Professional Documents
Culture Documents
Ulrich Witt
Paper prepared for the DIME workshop “Process, Technology, and Organization:
Towards a Useful Theory of Production”, Pisa, November 8-9, 2010
__________________________________
1. Introduction
Despite its now widely recognized importance, one obstacle for coming to
grips with knowledge and its role in the economy is the rather loose way in which
the very concept of “knowledge” is being made use of. Several interpretations,
analytical distinctions, and extensions exist that are only partly or not at all
commensurable. Hayek (ibid.), for instance, reflects on the role of (price)
information without explaining how knowledge and information relate to each
other. In the innovation literature, distinctions between different kinds of
knowledge like scientific vs. technological knowledge (Metcalfe 1998, Chap. 4) or
procedural vs. propositional knowledge (Zellner 2003) can be found; or between
knowing how vs. knowing that (Ryle 1949). Further internal vs. external
knowledge (Antonelli 1999), tacit vs. overt knowledge (Polanyi 1958) , knowledge
with the character of a public, a private, or a free good (Metcalfe 1998, Chap. 4)
and so on.
All these distinctions are relevant and, in fact, the list could be prolonged.
What seems to be lacking, though, is a comprehensive systematization of the
various facets of knowledge in an economic context, including a coherent taxonomy,
and a clarification of the problem of whether and how knowledge is measurable
either in quantitative or qualitative terms. Admittedly , knowledge is an elusive
phenomenon that is difficult to analyze with accuracy not only in economics (see,
e.g., Knorr Cetina 1999). Yet, in the economic domain, confusion at the conceptual
level seems particularly precarious given the unique and immensely important
feedback relationship between the production and growth of knowledge and
production and growth in the economy.
Foray (2004), the problems will for two reasons be framed here within an
evolutionary perspective. First, knowledge is a particularly significant agent in the
restructuring and growth of the economy and is itself subject to a characteristic
evolution (see Nelson and Nelson 2002). Following Metcalfe’s (2002) view,
evolutionary processes in the economy should therefore provide an excellent point
of reference for the following discussion. Second, a major source of inspiration for
these discussions is a comparison between knowledge created by evolution in
nature and the characteristics of human knowledge. In nature and the economy,
(re-) production is based on the generation, processing, and accumulation of
information. However, the very different ways in which this condition is met in the
two spheres – the different knowledge “technologies” – shed some light on
regularities and limitations inherent in the production and use of knowledge.
An example can illustrate the associative act. Imagine a test person, e.g. a
first grade economics student, who is given the information (i) “labor earns a wage
and labor provides services”. The two statements in (i) have the concept “labor” in
common and jointly represent the simplest case of what is called in cognition
theory a propositional network (Anderson 2000, Chap. 5) by which the properties
of that concept are mentally represented. Suppose the concept “labor” is a new,
unknown piece of information for the test person. Suppose further that the test
person associates the rest of statement (i) with a spontaneously recalled statement
(ii) “workers earn a wage and workers spend time and effort” – another
propositional network. The basis for this association is the cognitive cue contained
in (i) in the form of the attribute “earns a wage”. Since the propositional networks
(i) and (ii) have this attribute in common, it allows the test person to extend the
meaning of the concept “workers” o the concept “labor”. By logical inference, sense
can then be made of the “services” provided by labor, namely by identifying them
with “time and effort”.
The simple example suggests some more general conclusions. First, note that
information spontaneously activated in the associative act is in common parlance
considered part of the knowledge an agent commands. Following this
understanding, knowledge can be conceptualized as meaningful (semantically
connected) information that exists in the mind of an agent or that can be recalled
from memory or traced from extra-somatic stores for which the agent possesses the
interpretative code. Second, the fact that meaning attribution is based on an
association with already known cognitive concepts implies that newly received
information cannot be made sense of without already pre-existing knowledge. 1
Third, even though knowledge is composed of pre-existing and acquired
information, it is not only the mere sum of pieces of semantic information (each
with an own syntax). Their composition – the systematic, relational structure of the
information stored memory – is an essential feature of knowledge. Cognitive
concepts (scripts, schemas, prototypes) and propositional networks relating these
concepts to one another have been given as simple examples for building blocks
making up the relational structure. Examples of more complex ones are categorical
reductions, generic abstractions, logical implications, and intuitive theories on
causal and functional relationships.
