Rodrik - An Industrial Policy For Good Jobs

POLICY PROPOSAL | SEPTEMBER 2022
An Industrial Policy for Good Jobs
Dani Rodrik
ACKNOWLEDGMENTS
I am grateful to Wendy Edelberg, Lauren Bauer, Tim Bartik, Gordon
Hanson, and participants in the author’s workshop of The Hamilton
Project for comments and suggestions on a previous draft, Mitchell
Barnes for research assistance, and Christine Dziuba for editorial help.
THE HAMILTON PROJECT MISSION STATEMENT

The Hamilton Project seeks to advance America’s promise
of opportunity, prosperity, and growth.
We believe that today’s increasingly competitive global
economy demands public policy ideas commensurate with the
challenges of the 21st Century. The Project’s economic strategy
reflects a judgment that long-term prosperity is best achieved by
fostering economic growth and broad participation in that growth,
by enhancing individual economic security, and by embracing a role
for effective government in making needed public investments.
Our strategy calls for combining public investment, a secure
social safety net, and fiscal discipline. In that framework, the
Project puts forward innovative proposals from leading economic
thinkers—based on credible evidence and experience, not ideology
or doctrine—to introduce new and effective policy options into the
national debate.
The Project is named after Alexander Hamilton, the nation’s
first Treasury Secretary, who laid the foundation for the modern
American economy. Hamilton stood for sound fiscal policy, believed
that broad-based opportunity for advancement would drive
American economic growth, and recognized that “prudent aids
and encouragements on the part of government” are necessary
to enhance and guide market forces. The guiding principles of the
Project remain consistent with these views.
An Industrial Policy for Good Jobs
Dani Rodrik
John F. Kennedy School of Government at Harvard University
September 2022
This policy proposal is a proposal from the author. As emphasized in The Hamilton Project’s original strategy paper, the Proj-
ect was designed in part to provide a forum for leading thinkers across the nation to put forward innovative and potentially
important economic policy ideas that share the Project’s broad goals of promoting economic growth, broad-based participa-
tion in growth, and economic security. Authors are invited to express their own ideas in policy proposal, whether or not the
Project’s staff or advisory council agrees with the specific proposals. This policy proposal is offered in that spirit.
Abstract
Industrial policies have been with us for a long time, but often they have been carried out surreptitiously
and without clear motivation. The recent revival of discussions around industrial policy provides a welcome
opportunity for self-consciously crafting an improved set of policies. A modern approach to industrial policy
must respond to new circumstances. It must target “good-jobs externalities,” in addition to the traditional
learning, technological, and national security considerations. Relatedly, industrial policy’s traditional focus
on manufacturing and globally competitive industries has to be broadened to service sectors and smaller
and medium-sized firms. And the practice of industrial policy will need to rely less on traditional top-down
policy instruments—such as subsidies and tax incentives for firms—and more on collaborative, iterative
interaction whereby public agencies supply a portfolio of customized public services in exchange for firms
undertaking soft commitments on the quantity and quality of employment. With these objectives in mind,
this paper develops two types of specific initiatives: one at the local level and the other at the federal level.
The local approach builds on existing development and business assistance programs that take the form of
collaborative partnerships between local development agencies, firms, and other stakeholders aiming to
revitalize local communities and create good jobs. The federal initiative is an Advanced Research Projects
Agency (ARPA) focused on the promotion of employment-friendly technologies: ARPA-W(orkers).
ii The Hamilton Project • Brookings

Contents
Introduction..............................................................................................................................................1
The Challenge......................................................................................................................................3
The Why: Good-Job Externalities......................................................................................................3
The What: Moving Beyond Manufacturing......................................................................................6
The How: From Top-Down Subsidies to Collaborative, Customized Assistance...........................8
The Proposal...........................................................................................................................................12
A Federal Innovation Initiative to Promote Employment-Friendly Technologies.......................12
Local and Regional Industrial Policies............................................................................................14
Concerns.................................................................................................................................................17
Conclusion..............................................................................................................................................19
Endnotes..................................................................................................................................................20
References...............................................................................................................................................22
The Hamilton Project • Brookings iii

