The Economics of Artificial Intelligence

Home > Other > The Economics of Artificial Intelligence > Page 50
The Economics of Artificial Intelligence Page 50

by Ajay Agrawal


  demand.3 Baumol (1967) showed that the greater rate of technical change

  2. Other papers empirically analyzing the sector shifts include Dennis and Iscan (2009), Buera and Kaboski (2009), Kollmeyer (2009), Nickell, Redding, and Swaffi

  eld (2008), and Rowthorn

  and Ramaswamy (1999).

  3. Acemoglu and Guerrieri (2008) also propose an explanation based on diff erences in capital deepening.

  294 James Bessen

  Fig. 10.2 Manufacturing share of the labor force

  Sources: US Bureau of the Census 1975; BLS Current Employment Situation.

  Note: Labor force includes agricultural laborers.

  in manufacturing industries relative to services leads to a declining share of

  manufacturing employment under some conditions (see also Lawrence and

  Edwards 2013; Ngai and Pissarides 2007; Matsuyama 2009).

  But diff erences in productivity growth rates do not seem to explain the

  initial rise in employment. For example, during the nineteenth century, the

  share of employment in agriculture fell while employment in manufacturing

  industries such as textiles and steel soared both in absolute and relative

  terms. But labor productivity in these manufacturing industries grew faster

  than labor productivity in agricultural. Parker and Klein (1966) fi nd that

  labor productivity in corn, oats, and wheat grew 2.4 percent, 2.3 percent,

  and 2.6 percent per annum from 1840– 1860 to 1900– 10. In contrast, labor

  productivity in cotton textiles grew 3 percent per year from 1820 to 1900 and

  labor productivity in steel grew 3 percent from 1860 to 1900.4 Nevertheless,

  employment in cotton textiles, and in primary iron and steel manufacturing,

  grew rapidly then.

  The growth of manufacturing relative to agriculture surely involves some

  general equilibrium considerations, perhaps involving surplus labor in the

  agricultural sector (Lewis 1954). But at the industry level, rapid labor pro-

  ductivity growth along with job growth must mean a rapid growth in the

  4. My estimates, data described below.

  Artifi cial Intelligence and Jobs: The Role of Demand 295

  equilibrium level of demand—the amount consumed must increase suffi

  -

  ciently to off set the labor- saving eff ect of technology. For example, although

  labor productivity in cotton textiles increased nearly thirtyfold during the

  nineteenth century, consumption of cotton cloth increased one hundred-

  fold. The inverted- U thus seems to involve an interaction between produc-

  tivity growth and demand.

  A long- standing literature sees sectoral shifts arising from diff erences in

  the income elasticity of demand. Clark (1940), building on earlier statis-

  tical fi ndings by Engel (1857) and others, argued that necessities such as

  food, clothing, and housing have income elasticities that are less than one

  (see also Boppart 2014; Comin, Lashkari, and Mestieri 2015; Kongsamut,

  Rebelo, and Xie 2001; and Matsuyama 1992 for more general treatments

  of non homothetic preferences). The notion behind “Engel’s Law” is that

  demand for necessities becomes satiated as consumers can aff ord more, so

  that wealthier consumers spend a smaller share of their budgets on neces-

  sities. Similarly, this tendency is seen playing out dynamically. As nations

  develop and their incomes grow, the relative demand for agricultural and

  manufactured goods falls and, with labor productivity growth, relative

  employment in these sectors falls even faster.

  This explanation is also incomplete, however. While a low- income elastic-

  ity of demand might explain late twentieth century deindustrialization, it

  does not easily explain the rising demand for some of the same goods during

  the nineteenth century. By this account, cotton textiles are a necessity with

  an income elasticity of demand less than one. Yet, during the nineteenth

  century, the demand for cotton cloth grew dramatically as incomes rose.

  That is, cotton cloth must have been a “luxury” good then. Nothing in the

  theory explains why the supposedly innate characteristics of preferences for

  cloth changed.

  It would seem that the nature of demand changed over time. Matsuyama

  (2002) introduced a model where the income elasticity of demand changes

  as incomes grow (see also Foellmi and Zweimüller 2008). In this model, con-

  sumers have hierarchical preferences for diff erent products. As their incomes

  grow, consumer demand for existing products saturates and they progres-

  sively buy new products further down the hierarchy. Given heterogeneous

  incomes that grow over time, this model can explain the inverted- U pattern.

  It also corresponds, in a highly stylized way, to the sequence of growth across

  industries seen in fi gure 10.1.

  Yet, there are two reasons that this model might not fi t the evidence very

  well for individual industries. First, the timing of the growth of these indus-

  tries seems to have much more to do with particular innovations that began

  eras of accelerated productivity growth than with the progressive saturation

  of other markets. Cotton textile consumption soared following the introduc-

  tion of the power loom to US textile manufacture in 1814; steel consump-

  tion grew following the US adoption of the Bessemer steelmaking process

  296 James Bessen

  in 1856, and Henry Ford’s assembly line in 1913 initiated rapid growth in

  motor vehicles.