1
By backward induction it can then be inferred that there must be some ultimate
prior concepts which are not themselves acquired (learned) and from which the cumulative
knowledge acquisition process starts. These are likely to be innate concepts, a conclusion
supporting the “embodied knowledge” conjecture (Hayek 1952, Chomsky 1965), see Carey
(2009, Chap. 2) for a recent survey.
2
From the individual agent’s perspective, the validity of knowledge that she holds or
can access is of more or less tentative nature. Therefore the assessment of particular pieces
of knowledge can be, and implicitly usually is, based on assigning varying degrees of belief
to them. This concept is known from, and has systematically been elaborated in, subjective
the relevance, relative importance, and value also depends on the observer’s frame
of reference and the subjectively perceived context. Logical truth and degrees of
empirical confirmation indeed seem to be widely accepted criteria for assessing the
validity of factual statements and causal and functional hypotheses. For this
reason the corresponding assessments can be expected to more easily be agreed
upon than judgements on the relevance, importance, and value of information can
be. Nonetheless, the same reasons for which the non-identical knowledge of
different agents can only ordinally, and only in special cases, be compared (without
implicit value judgements) apply here too.
What has been said about the measurability and validity of knowledge by
and large applies to both propositional knowledge and procedural knowledge (the
knowing of how to do something). However, regarding the latter, some caveats need
to be made in other respects. They result from the fact that procedural knowledge
cannot always (fully) be articulated. At least in part it comes as “tacit knowledge”
3
that is not, or cannot reasonably be, encoded in semantic information and its
relational structure. Instead it exists in the mind of an agent in the form of non-
verbal concepts of sequences of actions or practices. Accordingly, the problem of
meaning attribution poses itself differently here. In some cases it takes the form of
interpreting visual information. Unlike the procedural knowledge that an agent X
holds that cannot itself be observed by others, its expression in the form of actions
carried out by X often can be observed. On that basis, an observer Y may be able to
make inferences about the underlying procedural knowledge even without getting
any verbal explanation from X. Y can attribute meaning to the visual information
she draws from observing X’s actions by associating it with concepts Y already
knows. Indeed, it is not accidental that in illiterate societies much of the knowledge
transmission between agents and generations of agents is based on non-verbal,
observational learning taking place in this way.
If, as has just been argued, knowledge is composed of information with a specific
relational structure, then one can as well speak of knowledge existing in a
genetically coded form in living nature as one can speak of knowledge existing in,
probability theory which, after all, was conceived as a theory of judgements on the validity
of human knowledge (Keynes 1921). It may be worthwhile to revive the insights on
personal attitudes towards the use of knowledge which this theory offers.
3
Polanyi (1958). The question of whether and, if so, when and why knowledge can be
tacit has triggered a long debate in which various other characteristics of knowledge have
been suggested as reasons, see, e.g., Polanyi (1966), Saviotti (1998), Cowan, David and
Foray (2000), Balconi (2002). Since the problem of tacitness is not relevant here except in
relation to procedural knowledge the debate will not be entered here. See, however, Witt,
Brökel and Brenner (2010).
or accessible to, the human mind (and, to a limited extent also in the mind of
higher animals). Let a knowledge technology be defined as the way in which
knowledge is generated, utilized, stored, and copied or communicated. It is
immediately clear then that the knowledge technologies underlying genetically
coded knowledge on the one side and human knowledge on the other are
fundamentally different, the fact not withstanding that in both cases the
technology is based on distributed and interactive knowledge. (Yet the interaction
mode differs, being based more on competition in nature and more on exchange and
communication in the human domain.)