Introduction
Industrial policy is as old as the state itself. Virtually every prosperity for all (Rodrik 2017; Sitaraman 2019). The rise of
government in history has engaged in policies to promote inequality, economic insecurity, and labor market polariza-
economic activities regarded as critical to national security, tion, as well as the disappearance of good jobs, have contrib-
economic well-being, or the sovereign’s coffers. Under the uted to a sense that governments need to be more proactive
influence of free-market ideas, the United States has often in the productive sphere. Second, there is greater recogni-
viewed itself as outside this tradition. Yet it was none other tion, even among economists, that dealing with climate
than one of the nation’s founding fathers, Alexander Ham- change requires interventions in production and investment
ilton, who articulated the earliest and one of the clearest ar- decisions beyond simply raising the market price of carbon:
guments for industrial policy. In his “Report on the Subject subsidizing green technologies is an indispensable compo-
of Manufactures,” Hamilton (1791) presented a powerful nent of a decarbonization strategy. Finally, there is grow-
case for subsidizing and protecting America’s nascent man- ing concern in national-security and business circles about
ufacturing establishments. China’s rise as a technological and business competitor, and
Hamilton forcefully took on his critics’ counterargu- a near-consensus that the United States (and the West more
ments, which still sound familiar today: supporting manu- broadly) needs to respond by reinvigorating innovation and
factures would be a misdirection of resources, would raise industrial capabilities. The last factor is reminiscent of the
domestic prices for industrial goods, would provide an un- manner in which the US government went into technologi-
fair advantage to a specific segment of the economy, would cal overdrive in response to the Soviets’ Sputnik challenge of
result in domestic monopolies, and would be abused by the the late 1950s.
recipients of government support. Hamilton did not dispute It is a good thing that today’s conversation is about the
that other economic activities, and agriculture in particu- how rather than the why of industrial policy. Since govern-
lar, may have been more remunerative in late 18th-century ments always engage in industrial policy, it is desirable that
America. But he maintained that expanding the manufac- they act deliberately and self-consciously, rather than surrep-
turing sector would enlarge the size of the overall economic titiously and without an overall strategic frame. The present
pie: it would “have the effect of rendering the total mass of economic and technological context for industrial policy is
useful and productive labor in a community, greater than it very different—not just from Alexander Hamilton’s day, but
would otherwise be” (original emphasis). Manufacturing had also from the heyday of industrial policy during the 1960s
the benefit, he wrote, of increasing returns to scale, employ- and 1970s. Moreover, we have learned much about what kind
ment creation, and more-rapid technological change. of industrial policies work better. The nature and contours of
The United States has never been without industrial best-practice industrial policies have changed considerably,
policy of some sort. Defense-related industries have always thanks to accumulated experience and knowledge. Present-
been big beneficiaries of government support through pro- day industrial policy in the United States must be shaped by
curement and other means. And even in the heyday of market this new understanding, and should not simply aim to repli-
fundamentalism during the 1980s, President Ronald Reagan cate old models that, regardless of how one views their suc-
actively used protectionist trade policies to prop up segments cess or failure in the past, are unlikely to work well at present.
of US manufacturing. Meanwhile economists have devel- In brief, policymakers must take into account the transfor-
oped a more full-fledged theoretical rationale for industrial mation of the why, what, and how of industrial policy.
policy—based on technological and learning externalities, I will make several arguments in this policy proposal.
and coordination failures—that support many of Hamilton’s • First, industrial policy must target what we might
arguments. Nevertheless, the debate on industrial policy has call good-jobs externalities, in addition to the
traditionally revolved around the question of whether govern- aforementioned learning, technological, or nation-
ments should engage in industrial policy at all, instead of the al security considerations. Increasing the supply of
more relevant (and useful) question of how they should do so. good jobs is an independent and important objec-
Industrial policy is back in fashion these days, as a re- tive in and of itself, and one that will not be met
sult of several developments. First, there has been a general as a by-product of pursuing the conventional ob-
dissatisfaction with neoliberalism and its maintained as- jectives of spurring innovation in technologically
sumption that a hands-off government would produce broad advanced industries or fostering national security.
An Industrial Policy for Good Jobs 1

