying superstitions rather
than building more complicated encoding devices.
I think your observations are well founded. I think that the cost/benefit
argument is one that probably can never be solved -- that there's
no way of making that animal hold still. We're not going to model
this process simply in order to look at it and to study it. This is
another case where the real world is probably the only model that
we can derive. But rather than passing on it, let me just say that
I think that the cost of it (standard building) is probably not all
that significant relative to other consideration. Secondly, collectively
we are not looking for perfect or ideal models. I'm just saying they
can't be built and, therefore, that we have to adopt, if you wish,
a system that will approximate useful models for a period of
time. In that connection, I think what the Trueblood Committee did
was quite worthwhile. (Since you spent a lot of time on it, and I
didn't, I think it was quite worthwhile.) I also think it's quite
worthwhile and necessary that the FASB has spent three years on the
Conceptual Framework; frankly, the Conceptual Framework is another
attempt to enable us to look at things from a general viewpoint given
certain constraints about the rules. The thesis repeats itself --
if we know what those constraints are, and what the output is given
those constraints, they are adaptable. Adapting from rules is a little
easier in use, so I think it will be worthwhile to have the Conceptual
Framework. At least it will quiet some people that have been saying,
"Don't look at this problem until you have the Conceptual Framework."
At the same time, I very, very much want to stress that when the Conceptual
Framework is finally with us, accept it, adopt it, and put it into
standards. Remember the way I use the word "standards."
It is not an end, all it is a relative measure to be used in another
context with adaptive rules. I hope that we may be able to use it
well, as I said, for perhaps twelve years; although experience tells
me, once it gets in, it will be used, perhaps as a superstition for
maybe ten more years. The Conceptual Framework is no panacea. I think
that no one from any group, the AICPA, the FEI, the FASB, the government
or anybody else, should hold his breath and say that these problems
are going to be solved simply. If we do agree on a conceptual framework,
however, it will give us boundaries within which we can operate.
You can get all your Ph.D. students working on this problem, on developing
adaptive rules therefrom. I think that's the important thing. We have
it in measurement, in spades. You know, in fact, the only thing that
"scientists" have over us "social scientists"
is that they are more precise with their adaptive rules. They are
more precise in their measurements because they do their jobs a little
narrower, and they have better adaptive rules. Their rules are such
that they can replicate their experiments, not all the time, but generally
most of the time.