Technology Creation, Diffusion, and Growth Cycles

John D. Stiver∗

This Draft: July 2003

Abstract

Standard macroeconomic models that assume an exogenous stochastic process for mul-

tifactor productivity offer the interpretation that recessions are the result of ”bad news”

(technological regress) and expansions are the result of ”good news” (technological advance-

ment). The view taken here is that both expansions and recessions are the result of ”good

news” in the sense that in both cases, aggregate production possibilities have increased.

Recessions can be thought of as the transition from one technological frontier to the next.

∗
Department of Economics, University of Connecticut, One University Place, Stamford, CT 06901. Tel:
(203)251-8433. FAx: (203)251-8592. Email: John.Stiver@uconn.edu

1
1 Introduction

Standard neoclassical models of economic fluctuations often rely on an exogenous autoregressive

process for multifactor productivity. The interpretation is that recessoins are the result of

a temporary decline in productivity while expansions are the result of a temporary increase

in productivity. In other words, downturns are the result of ’bad” news and expansions are

the result of ”good” news. The view taken here is that neither recessions nor expansions are

necessarily good or bad news, and that both are, in fact, associated with increasing production

possibilities. The driving force behind an economy’s development lies its ability to produce

newer, more productive capital goods. However, when a new, more productive capital good is

introduced into an economy, it takes time to learn how to use it appropriately and to improve

on its operation and efficiency. Therefore, there will be a natural slowdown as the economy

transitions to the new ”technological frontier”. Once the economy has sufficient knowledge of

the new product, it begins to improve on it - producing the next generation. However, as

each new generation is introduced, the marginal productivity gains become smaller and smaller.

Eventually, the marginal gains will become small enough that the costs outweugh the benefits of

improvement. At that point, when a capital good is ”played out”, a new, revolutionary product

must be developed - and the cycle starts over.

For example, Gordon (2000) uses the example of word processing. The invention of the

typewriter surely represented a dramatic increase in productivity once individuals learned how to

type. The next generation memorey typewriter, which eliminated much repetitive retyping, also

marked a substantial productivity gain. However, as we move through successive improvements

(a DOS PC with Wordperfect 4.2, Wordperfect 6.0, Word for Windows, Word 95, Word 98, Word

2
2000), one can imagine the marginal productivity improvements rapidly approaching zero. From

my own experience, I can say that I am no more productive using windows 98 than I am using

windows 95. However, as we speak, Microsoft is working on perfecting word processing software

with voice recognition - possibly the next ”revolution” in word processing software. From a

more historical context, Gordon (2000) examines multifactor productivity frmo 1870 - 1999. He

finds that the show a growth pattern of ”slow - fast - slow- fast”.

P eriod 1870 − 1913 1913 − 1972 1972 − 1995 1995 − 1999

MF P (%4) .47 1.08 .02 1.25

Gordon attributes this pattern to patterns of innovation. The Late 1800’s produced many

revolutionary inventions. For example, the electric light (1879), the internal combustion engine

(1877),the telephone (1876). Gordon attributes the fast growth 1913 - 1972 to such inventions

as listed above. While it is difficult to distinctly classify inventions by generations, one might

suggest that the first generation electric light led to future generations such as the power station,

electric train, radios, and air conditioning. Similarly, the internal combustion engine led to future

generations such as automobiles, airplanes etc.

Similarly, the rapid expansion of the late 90’s could be attributed to the invention of the

microprocessor (first marketed by INTEL in 1971). As a first generation invention, the microchip

has led to future generation products such as pentium processors, cellular telecommunications,

and the world wide web. The relative length of the ”slowdowns” compared the expansions

should depend on two factors: (1) The difficulty involved in perfecting and learning the new

3
technology and (2) the pace at which the new product diffuses through the economy. A new

product which is difficult to learn and costly to adopt will be associated with a large period of

low productivity and a corresponding long, slow recovery. On the other hand, a product that is

easy to learn and rather costlesss to adopt will be associated with a very short contraction and

a quick recovery.

Andolfatto and Macdonald use a model of technological adoption to examine the U.S. from

1940 to the present. They discover that the period is represented well by the production and

adoption of five new technologies. The first was developed in the late forties, was easy to learn,

diffused rapidly, and quickly developed into widespread usage (over 70%). In the mid fifties, a

new product came out that was also easy to use, but was a small improvement and, hence, while

it diffused rapidly, it never reached more that a 40% acceptance rate. More recently, another

innovation began in the early seventies (while the authors don’t identify these innovations as

particular products, it is reasonable to believe that this one is computer related). Unlike previous

technologies, this one was more difficult to learn and more difficult to adopt. However, eventually

(by the mid eighties), it reached a 60% adoption rate. Their model, however, took the size of

technological advance as exogenous. This paper plans to endogenonize that variable.