• Second, and relatedly, industrial policy’s tradition- creation. Such partnerships align with a new, more-flexible,
al focus on manufacturing and globally competi-
and contextual model of industrial policy that is better suit-
tive industries will need to be broadened to service
sectors and smaller and medium-sized firms. Man- ed to the challenge of creating good jobs.
ufacturing is unlikely to generate much employ- The federal initiative would be the establishment of an
ment even if its value-added share in the national Advanced Research Projects Agency (ARPA) focused on the
economy starts to climb; the bulk of good jobs will promotion of employment-friendly technologies: ARPA-
have to come from outside manufacturing. W(orkers). Starting from the premise that innovations that
complement rather than displace workers are feasible yet
• Third, the practice of industrial policy will need currently undersupplied, ARPA-W would promote early-
to rely less on traditional top-down policy instru- stage investments in digital and other technologies that en-
ments—such as subsidies and tax incentives for
hance prevailing worker skills and create good jobs.
firms—and more on collaborative, iterative inter-
action whereby public agencies supply a portfolio Two final introductory remarks about the scope and
of customized public services in exchange for firms limitations of this proposal. First, the policies I describe here
undertaking soft commitments on the quantity and are not sufficient in themselves to create an inclusive, good-
quality of employment. The state cannot systemati- jobs economy; they are restricted to interventions that one
cally pick winners, as the critics of industrial policy would consider as belonging under the rubric of industrial
rightly point out. But appropriately structured gov- policies. They must be supported and complemented by a va-
ernance arrangements can act as an information riety of regulatory, social insurance, and macroeconomic ar-
revelation mechanism facilitating learning about rangements that increase the bargaining and organizational
what works and what fails, allowing government power of labor in the workplace, provide a robust safety net
agencies to abandon failing initiatives and focus on for workers and their families, and ensure adequate levels of
supporting those with the most potential.
aggregate demand to run a tight labor market.1 They are not
I will draw out the policy implications of this approach a substitute for the other remedies and protections that an
by considering two types of specific initiatives: one at the lo- inclusive economy needs.
cal level and the other at the federal level. Second, I do not claim that good jobs must be the only
The local approach would build on existing develop- focus of industrial policies. A robust set of industrial poli-
ment and business assistance programs that are already cies would also target the climate transition (to promote
loosely structured along the lines advocated here. These green technologies), the high-tech and digital economy
are collaborative partnerships between local development (to promote general innovation), and the rebuilding of do-
agencies, firms, and other partners aiming to revitalize lo- mestic supply chains (to create a more resilient economy).
cal communities and create good jobs. They are organized The governance principles I will describe are quite possibly
around an implicit (and evolving) quid pro quo: the provi- relevant to all these areas. But just as traditional industrial
sion of public services (such as business extension services, policies cannot be relied on to serve the needs of a good-jobs
infrastructure, or customized training) in return for soft economy, the specific policies I advocate here do not neces-
commitments by firms on investment and employment sarily fulfill those other objectives.
2 The Hamilton Project • Brookings