In nature, knowledge is hard-wired into the DNA and RNA strings in the
cells of the carrying organism. However, it is not the mere existence of that genetic
information that matters, but the effects that it develops. Genetic knowledge
spontaneously expresses itself physically in the phenotypic organism, i.e. in the
form of processes like those guiding the organism’s morphological development or
the controls it exerts on the life functionings. A unique property of gene-coded
knowledge is thus its “self-executing” capacity – provided the necessary chemical
energy flows are availbale. Genetic knowledge includes a copying code by which it
spontaneously reproduces itself in its physical carriers. It interprets itself (its
“meaning”) through the biophysical signal cascades it codes. The copying process
also generates new genetic knowledge in a replication of the gene code between
generations that is subject to recombination and mutation. The emergence and
dissemination of novelty is thus part of a pre-programmed, self-executing
automatism shaped together with its outcome by natural selection.
it to other agents and generations of agents who need to acquire it for themselves
usually with a significant expense of effort and time before it can be put to action.
Given the limited lifetime of the carriers of knowledge this fact means that an
incessant transfer of knowledge between, and its reproduction by, successive
generations needs to be organized where in nature the inbuilt copying automatism
ensures the transmission.
There are also important differences between the purposes that the two
knowledge technologies serve. The differences explain in part the historical
emergence of ever new kinds of knowledge. Human knowledge is not exclusively
oriented towards enhancing reproductive success. In fact, it is the less so, the
economically more prosperous and less exposed to selection pressure humans have
become. For the early hominids – as for the animal kingdom up to the present day
– it can be conjectured that all non-inherited kinds of knowledge aimed at
augmenting the forms of production accomplished by means of the inherited
behavioral repertoire. For example, hunter-gatherer societies (but even more so the
emerging agricultural societies) developed know-how by which the productive
knowledge of nature that is embodied in other species could be manipulated and
instrumentalized to support the own life functions.
from ever new sources with ever new techniques is not the only route on which
human knowledge has expanded. The very deviation from nature’s production
mode meant that the productive activities were increasingly less coordinated by the
natural feed backs of the underlying eco-systems. The new kinds of production
based on specialization, division of labor, and trade therefore required new kinds
of knowledge: that of organization (and, one may add, of entrepreneurship), of
market, and of how to socially interact more generally (Metcalfe 2002).
Furthermore, with its success, the human economy generates and accumulates
exponentially growing amounts of wealth demanding commercial knowledge of how
to secure against risks, of banking, and of wealth administration. Likewise,
growing wealth fuels the demand of all sorts of services between men with the
corresponding service knowledge reaching from military to medical know how, from
educational to governance know how, and from the know how in arts and crafts to
the know how related to creation, administration, and coordination of knowledge
and knowledge production in science and technology.
In the previous section the successive emergence of new kinds of human knowledge
has been portrayed as a concomitant of increasing productivity, the growth in the
variety of economic activities, and the necessary organizational and administrative
adaptations. Regarding the production side, the historical sequence in which these
activities emerged matches roughly with the historical sequence in which the
traditional sectors of the economy emerged and began to dominate: agriculture,
manufacturing industries, and service industries. It may be asked, therefore,
whether the historical restructuring of the growing economy, observable in
changing value added and/or employment shares of economic sectors, is correlated
with, or even a consequence of, corresponding changes in sector-specific knowledge
and/or knowledge technologies. This question is center stage in the neo-
Schumpeterian theories of structural growth differentials between firms and
industries (Foster, Metcalfe, and Ramlogan 2006) and between countries
(Fagerberg and Verspagen 2002). By asking more generally whether economic
growth is ultimately a consequence of the growth of knowledge (presumably in the
sense of an increasing variety) the question also motivates a good part of new
Thus, while it seems a truism that human knowledge has been, and still is,
growing, this is hard to give an accurate account of. New economic activities
contingent on new varieties of knowledge and advances in information and
communication technology have brought forth a progressive expansion in the mere
volume of information stored and transmitted (for which signal-based, cardinal
measures are possible). Yet, given the insight in the previous section that stored
human knowledge is not self-expressing, these achievements do not mean much.