Authors such as Greenwood and Jovanovic (2000) and Boldrin and Levine(2001) offer further

evidence of this type of growth cycle by looking at the stock market. Boldrin and Levine note

that the stock market tends to rise slowly, but drop very rapidly with periodic crashes. While

many suggest that this market behavior is the result of bubbles, assymetric technology shocks,

etc., Boldrin/Levine argue that this market behavior is the natural result of innovation cycles.

The market rises during periods in which an existing technology is being improved on. Note

4
that this is a period of very little uncertainty. However, eventually, the market learns that

the existing technology is ”played out”. At that point, there is a great deal of uncertainty as

to how the new technology will evolve, how difficult the diffusion process will take, etc. - the

market responds with a sharp downturn until this uncertainty is resolved (ie, a slow recovery).

Greenwood and Jovanovic use this type of explanation to explain the behavior of the stock market

during the 1970’s. In 1970, the ratio of market capitalization to GDP was approximately 1.

In 1973, that ration dropped dramatically to approximately .4 and stayed at .4 until the late

eighties. The current ratio is currently around 2. Greenwood/Jovanivic attribute this to the

development of the microchip in 1971. The market recognized the complexity and difficulty

involved in computerization and adjusted prices accordingly.

There is an important lesson to be learned from the above examples. Economic slowdowns

need not be caused by bad news. The introduction of the computer chip, while taking place

during a period of sluggish (even negative) economic growth, represented an increase, not a de-

crease in the productive cababilities of the U.S.. The measured drop in MFP is simply represents

the learning processas the economy transitions from old to new technology.

This paper formalizes the ideas discussed above by developing a general equilibrium model

of technology creation. The model is along the lines of Boldrin and Levine (2001). However, in

Boldrin and Levine, the decisions concerning the timing of new product develoment, upgrading

and reproducing existing capital are determined primarily by parameter values and technological

restrictions. Here, those decisions will be determined endogenously. A technology sector ex-

pends resources to produce new technologies. Periodically, the technology sector will develop a

”revolutionary product”, but most of the time, it simply improves on existing technologies. The

5
decision to ”revolutionize” will depend on the shape of the learning curve of the new technology.

Eventually, it will be the case that upgrading existing technologies will not be worth the effort

required. At that point, resourced will be moved into the development of new technologies.

When new technologies are developed, the manufacturing sector automatically (and costlessly)

adopts them.

2 The Model

2.1 The Capital Goods Sector

The capital goods sector of the economy uses the existing stock of capital goods to produce

new capital goods to be used by the production sector. It also uses the knowledge embodied

in capital goods to improve on existing capital goods. Specifically, capital in the economy be

indexed by i and j where i represents the type of capital and j represents the generation of

capital. For example, suppose that we consider communications equipment to be a type of

capital good. Then, generation 1 might represent the telegraph, while generation 5 might be

the cellular telephone. Assume that the representative firm in the technology sector has 1 unit

of type i, generation j capital , or kij . The production possibilities for t are

1. Investment: Each unit of kij produces A Â 1 units of the same type and generation of

capital (type i, generation j)

³ ´
Φ
2. Upgrading: Each unit of kij+1 produces 1+ωj
L units of the next generation capital type

i , generation j + 1

6
 The capital goods sector will be restricted to only undertake one activity at a time. This

shouldn’t qualitatively change the results and will allow the analysis to go through without the

need for tracking the distribution of capital across sectors. Furter, diminishing returns to labor

are excluded to avoid the necessity of tracking the distrubution of labor across sectors.

The intuition here is straightforward. The reproducting of existing capital goods is a simple,

automated task. Note that it will be possible for the economy to maintain a constant long run

rate of growth through this process. If no new capital are created and no upgrading takes place,

the model is a simple AK economy. Upgrading is a more complicated task which requires the

addition of skilled labor as well as capital. Most importantly, ungrading exhibits diminishing

returns. The idea here is that each upgrade will produce small and smaller increases in produc-

tivity. The parameter ω is meant to capture the size of these diminishing returns. For example,

when Microsoft released its first word processing software program (think of this as the first

generation of a new capital goos), the improvement in productivity over, say, electric typwrit-

ers (the latest generation of the previous type of capital) was sizeable. However, as Microsoft

proceeded through Word, Word95, Word98, etc., the productivity gains from each new edition

become smaller and smaller. This is the key to the entire model. An economy cannot grow

indefinately from improving on existing capital. Eventually, it must either settle with simply

increasing the stock of capital or it must begin searcing for new, novel ideas. Sticking with the

microsoft example, as the improvment opportunities dimished with traditional word processing

software,Microsoft began shifting resources towards things like voice recognition software (think

of voice recognition as the next type of capital).