The Challenge
The Why: Good-Job indicator on earnings, labor market security, and the quality
of the working environment as subdimensions (OECD n.d.).
Externalities The OECD database allows disaggregation by gender, age
groups, and education as well as comparison across coun-
What constitutes a good job is a subjective and highly con- tries—showing that the United States does worse than many
textual matter. Generally speaking, good jobs are those that others, such as Australia, Austria, Denmark, Finland, Ger-
provide a middle-class living standard, adequate benefits, many, and Switzerland. And in April 2022, the Families and
reasonable levels of personal autonomy, economic secu- Workers Fund launched a collaborative effort with the US
rity, and career ladders.2 Regardless of the specifics of the Department of Labor to measure the quality of American
definition, however, the sine qua non of a good job is a high jobs, called the Job Quality Measurement Initiative (Fami-
enough level of labor productivity. While bargaining power lies and Workers Fund 2022).
can affect the division of enterprise surplus, it is ultimately For workers without college degrees, manufactur-
productivity that enables the provision of adequate wages ing and related services have been the traditional source of
and benefits. As I will explain below, this link between pro- good jobs—the basis for income mobility and a pathway to
ductivity and wages establishes a parallel with the tradition- the middle class. But globalization, deindustrialization, au-
al case for industrial policy, but with a focus on the quantity tomation, and generally skill-biased technological change
and quality of jobs rather than on the profitability or com- have reduced demand for workers who lack higher levels of
petitiveness of firms per se. education, shrinking the supply of such employment oppor-
In practice, the idea of good jobs needs to be operation- tunities. Since the 1980s there has been an evident failure
alized through an evolving set of standards that reflect local of the economy to produce adequate numbers of good jobs
conditions and preferences. Metrics to assess the availability to sustain a prosperous and growing middle class. Medium-
of good jobs can be developed, based on surveys of work- pay jobs have seen an absolute decline while low- and high-
ers’ perceptions or objective statistical criteria. There are, pay jobs have expanded, a phenomenon known as labor
in fact, many such measures. Since 2017, the Boston-based market polarization (see figure 1). The waning of factory and
workforce development agency Jewish Vocational Service office/clerical/sales jobs, along with the weakening of work-
(JVS) has maintained a job quality index, a composite that ers’ bargaining power, has in turn produced a stagnation in
measures wages, benefits, scheduling flexibility and predict- the average real wages of production and nonsupervisory
ability, access to career ladders, and the degree to which the workers and greater economic insecurity (Autor and Dorn
work environment is supportive (JVS n.d.). A 2020 Gallup 2013; Autor, Mindell, and Reynolds 2020; Eurofound 2017;
survey measured job quality based on a weighted average OECD 2019). Another indication of the scarcity of good jobs
of respondents’ satisfaction on 10 dimensions of work, in- is that the American middle class, measured by the share of
cluding “having a sense of purpose and dignity at work” and adults with pretax earnings between the 30th and 70th per-
“having the power to change things that are unsatisfying at centiles, has shrunk significantly. As figure 2 shows, while
work” (Gallup 2020). The survey found fewer than half of other advanced economies have experienced a middle-class
workers are in good jobs, defined as scoring three or more squeeze as well, the downward trend in the United States has
on the five-point combined index. The survey also found been exceptionally dramatic.
that non-pay-related aspects of work are typically more im- The failure to generate good jobs has significant eco-
portant than level of pay when workers judge job quality, nomic, social, and political costs. On the economic side, the
even for those in the bottom quintile of incomes. distributional costs are compounded by the implications
Similarly, the Good Jobs Institute provides a scorecard for overall productivity and economic growth. Labor mar-
for employers, focused on employee basic needs and stabili- ket polarization slows down the dissemination of innova-
ty, to allow firms to understand and track their performance tion from the more advanced sectors and firms to the rest
on this dimension (Good Jobs Institute n.d.). The Organisa- of the economy that often occurs through the creation of
tion for Economic Co-operation and Development (OECD) more-productive jobs in the middle of the skill distribution.
maintains a database on job quality, with objective statistical This deficit of middle-skill jobs may well be connected to the