  Second, there is the general problem of looking at the income elasticity

  of demand as the main driver of structural change: the data suggests that

  prices were often far more consequential for consumers than income. From

  1810 to 2011, real gross domestic product (GDP) per capita rose thirtyfold,

  but output per hour in cotton textiles rose over eight hundredfold; infl ation-

  adjusted prices correspondingly fell by three orders of magnitude. Similarly,

  from 1860 to 2011, real GDP per capita rose seventeenfold, but output per

  hour in steel production rose over 100 times and prices fell by a similar

  proportion. The literature on structural change has focused on the income

  elasticity of demand, often ignoring price changes. Yet these magnitudes

  suggest that low prices might substantially contribute to any satiation of

  demand. I develop a model that includes both income and price eff ects on

  demand, allowing both to have changing elasticities over time.

  The inverted- U pattern in industry employment can be explained by a

  declining price elasticity of demand. If we assume that rapid productivity

  growth generated rapid price declines in competitive product markets, then

  these price declines would be a major source of demand growth. During the

  rising phase of employment, equilibrium demand had to increase propor-

  tionally faster than the fall in prices in response to productivity gains. During

  the deindustrialization phase, demand must have increased proportionally

  less than prices. Below I obtain estimates that show the price elasticity of

  demand falling in just this manner.

  To understand why this may have happened, it is helpful to return to the

  origins of the notion of a demand curve. Dupuit (1844) recognized that

  consumers placed diff erent values on goods use
d for diff erent purposes. A

  decrease in the price of stone would benefi t the existing users of stone, but

  consumers would also buy stone at the lower price for new uses such as

  replacing brick or wood in construction or for paving roads. In this way,

  Dupuit showed how the distribution of uses at diff erent values gives rise to

  what we now call a demand curve, allowing for a calculation of consumer

  surplus.

  This chapter proposes a parsimonious explanation for the rise and fall of

  industry employment based on a simple model where consumer preferences

  follow such a distribution function. The basic intuition is that when most

  consumers are priced out of the market (the upper tail of the distribution),

  demand elasticity will tend to be high for many common distribution func-

  tions. When, thanks to technical change, price falls or income rises to the

  point where most consumer needs are met (the lower tail), then the price

  and income elasticities of demand will be small. The elasticity of demand

  thus changes as technology brings lower prices to the aff ected industries and

  higher income to consumers generally.

  Artifi cial Intelligence and Jobs: The Role of Demand 297

  10.2 Model

  10.2.1 Simple Model of the Inverted- U

  Consider production and consumption of two goods—cloth and a

  general composite good—in autarky. The model will focus on the impact

  of technology on employment in the textile industry under the assumption

  that the output and employment in the textile industry are only a small part

  of the total economy.

  Production

  Let the output of cloth be q = A · L, where L is textile labor and A is a measure of technical effi

  ciency. Changes in A represent labor- augmenting

  technical change. Note that this is distinct from those cases where automa-

  tion completely replaces human labor. Bessen (2016) shows that such cases

  are rare, and that the main impact of automation consists of technology

  augmenting human labor.

  I initially assume that product and labor markets are competitive so that

  the price of cloth is

  (1)

  p = w/ A,

  where w is the wage. Below, I will test whether this assumption holds in the

  cotton and steel industries.

  Then, given a demand function, D( p), equating demand with output implies D( p) = q = A L

  or

  (2)

  L = D( p) / A.

  We seek to understand whether an increase in A, representing technical

  improvement, results in a decrease or increase in employment L. That

  depends on the price elasticity of demand, , assuming income is constant.

  Taking the partial derivative of the log of equation (2) with respect to the

  log of A,

  ln L

  ln p

  ln D( p)

  = ln D( p)

  1 =

  1,

  .

  ln A

  ln p

  ln A

  ln p

  If the demand is elastic ( > 1), technical change will increase employment;

  if demand is inelastic ( < 1), jobs will be lost. In addition to this price eff ect, changing income might also aff ect demand as I develop below.