7
2.2 The Research Sector

The research sector of the economy develops the blueprints for new capital goods and builds

the prototype. When new types of capital goods are produced, they are initially unproductive

relative to state of the art versions of earlier types. This can be thought of two ways. First,

early generations of capital are ”prototypes”, or ”betas”. They are not very productive due

to minor flaws or ”bugs” in the design. They must be tested and redesigned before they

will be used to produce consumption goods. Second, new capital have a ”learning by doing”

component. Initially, workers must learn how to use a new type of capital. As they redesign it,

they become more familiar and,hence,more productive. As with investment, R&D production is

linear. Further,it assumed that only labor is used in the research sector. This assumption is not

necessary, but allows for easier calculation. All that is necessary is for they process of upgrading

to be less labor intensive than R&D, but more labor intensive than Investment. First, a plan

for new capital must be developed. Each unit of labor applied to research can produce some

fraction B of the completed blueprint. Let φ represent the cumulative research,then

e
φ0 = φ + B L (1)

Note that it will require at least (1/B) periods to complete a new research plan (total labor

supply in the economy is set to 1). Once a plan is completed (φ = 1), any capital of type i,

generation j can be converted to type i + 1, generation 0 on a one for one basis.

8
2.3 The Production Sector

The Production sector of the economy converts type i , generation j capital into consumption

using the following technology. It should be noted here that consumption here is not measured in

physical units, but rather in quality units. That is, over time, we can produce more cunsumption

as well as better quality consumption.

c = γ i+j kij (2)

Note that there are three potential sources of growth in this economy. First is growth due to

an increase in the stock of physical capital. The second source is due the continual upgrading if

the existing capital stock and the periodic development of new, more productive capital goods.

However, upgrading capital will not represent a long run (sustainable) source of growth due to

diminishing returns to upgrading.

2.4 Consumers

Consumers have preferences defined over streams of consumption represented by the expected

utility function. Again, for simplicity, labor will be supplied exogenously at 1 unit.

X
∞
β t W (ct ) (3)
t=0

9
 1
W (c) = − c−θ
θ

where c represents consumption and β ≺ 1 is the discount rate.

3 Analysis

To begin, we need to place some initial restrictions on some of the parameters of the model to

avoid some uninteresting results. Suppose that there exists 1 unit of type zero, generation zero

capital. There are two options available to produce consumption to be available for the next

period. First, we could simply build more of the type zero, generation zero capital (we could

produce A units). This would provide us with A units of consumption (γ 0+0 A). The second

choice is to upgrade our capital to type zero, generation one. Using thuis method yields at most

Φ units of capital next period (assuming that all available labor is dedicated to upgrading rather

than innovating) which can be used to produce γΦ units of consumption next period. We want

to be sure that there is always some range for which upgrading is a preferable option to simply

reproducing existing capital. Otherwise the model collapses into a standard AK framework with

exogenous growth rate A. Therefore, it must be the case that

γΦ Â A (4)

Secondly, we know that labor will be atracted to the activity with the highest wage (i.e

10
the highest marginal product of labor). Once again, we have to be sure that there exists a

range of parameter values for which the marginal product of labor in the capital goods sector

(specifically, the marginal product of labor in upgrading) is greater than the marginal product

of labor in research. Again, assume that we are starting with 1 unit of type zero, generation

zero capital. One unit of labor used in upgradin will generate Φ units of type zero generation

one capital. On the other hand, each unit of labor dedicated to research produces B units of a

blueprint for type 1, generation zero capital. Note that the lifetime value of type one capital is

γ times that of type zero capital. Therefore, it must be the case that

Φ Â γB (5)

to be sure that there exists a labor allocation with a positive amount of labor engaged in

upgrading. Otherwise the model would once again reduce down to an AK model (with an

average growth rate of γA).

Given the above parameter restrictions, the model will revolve between three distinct regimes.