figure 1
Employment Growth by Occupation and Annual Pay, 1980 to 2019

8
6
Low Pay
4
Change in occupation share
Medium Pay
2
-2
High Pay
-4
-6
-8
Personal Clean/ Laborers Production Office/ Sales Technicians Professionals Managers/
Services Protect Admin Executives
Services
Source: Census Bureau American Community Survey (ACS) 1980-2019; author’s calculations.
Note: Data for 2019 reflects pooled employment and wage earnings from 2017–2019. Sample includes working-ages
16–64, excluding those in the military. Occupations are harmonized across periods and grouped based on the clas-
sification scheme developed by Dorn (2009).
general slowdown in US aggregate economic productivity, a These problems get transmitted from one generation to an-
fact that is otherwise puzzling given the significant rates of other, with joblessness among today’s adults making jobless-
innovation in the advanced sectors of the economy. ness among tomorrow’s adults more likely.
The broader complication with the shortage of good Then there are the political consequences of the scarcity
jobs is the undermining of social structures that underpin of good jobs. There is considerable evidence from a number
economic prosperity. Communities where middle-class jobs of advanced market economies that links the rise of nativist
have become scarce suffer from a variety of social ailments. populist political movements to adverse labor market devel-
In his pathbreaking book When Work Disappears, sociolo- opments (see Rodrik 2021 for a review and discussion). Autor
gist William Julius Wilson (1996) described at length the so- et al. (2017) have shown that, in the United States, the China
cietal costs of the decline in manufacturing and blue-collar trade shock had a significant impact on political polarization.
jobs on racial minorities living in urban areas; those costs Holding constant initial political conditions in 2002, districts
include household restructuring, addiction, and crime. His that experienced sharper increases in import competition
analysis applies more broadly, however. More recently, Au- were less likely to elect a moderate legislator in 2010. New
tor, Dorn, and Hanson (2019, 2021) have studied commu- legislators elected in hardest-hit areas tend to occupy more-
nities across the entire United States, differentiating them extreme positions on the ideological spectrum, especially on
by the degree to which they were affected by import com- the right. Districts initially in Republican hands were sub-
petition with China. Communities where jobs came under stantially more likely to elect a GOP conservative. The China
greatest pressure from Chinese imports experienced long- trade shock on local labor markets may have even been di-
term increases in “idleness” among young men (i.e., neither rectly responsible for President Donald Trump’s electoral vic-
employed nor in school) and a rise in male mortality due to tory in 2016. Autor et al. (2017) undertake a counterfactual
drug and alcohol abuse, HIV/AIDS, and homicide. Job loss analysis in which they assume the growth of Chinese import
also led to an increase in the fraction of single mothers, of penetration is 50 percent lower than the realized rate over the
children in single-headed households, and of children living 2002–14 period. Their estimates for the electoral consequenc-
in poverty. Finally, in their evocatively titled book Deaths es indicate that a Democrat instead of a Republican presiden-
of Despair, Anne Case and Angus Deaton (2020) have de- tial candidate would have been elected in 2016 in the swing
scribed the staggering costs in terms of disease and mortal- states of Michigan, Pennsylvania, and Wisconsin, preventing
ity when economic opportunities desert local communities. Trump from garnering a majority in the Electoral College.

figure 2
Middle Class Pretax Income Shares in Select Countries, 1980–2021

40
38
36
Sweden
34
32
France
Share
30
28 Germany
26
24
United States
22
20
1980 1984 1988 1992 1996 2000 2004 2008 2012 2016 2021
Source: World Inequality Lab 2021.