  Consumption

  Now, consider a consumer’s demand for cloth. Suppose that the con-

  sumer places diff erent values on diff erent uses of cloth. The consumer’s fi rst

  298 James Bessen

  set of clothing might be very valuable and the consumer might be willing

  to purchase even if the price is quite high. But cloth draperies might be a

  luxury that the consumer would not be willing to purchase unless the price

  is modest. Following Dupuit (1844) and the derivation of consumer surplus

  used in industrial organization theory, these diff erent values can be repre-

  sented by a distribution function. Suppose that the consumer has a number

  of uses for cloth that each give her value v, no more, no less. The total yards of cloth that these uses require can be represented as f ( v). That is, when the uses are ordered by increasing value, f ( v) is a scaled density function giving the yards of cloth for value v. If we suppose that our consumer will purchase cloth for all uses where the value received exceeds the price of cloth, v > p, then for price p, her demand is

  p

  D( p) = f ( z) dz = 1 F( p), F( p) f ( z) dz,

  p

  0

  where I have normalized demand so that maximum demand is 1. With this

  normalization, f is the density function and F is the cumulative distribution function. I assume that these functions are continuous, with continuous

  derivatives for p > 0.

  The total value she receives from these purchases is then the sum of the

  values of all uses purchased,

  U ( p) = z f ( z) dz.

  p

  This quantity measures the gross consumer surplus and can be related to the

  standard measure of net consumer surplus used in industrial organization

  theory (Tirole 1988, 8) after integrating by parts:

  U ( p) = z f ( z) dz = – z D ( z) dz = p D( p) + D( z) dz.

  p

  p

  p

  In words, gross consumer surplus equals the consumer’s expenditure plus

  net consumer surplus. I interpret U as the utility that the consumer derives

  from cloth.5

  The consumer also derives utility from consumption of the general good,

  x, and from leisure time. Let the portion of time the consumer works be l so that leisure time is 1 – l. Assume that the utility from these goods is additively separable from the utility of cloth so that total utility is

  5. Note that in order to use this model of preferences to analyze demand over time, one of two assumptions must hold. Either there are no signifi cant close substitutes for cloth or the prices of these close substitutes change relatively little. Otherwise, consumers would have to take the changing price of the potential substitute into account before deciding which to purchase.

  If there is a close substitute with a relatively static price, the value v can be reinterpreted as the value relative to the alternative. Below I look specifi cally at the role of close substitutes for cotton cloth, steel, and motor vehicles.

  Artifi cial Intelligence and Jobs: The Role of Demand 299

  U ( v) + G( x,1 l),

  where G is a concave diff erentiable function. The consumer will select v, x, and l to maximize total utility subject to the budget constraint

  wl

  x + pD( v),

  where the price of the composite good is taken as numeraire. The consumer’s

  Lagrangean can be written

  L( v, x, l) = U( v) + G( x,1 l) + ( wl x p D( v)).

  Taking the fi rst order conditions, and recalling that under competitive mar-

  kets, p = w / A, we get

  p

  G

  ˆ

  v = G

  = Gl , G

  ;

  l w

  A

  l

  l

  G represents the marginal value of leisure time and the second equality

  l

  results from applying assumption (1). In eff ect, the consumer will purchase

  cloth for uses that are at least as valuable as the real cost of cloth valued

  relative to leisure time. Note that if G is constant, the eff ect of prices and the l

  eff ect of income are inversely related. This means that the price elasticity of

  demand will equal the income elasticity of demand. However, the mar
ginal

  value of leisure time might very well increase or decrease with income; for

  example, if the labor supply is backward bending, greater income might

  decrease equilibrium G so that leisure time increases. To capture that notion, l

  I parameterize G = w so that

  l

  (3)

  ˆ

  v = w / A = w 1 p, D( ˆ v) = 1 F( ˆ v).

  10.2.2 Elasticities

  Using equation (3), the price elasticity of demand holding wages constant

  solves to

  ln ˆ

  v

  =

  ln D = ln D( ˆ v)

  = pf ( ˆ v) w 1,

  ln p

  ln ˆ

  v

  ln p

  1

  F ( ˆ

  v)

  and the income (wage) elasticity of demand holding price constant is

  ln ˆ

  v

  = ln D = ln D( ˆ v)

  = 1

  (

  ) .

  ln w

  ln ˆ

  v

  ln w

  These elasticities change with prices and wages or alternatively with

  changes in labor productivity, A. The changes can create an inverted- U in

  employment. Specifi cally, if the price elasticity of demand, ε, is greater than

  1 at high prices and lower than 1 at low prices, then employment will trace an

  inverted- U as prices decline with productivity growth. At high prices relative

  to income, productivity improvements will create suffi

  cient demand to off set

  job losses; at low prices relative to income, they will not.

  300 James Bessen

  A preference distribution function with this property can generate a kind

  of industry life cycle as technology continually improves labor productiv-

  ity over a long period of time. An early stage industry will have high prices

  and large unmet demand, so that price decreases result in sharp increases in

  demand; a mature industry will have satiated demand so further price drops

  only produce an anemic increase in demand.

  A necessary condition for this pattern is that the price elasticity of demand

  must increase with price over some signifi cant domain, so that it is smaller

  than 1 at low prices but larger than 1 at high prices. It turns out that many

  distribution functions have this property. This can be seen from the following

  propositions (proofs in the appendix):

 

‹ Prev