While the current type of capital is quite new, the costs of upgrading are small relative to the

benefits. Thertefore, early on, all the economy’s resources not devoted to consumption will be

used upgrading the economy’s capital stock. In this region, growth is a consequence of technology

accumulation as well as physical capital accumulation. Eventually, the current technology will

”play out”. That is, the costs of upgrading will drop below the benefits. This will result in the

labor force shifting from upgrading into research and the capital stock shifting out of upgrading

and into traditional investment. Growth in this region slows because technology accumulation

11
stops temporarily (actually, technology accumulation is still taking place but hasn’t manifested

in increased productive capability yet). To determine the upper bound for upgrading, set the

marginal product of capital in investment equal to the marginal product of capital in upgrading.

Note that the marginal product in upgrading must be scaled up by the factor γ because new

generations of capital produce more consumption than older generations.

γΦ
A =
1 + ωj ∗
µ ¶µ ¶
1 γΦ
j∗ = −1 (6)
ω A

This solution makes sense. The ”technological frontier” in inversely related to ω. Recall

that this parameter governs the degree to which diminishing returns result with each upgrade.
γΦ
The term A
represents the baseline productivity in upgrading relative to investing. The above

paremeter restriction guarentees that this is bigger that one.

The length of the second phase - one in which physical capital simply reproduces itself is
¡1¢
determined through the parameter value for B. A completeed plan takes B
periods. Once

the plan is completed, some of the economy’s capital will be converted into the new type of

capital. Note that new capital is initially quite unproductive relative to the existing capital

(existing capital is generation j ∗ while the new type will begin at generation 0). The new type,

however, has a very steep learning curve and can be upgraded so that eventually it will be more

productive than the existing stock. Suppose that type one capital has just been invented. You

have one unit of type zero, generation j ∗ capital at your disposal. You have two options for

producing consumption at some future time period M ∗ . First, you could simply reproduce your

12
existing capital for (M ∗ − 1) periods and then convert that capital into consumption. That
∗ −1
would provide AM units of consumption. The second option would be to convert your capital

into type one in the current period, upgrade that capital for M ∗ − 2 periods, and then convert

the resulting capital into consumption. This would result in

"M ∗ −2 µ ¶#
Y Φ
M ∗ −2
γ (7)
i=0
1 + ωi

units of capital in time M ∗ . Setteing the two equal will allow us to solve for M ∗ . Note

that M ∗ will dictate the point at the new type of capital becomes useful for producing current

consumption. Hence any of the old type of capital should be used up at M ∗ . At this point, the

cycle starts anew.

3.0.1 Stage 1: Upgrading

The economy initially begins with an initial stock of capital. For simplicity, assume that the

initial stock is 1 unit of type 0, generation 0 capital (k00 ) . For periods 1 through j ∗ the capital

stock is upgraded to the next generation every period. Labor is entirely devoted to upgrading

the capital stock. The economy grows at the rate of capital accumulation plus technology growth.

The first order condition governing consumption give us optimal consumption growth.

µ ¶θ+1 µ ¶−1
ci γ iΦ
β =
ci+1 1 + ωi
µ ¶ 1+θ
1
ci+1 γ iΦ
= (8)
ci 1 + ωi

13
 Note that during this phase, capital becomes more productive, but also becomes more ex-

pensive. Initially, the stock of physical capital grows, but eventually, begins to shrink. In fact,

at j ∗ we know the growth of potential output through invesment (gk = A) is equal to the growth

of potential output through upgrades (gγ + gk ) .

3.0.2 Stage 2: Research

At j ∗ the rising cost of new capital finally outpaces productivity and the process of upgrading

stops. Labor is shifted into research and capital is shifted into investment. During this period,

the capital stock grows at the exogenous rate A. From the first order condition for consumption,

we know that consumption grows at the rate

ct+1 1
= (Aβ) 1+θ (9)
ct

The length of the research stage is determined by the exogenous parameter B.

3.0.3 Stage 3: Development

At time j ∗ + B the blueprint for a new type of capital becomes available. At this point, some

of the capital stock will be reproduced and used for current consumption while the remaining

capital stock is converted to the new type and then upgraded until it is useful for producing

consumption goods. Therefore, the period j ∗ + B budget constraint is given by

14
 kij = c + A−1 kij0 + ki+1,0 (10)

During the economy’s ”development” phase, the capital used for consumption grows at an

exogenous rate of A. Therefore, condition consumption consumption grows at rate

ct+1 1
= (Aβ) 1+θ (11)
ct

³ ´
γΦ
New capital in the economy grows at the rate 1+ωj
= gk .