Knowledge encoded and stored somewhere is “dead” knowledge. Greater capacity
of transmitting encoded information from one locus to another does not change this
status (though it is a decisive prerequisite for a broader and more rapid use of
stored knowledge, see Dudley 1999). A growing variety of encoded knowledge can
develop an effect and become economically relevant only qua media which have the
capacity to express that knowledge (and have acquired it before). This is true for
both propositional and procedural knowledge.
the activities expressing the same knowledge repeatedly or in parallel on the one
hand and activities expressing a changing variety of knowledge on the other hand.
Some insights seem indeed possible in this basis. In order to show this, the media
by which human knowledge is expressed need to be specified, however.
First of all, there is the medium of the human mind. A major part, if not all,
of the activities of economic agents, productive and consumptive ones, involve an
expression of previously acquired knowledge, complemented by more or less
physical and mental effort. If, for expository convenience, the effort variable is
assumed to be distributed with an invariable mean, the working time of both the
employed and self-employed agents in the economy can be assumed to map into a
measure for the knowledge that is put to an economically valuable use. However,
the human mind is not the only medium capable of expressing man-made
knowledge. A historically increasingly significant role is also played by other
media, viz. machinery, instruments, and automata which are to be distinguished
here from mere tools. The unique combination of intelligence and fine-tuned
motility has allowed humans fairly early to create tools. This presupposes an at
least latent understanding of a cause-effect relationship which the tool is meant to
facilitate and which is encoded in the shape and functionality of the tool. A wedge
, for example, made of stone or bone suggests by its sharp edge to an observer who
finds this artefact to use it for cutting or carving. Once created, the tool can thus be
seen as “storing” this technological knowledge. Yet, as all stored knowledge, it is
ineffective as long as it is not implicitly expressed by the human action of using the
tool – provided the user is capable of interpreting the encoded information, i.e. of
inferring what the tool is good for. Knowledge expression of this kind thus involves
a human mind and more or less physical labor. (Once the tool has been created by
someone, it can of course be used by someone else without any practical knowledge
about how it was produced.)
4
The users do not have to acquire the knowledge that has been programmed into the
equipments by their developers, except what may be necessary for occasional service and
repair. All that is required is a usually rather limited know how concerning the initiation
and control of the equipment.
effects result from technical progress in the equipments and are long term. If one
capital good can be replaced by another one with the same price but a capacity to
express (re-use) more preprogrammed knowledge (i.e. to substitute more human
time for knowledge expression), this leads to a higher level of output for
(nominally) the same capital input, i.e. an upward shift of the corresponding
engineering production function.
In this respect it is a historical fact that in the 19th century the lions’s share
of the technical improvements occurred in mechanical and chemical processing
technology (exactly what substituted human knowledge expression in those
productive activities), favoring a rising capital-labor ratio in the primary industries
and manufacturing. It was only in the mid 20th century that the electronics
revolution allowed substantial technical progress also to be made in information
processing and communication. Standardization of many information processing
and communication tasks – a substantial part of the productive activities in the
service sector – seems to have occurred earlier already. The financial sector, for
example, used standardized forms and their electro-mechanical processing in the
book keeping of transactions already at the beginning of the 20th century (Austrian
1982). However, the transfer of knowledge expression from man to machine and
the re-use of processing know how programmed into the equipments took on mass
The very technical change that made this growth possible also meant, of
course, that new (technological) varieties of knowledge – the very essence of
innovations – were created and actively expressed. For the reasons mentioned, it
is not possible, however, to infer from this fact that the overall variety of knowledge
that can be expressed in an economy has been growing. There may be good reasons
to assume that this did happen in terms of both stored and active knowledge. Even
so, however, it is not clear whether such a growth is a causal factor for economic
growth in the past and will continue to be conducive to future growth. To explore
these questions, further conceptual clarifications are necessary that revolve around
the variety of knowledge and its growth.