Note: Figure shows pretax income share going to the middle 30th–70th percentiles for each country.
Similar results showing an association between labor market In short, bad jobs lead to lagging communities with poor
problems and the rise of the authoritarian, nativist right have social outcomes (poor health, inferior education, high crime)
been obtained in several European nations (Colantone and and social and political strife (populist backlash, democratic
Stanig 2016, 2017; Dal Bò et al. 2018; Guiso et al. 2017). malfunction). In the absence of incentives that prompt them
More broadly, adverse labor market developments to do so, private employers fail to take these costs into ac-
weaken support for democracy and foster nativist and au- count.3 These negative externalities can be substantial—per-
thoritarian attitudes. Economic stagnation or decline haps so great that they threaten the economic order under-
among the middle classes undermines the set of moral val- pinning our form of government. Good jobs, conversely,
ues and beliefs that sustain liberal democracy (Friedman have enormous positive externalities. The external costs as-
2005). The association between economic crisis and the sociated with the failure of the private sector to create good
rise of fascism in interwar Europe is well known (Frieden jobs provide a motive for industrial policies that is broadly
2006). There is evidence that some of the same tendencies similar to the traditional economic case for such policies.
may be at play currently. In the United States, individuals The case for industrial policy rests on two fundamental
located in local labor markets that were more substantially rationales: externalities and coordination failures. External-
affected by imports from China appear to have developed ities are costs or benefits that producers’ actions create for
values that are more authoritarian (Ballard-Rosa, Jensen, society at large and that they do not pay or receive a reward
and Scheve 2018). Similarly, individuals living in European for, at the margin. Learning, technological, and agglomera-
regions that received more-negative globalization shocks tion externalities are at the core of the traditional case for
were systematically less supportive of democracy and liberal industrial policy. These externalities occur when, for ex-
values and more in favor of authoritarian leaders (Colantone ample, a firm invests in new technologies that other firms—
and Stanig 2018). Strain and Veuger (2019) find increases in those that are nearby, suppliers, or direct competitors—can
Chinese import penetration are associated with hardening benefit from without having to pay for them. Since the social
of preexisting attitudes among white Americans toward im- benefit of such investments exceed the benefits to the firm it-
migrants, minorities, guns, and religion. Cerrato, Ferrara, self, they would be under-provided in the absence of explicit
and Ruggieri (2018) argue that the political impact of the encouragement. The optimal policy here would be the pay-
China trade shock played out primarily through a cultural ment by the government to the firm of a Pigovian subsidy to
backlash: greater disruption of local labor markets produced internalize the externality, equal to the difference between
negative attitudes toward immigrants and racial/ethnic mi- social and private marginal benefits of research and devel-
norities, including Muslims. opment (R&D).4