Old capital can be used for either consumption or investment. Therefore, the period by period

budget constraint for the periods j ∗ + B ≤ t ≤ j ∗ + B + M ∗ is given by

kt = ct+1 + A−1 kt+1 (12)

Finally, note that any optimal plan will use up the old type capital once the new capital is
¡ ∗ ¢
ready for consumption purposes kijM = 0 .

To solve the model, first note that we have a starting capital stock (of type 0 generation 0)

equal to one which is allocated towards consumption and upgrading.

15
 µ ¶−1
Φ
k00 = 1 = c + k01
1 + ωi

At time j ∗ the process of investment begins. The period j ∗ budget constraint is equal to

k0j ∗ = 1 = c + (A)−1 k0j

At period period j ∗ + B , the new capital becomesbudget constraint is given by

kij = c + A−1 kij0 + ki+1,0 (13)

Lastly, we have the ternminal condition that the old style capital is completely used up at
¡ M∗ ¢
the time that the new style capital is productive enough to produce consumption. kij ∗ = 0 .

Solving the budget constraints forward and inserting the starting and terminal condition results

in the lifetime budget constraint. Next, at time j ∗ +B (when the plans for the new type of capital

become available) resources allocated towards new capital creation must provide equal lifetime

utility at the margin as resources allocated towards consumption. Let V00 denote the lifetime

value of one unit of type 0 generation 0 capital. The value of kij unite of type i generation j capital
−θ
will produce (γ i+j kij ) V00 units of welfare. Therefore, the optimality condition governing new

capital formation can be written as

16
 " M µ
∗ ¶ #−θ
Y Φ
−(1+θ) M ∗ +1 M∗
β (c) =β γ k10 V00 (14)
i=1
1 + ωi

Finally, we have the bellman equation defining the value function.

j ∗ +B+M ∗
" M µ
∗ ¶ #−θ
X Y Φ
j ∗ +B+M ∗ +1 M ∗ +1
V00 = β i U (ci ) + β γ k10 V00 (15)
i=1 i=1
1 + ωi

Equations (14), (15), and the lifetime budget constraint define a solution for k10 , c,and V00 .

Choosing parameter values is the next step.

A (Productivity in Investment) 1.03

γ (Rate of Growth in Technology) 1.05

Φ (Baseline productivity in Upgrading) 1.06

B (Productivity in Research) .2

ω (Degree of Diminishing returns in upgrading) .007

θ (Intertemporal Elasticity of Substitution in Consumption) -.4

However, the economy actually averages 4% annual growth by moving back and forth between

the low growth state (research and development) and the high growth state (upgrading). Given

the parameters, the maximum growth of consumption is

17
 ci+1 1
= (γΦ) 1+θ = 1.19
ci

However, this growth rate continually declines. The first switching point j ∗ is equal to 12

(capital reaches twelve generations before upgrading becones to expensive. At j ∗ (and until

j ∗ + B + M ∗ ), growth of the economy is dictated by the growth of physical capital through

investment. During this period, we get a growth of consumption equal to

ct+1 1
= (Aβ) 1+θ = .98
ct

So consumption shrinks during the ”recession”. The length of this contraction is determined

by the exogenous parameter B. (In this case,the research phase takes 5 periods) and M ∗ (in this

case, M ∗ is determined to b 5). Therefore, the economy experiences alternating expansions (of

length equal to 12 periods) where consumption grows at an average rate of 8.5% and recessions

(of length equal to 10 periods) where consumption contracts at a rate of 2%. Figures (1) − (4)

plot the transitions for new capital, the aggregate capital stock, consumption, and productivity

(units of new capital per unit of old capital ).

4 Conclusions

Standard macroeconomic models stress the fact that expansions and contractions are the result of

random fluctuations to productivity. Expansions are the result of ”good” news while recessions

18
are the result of ”bad” news. The view taken here is that both recessions and expansions are

both the result of the development of new technologies.

When new technologies are invented, they are initially very unproductive. However, over

time, they are modified and improved upon. Each successive generation becomes more produc-

tive at producing consumption goods, but becomes more costly to update. For example, the

telegraph was a new type of capital used in the production of communication services, then suc-

cessive generations would be the telephone, cordless phone, cellular phones, etc.. The constant

pattern of research, invention, development and replacement create a natural cycle of economic

activity.

• When a new type of capital is in its early stages, the cost up upgrading is low relative

to the productivity gains. Consumption grows at above trend rates as the capital stock

is constantly upgraded. However, as upgrades become more and more expensive, more

capital is devoted to development rather than consumption. Eventually, a technology

becomes ”played out” in the sense that it is no longer cost effective to continue updating.