What has been left open so far is a closer characterization of the knowledge that
the agents are able to express on the job and the question of whether there have
been changes over time in the knowledge requirement due to technical progress
and the growth of knowledge. In order to address these issues a qualitative
assessment of the variety knowledge by some criterion (similarity, relevance,
validity, importance, value etc.) is necessary. As was discussed in Section 2, this
implies additional measurement problems so that at best ordinal inter-personal or
inter-temporal comparisons of knowledge varieties are possible under rather
restrictive conditions. Given these difficulties it seems straight forward to consider
here too an activity measure as a substitute for measuring qualitative aspects of
knowledge directly – this time one that relates to the knowledge acquisition
activities.
Knowledge acquisition is, of course, costly – the more costly on average, the
more time and effort are necessary. 5 In order to be able to put knowledge of a
particular variety to work, agents therefore have to be remunerated not only for the
knowledge expression activities which their occupations require, but also for the
preceding knowledge acquisition activities. Before this background, value
judgements of a special kind are made in the markets regarding the desirability of
the combined acquisition and expression activities: the judgements consolidated
into the prices which the demand side is willing to pay as compensation to the
agents capable of expressing the desired variety of knowledge. Since on average the
cost of acquiring the desired variety are the higher, the greater that variety,
capabilities of expressing such knowledge will only be held in the market, if the
market remuneration for the expression activities indirectly also compensates the
agents for the on average higher acquisition costs. (Where this condition is not met
– as in many domains of research and education – expression of high end varieties
of knowledge that is deemed desirable by the public must be commissioned and
compensated by public spending.)
It is an empirical fact that the level of education, i.e. the average length of
training and study time in a non-employed status, has been rising in all developed
economies (the frequently stated trend towards a “learning society” described in
Lundvall 1992). Individual knowledge acquisition requires increasingly more time
to absorb the state-of-the-art knowledge (Jones 2005). Even if not all of that
education is qualifying for a later occupation, it is likely that the increasing level
of education of the work force reflects a trend by which the variety of individually
acquired knowledge is becoming greater. At the same time, the average capital-
labor ratio in the economy, average labor productivity, and real wages per hour
worked (i.e. the time during which knowledge that has previously been acquired is
expressed on the job) has been rising dramatically. However, the increase in these
variables over time is lower, if the longer, unpaid education time is included as just
suggested.
A first insight into the relationships between the growth of the variety of
knowledge and economic growth is therefore the following. If the growth of the
average variety of knowledge has in some way been causal for raising the capital-
labor ratio, labor productivity and real wages, then its impact is much less
dramatic, when taking its rising acquisition costs into account. In aggregate growth
accounting models in which labor inputs are typically measured in paid working
time this means that a good part of the total factor productivity – usually estimated
to be a quite dramatic, unexplained residual, see Hulten (2009) – can actually be
5
The costs also depends on the quality and efficiency of education and training, on
the agents’ natural learning capabilities and their individual effort and skills, on the pace
with which individually acquired knowledge becomes obsolete or is forgotten. These factors
vary not only across agents. As a consequence of increasing efficiency (e.g. resulting from
learning curve effects) or a changing motivation they can also vary over time for each
single agents.
attributed to the neglected rise in unpaid labor inputs for acquiring productive
knowledge.
The variety of encoded and stored knowledge that has been accumulated has
visibly grown (as can be seen, for example, from the ever expanding texts of
compendia describing that knowledge). At the same time each human generation
has to acquire the growing variety with its limited cognitive resources anew in
order to turn it into active individual knowledge. In response to this increasingly
more demanding challenge, the extent has grown to which knowledge is
partitioned among individuals who specialize differently in the variety of
knowledge they acquire, thus enlarging the overall variety of active knowledge
(Pavitt 1998). Nonetheless, the acquisition still seems to require increasingly more
time to absorb the state-of-the-art knowledge so that a pressure to individually
specialize further is upheld (Jones 2005). As a result of the attempt to cope with
the growth of accumulated knowledge, this process is likely to go on. Yet, it will not
do so indefinitely.