For policymakers, national security has always loomed The What: Moving Beyond
large in arguments for industrial policy. Alexander Hamil-
ton argued in his 1791 report that supporting manufactures Manufacturing
“will tend to render the United States, independent on for-
eign nations, for military and other essential supplies.” The The economic rationale for industrial policy is that society
national-security case for industrial policy can also be made benefits when government promotes certain economic ac-
in terms of externalities. A strong national defense, and in tivities—implicitly at the expense of others—because those
particular a dependable supply of critical inputs, benefits all activities are the repository of certain desirable features (i.e.,
firms in an economy. Individual firms may not sufficiently they produce positive externalities). As such, the argument
internalize these objectives, prioritizing the benefits of off- in favor of industrial policy has nothing to do with indus-
shoring and outsourcing. Government inducements may be try per se. But, as the label makes clear, industrial policy has
needed to ensure there is adequate domestic capacity in crit- historically been regarded as synonymous with support for
ical products that foreign nations (i.e., China) may choose to manufacturing. Since Hamilton’s day, manufacturing has
withhold or raise prices on to extract geopolitical leverage. been viewed as special—as a source of positive externalities,
Good-job externalities can be thought of in similar dynamism, innovation, and as a growth engine for the entire
terms, though the external effects in question may be local economy. There is good reason for this belief. Modern eco-
rather than national. Good-jobs externalities exist when nomic growth started with the Industrial Revolution. And,
the private costs to firms of hiring workers with good pay, while advanced economies have become postindustrial, to
benefits, and working conditions exceed the social costs, or this day manufacturing plays an outsized role in R&D and
alternatively, when the private benefits to employers of cre- innovation for the economy as a whole (Fuchs et al. 2022).
ating these good jobs fall short of the social benefits.5 The However, manufacturing today is no longer a plentiful
considerations discussed earlier suggest such gaps can be or reliable source of good jobs, and neither is it likely to be-
quite significant. come so in the future, as I will argue below (see also Law-
Coordination failures arise in the presence of strong rence 2022). These considerations require us to go back to
scale economies (that are in turn due to either declining the original rationale for industrial policy and reconceive it
marginal costs or large fixed costs). A firm considering in a broader sense, deemphasizing the role of manufactur-
making a large investment in a particular region will need ing per se.
to ensure that there is adequate infrastructure, high-quality Remarkably, the share of manufacturing in US GDP at
suppliers nearby, and the requisite specialized skills in the constant prices has remained quite stable since the end of
local workforce. But these critical inputs may not be avail- World War II (figure 3). In real terms, the United States has
able in the absence of the investment from the firm in the not experienced any deindustrialization in output terms. But
first place, because they would not be profitable on their because productivity has advanced more rapidly in manu-
own. A simultaneous package of investments in all these facturing than it has in the rest of the economy, the relative
areas would yield profits to all, though none makes com- price of manufacturing has fallen and the share of manu-
mercial sense in the absence of other complementary invest- facturing at current prices has come down steadily from
ments. Timing discrepancies—firms need specialized skills 26 percent in the early 1950s to around 11 percent currently.
today for which investments ought to have been made yes- There has been an even more striking reduction in the share
terday—add a further layer of complications. In such cases of jobs generated by manufacturing. The proportion of em-
government agencies have the potentially important role of ployment in manufacturing has declined from more than a
coordinating investments by the different parties involved. third during World War II to less than 9 percent at present.
In principle, coordination failures do not require gov- Manufacturing presently supplies less than one in ten total
ernment financial assistance to be resolved. Subsidies need jobs in the US economy. Since 1979, the total number of jobs
not be paid since the private investments will more than pay in goods-producing sectors has fallen by 4 million, while
for themselves ex post. But the promise of a subsidy—such jobs in services have increased by 59 million (figure 4).
as a loan guarantee—may be a useful inducement to get the The decline in US manufacturing employment is close-
investments started. If the logic of coordination failures ly linked to patterns of technological change, which have
holds, the government guarantee will never be called, and been skill- and capital-biased. Since the late 1980s the labor
no subsidy will have to be paid ex post.6 share in US manufacturing has fallen while holding gener-
While job creation is almost always one of the stated mo- ally steady in the service industries (Acemoglu and Restrepo
tives of industrial policy, good jobs are rarely targeted spe- 2019). It is reasonable to suppose that global competition
cifically. It is typically assumed that increased investments from countries, such as China, with low-cost labor has stim-
in physical capital or innovation will bear fruit in terms of ulated labor-saving innovation in manufacturing that has
improved labor market outcomes as well. But such outcomes remained in the United States. Both globalization and au-
are not always assured. The evidence reviewed above sug- tonomous trends in innovation would appear to have played
gests good-job externalities must play an explicit and much a role in driving employment deindustrialization.
more significant role in the design of industrial policy. I emphasized above the importance of good-job ex-
ternalities. Secure, well-paid jobs in manufacturing have