Consumption growth falls and labor is reallocated to research activities (i.e., the invention

of new types of capital). Note that the length of this expansionary period depends on the

degree of decreasing returns to upgrading (ω) .

• Consumption growth remains low during the research period. Capital is simply reproduced

rather than upgraded. Labor is devoted to reserach activities. Eventually, a new type of

capital is discovered (discovery time is treated as exogenous in this framework). However,

the new capiatl is still unproductive relative to the existing capital stock. A development

phase must take place before the new capital can be used for consumption purposes. Note

19
 that while new capital is being improved, the old capital is in the process of being used

up in anticipation of the new capital coming online. Therefore, the aggregate capital

stock falls. This period looks very much like a recession - a declining capital stock and

low consumption growth - but actually, lifetime consumption possibilities have actually

increased. The length of the development also hinges crucially on the curvature of the

upgrading process (ω) and well as the baseline level of productivity (Φ) . Once the new

type of capital is productive enough to produce consumption goods, the cycle begins anew.

Obviously, there are some important drawbacks in the current framework. With only one

production sector, there is no possibility of simultaneously operating multiple technologies -

each in a different developmental stage. Further, the research stage should ideally be treated

as endogenous with a tradeoff between research time and certain features of the new capital

(i.e. initial productivity or developmental path). These are questions to be looked at in future

research.

5 References

Andolfatto, David and Glen MacDonald (1998), ”Technology Diffusion and Aggregtae Dynam-

ics”, Review of Economic Dynamics, 1(2), 338-370.

Aghion, P. and P. Howitt (1996), ”On the Microeconomic Effects of Technological Change”,

University College London

Aghion, P. and P. Howitt (1992), ”A Model of Growth Through Creative Destruction”,

Econometrica, 60, 323-352.

20
 Atkeson, Andrew and Parick Kehoe (1997), ”Industry Evolution and Transition: A Neoclas-

sical Benchmark”, NBER Working Paper #6005.

Boldrin, Michael and David Levine (2001), ”Growth Cycles and Market Crashes”, Journal of

Economic Theory, 96, 13-39.

Campbell, Jeffrey (1998), ”Entry, Exit, Embodied Technology and the Business Cycle”, Re-

view of Economic Dynamics, 1(2), 371-408.

J. Gali (1999), ”Technology, Employment, and The Business Cycle”, American Economic

Review, 89, 249-271

Gordon, Robert (2000), ”Does the ’New Economy’ Measure Up to the Great Inventions of

the Past?”, NBER Working Paper #7833.

Gort, Michael and S. Klepper (1982), ”Time Paths and the Diffusion of Product Innovations”,

Economic Journal, 92, 630-653.

Greenwood, Jeremy and Mehmet Yorukoglu (1997), ”1974”, Carnegie-Rochester Conference

Series on Public Policy, 46, 49-96.

Greenwood, Jeremy, Zvi Hercowitz and Per Krusell (1997), ”Long Run Implications of In-

vestment Specific Technological Change”, American Economic Review, 87, 342-62.

Greenwood, Jeremy and Boyan Jovanovic (1999), ”The IT Revolution and the Stock Market”,

AER Papers and Proceedings, 89, 116-122.

Jones, Charles (1997), ”The Upcoming Slowdown in U.S. Economic Growth”, NBER Working

Paper #6284.

21
 Jones, Charles (1995), ”R&D Based Models of Economic Growth”, Journal of Political Econ-

omy, 103, 759-784.

Jovanovic, Boyan and S. Lach (1997), ”Product Innovation and the Business Cycle”, Inter-

national Economic Review, 38(1), 3-22.

Jovanovic, Boyan and R. Rob (1990), ”Long Waves and Short Waves”, Econometrica, 58,

1391-1409.

Hornstein, A. and Per Krusell (1996), ”Can Technology Improvementc Cause Productivity

Slowdowns?”, NBER Macroeconomics Annual 11, 209-259.

22
 Figure 1:Capital Figure 2:Capital stock
1.4 1.15

1.2
1.1
1

0.8 1.05

0.6 1
0.4
0.95
0.2

0 0.9
0 5 10 15 20 0 5 10 15 20

Figure 3:Consumption Figure 4:Productivity

0.018 1.12

1.1
0.016
1.08
0.014
1.06
0.012
1.04

0.01 1.02
0 5 10 15 20 0 5 10 15 20