6
An extreme partitioning would be a distribution of knowledge among the agents in
an economy such that every agent commands a different variety of knowledge except the
common core (an interpretative code) required for attributing meaning to information
communicated between individuals and for coordinating a joint expression of individually
held pieces of knowledge where this is necessary in the production process.
somatic memories can. But since only knowledge that is expressed in some way
develops an economic effect, the possibility of an unlimited growth of encoded and
stored knowledge is economically irrelevant. 7 Hence, if, as new growth theory
seems to presume, the growth of the variety of active knowledge in the economy is
a cause of economic growth, then the existence of a knowledge constraint implies
a limit to economic growth. While the individual limits in knowledge acquisition
are trivially experienced by everyone, the limits to the overall variety of active
knowledge seem to go largely unnoticed. A reason for this asymmetry is
presumably that these limits seem far away, not least because of the fact that the
human population is still growing. If properly educated, there is still a huge
number of agents who can be included in the further partitioning of a growing
variety of active human knowledge.
6. Conclusions
This paper has made an attempt to provide a systematic account of knowledge, its
different uses, and their significance for production and growth. A concept of
knowledge and its different facets has been developed, followed by a short inquiry
into the measurement problem. The framework that inspired the discussion of the
role of knowledge was a comparison between the accumulation of genetically stored
information that is physically expressed in living organisms (genetic “knowledge”)
and the conditions under which man-made knowledge is accumulated. An essential
insight derived from that comparison was that human knowledge also requires a
medium for expressing it, a fact that is often not sufficiently acknowledged. The
media of knowledge expression that have been identified are the human mind that
expressing knowledge in the actions it triggers and machinery, instruments, and
automata which embody human knowledge that has been built into them. Several
implications have been discussed that shed some new light on phenomena like
scale economies in production, sectoral structural change, and the boundaries of
knowledge growth impacting on economic growth.
7
This verdict is valid as long as no intelligent automata exist that can acquire and
use knowledge by themselves; for a discussion of such a constellation, presently considered
utopian, see Wadman (2001).
References
Pavitt, K. (1998)
“Technologies, Products and Organization in the Innovating Firm: What
Adam Smith Tells Us and Joseph Schumpeter Doesn’t”, Industrial and
Corporate Change, Vol. 7, 433-451.
Polanyi, M. (1958),
Personal Knowledge: Towards a Post-Critical Philosophy, London:
Routledge.
Polanyi, M. (1983).
The Tacit Dimension. Gloucester: Peter Smith.
Ryle, G. (1949)
The Concept of Mind, London: Hutchinson.
Saviotti, P. P. (1998).
On the dynamics of appropriability, of tacit and of codified knowledge.
Research Policy, 26, 843–856.
Usher, A.P. (1954)
A History of Mechanical Inventions. New York: Dover Publications.
Wadman, W. (2001)
“Technical Change, Human Capital and Extinction”, Journal of
Bioeconomics, 3, 65-68.
Witt, U. (2009)
“Propositions About Novelty”, Journal of Economic Behavior and
Organization, 70, 311-320.
Witt, U., Brökel, T., Brenner, T. (2010),
“Knowledge and its Economic Characteristics – A Conceptual Clarification”,
in: R.Arena, A. Festré, N. Lazaric (eds.), Handbook of Knowledge and
Economics, Edward Elgar, forthcoming.
Zellner, C. (2003)
“The Economic Effects of Basic Research: Evidence for Embodied Knowledge
Transfer via Scientists’ Migration”, Research Policy 32, 1881-1895.