figure 3
U.S. Manufacturing Shares of Total Employment and GDP, 1939–2020

40
35
Employment
Share of total employment and GDP
30
25
Value-added
20
15
10
Real value-added
5
0
1939 1948 1957 1966 1975 1984 1993 2002 2011 2020
Source: Bureau of Economic Analysis (BEA) 2022; Bureau of Labor Statistics (BLS) 2022; Grogingen Growth and De-
velopment Centre (GGDC) 2014; Timmer et al. 2015.
Note: Historical value-added series are sourced from GGDC database based on Timmer et al. (2015), which estimates
manufacturing’s share of total value added aggregated across all observed sectors, where real series are based on
constant 2005 prices. Recent series for 2005–2020 shown as lighter shaded lines reflect BEA estimates of real and
nominal contributions of manufacturing to total GDP, where real series uses constant 2012 prices. Due to sector-level
adjustments, aggregate totals underlying GGDC shares are not fully consistent with national accounts aggregates.
traditionally served as a vehicle for broadening the middle Singapore, South Korea, Taiwan, and Turkey. The chart for
class. From this perspective, employment deindustrializa- each country shows manufacturing employment shares as
tion is especially significant: it raises the question of wheth- well as manufacturing value-added shares in GDP at con-
er the traditional focus on manufacturing can be effective stant prices. Several of these countries have significantly
in generating significant quantities of good jobs when the raised real manufacturing shares in the economy in recent
employment share has already fallen to less than 10 percent. decades, with Japan, South Korea, Malaysia, and Taiwan in
The answer depends on the likely future prospects for man- particular standing out. Nevertheless, none has managed to
ufacturing employment. If successful industrial policy can prevent a noticeable decline in manufacturing employment
boost manufacturing employment alongside manufacturing shares.
output and innovation, we can remain reasonably optimistic South Korea is an interesting example. The output share
about the efficacy of traditional industrial policies. If not, we of manufacturing in the economy has risen (in real terms,
need to look to other parts of the economy and broaden the at 2015 prices) from 20 to 29 percent, which is an increase
definition of industrial policy. of nearly 50 percent. Meanwhile the employment share of
The experience of other countries provides an impor- manufacturing has fallen from 26 to 17 percent. As South
tant clue regarding the prospects for manufacturing em- Korean manufacturing has become more productive and
ployment. The precarious state of US manufacturing is often more competitive in global markets, it has become signifi-
compared to the apparently more encouraging situation in cantly less intensive in labor. In fact, the adoption of labor-
some countries in East Asia where policies and other factors saving technologies such as automation and robots have
are said to have been more advantageous. Yet the compara- been an important cause for South Korea’s continued manu-
tive experience provides very little hope that employment facturing success. The same process appears to be playing
deindustrialization can be reversed. In fact, the picture oth- out in China as well, as Chinese manufacturing firms react
er countries paint is rather pessimistic for the prospects of to rising domestic labor costs.
manufacturing employment. In theory, it is possible that the adverse employment
Figure 5 summarizes manufacturing trends in eight consequences of the substitution of capital for labor could
comparator countries: China, Japan, Malaysia, Mexico, be offset by increased scale (an expansion of manufacturing

figure 4
Private Employment in Goods and Services Sectors, 1939–2022

120
Service-providing sector
100
80
Millions of workers
+59.0m
60
40
20 −4.3m
Goods-producing sector
0
1939 1944 1950 1955 1961 1966 1972 1977 1983 1988 1994 1999 2005 2010 2016 2022
Source: Bureau of Labor Statistics 2022.

Note: Gray bars indicate recession periods.
output as a whole). But there does not seem to be a single

country where this has actually happened.
The How: From Top-Down
The bottom line is that employment deindustrializa- Subsidies to Collaborative,
tion seems to be a universal phenomenon. In fact, greater
success in manufacturing seems to be associated with the Customized Assistance
adoption of labor-saving technologies that contribute to the
fall in employment shares. We simply cannot presume that In his “Report on the Subject of Manufactures,” Alexander
conventional industrial policy, focusing on manufacturing, Hamilton (1791) considered different policy instruments
will serve, as a by-product, the objective of increasing the for promoting industrialization: import duties, import pro-
potential supply of good jobs in manufacturing. Therefore, hibitions, and export prohibitions on key material inputs
the remit of industrial policy will have to be broadened sig- to manufactures, among others. But he especially favored
nificantly beyond manufacturing, to encompass firms in “bounties” (i.e., subsidies) as the most direct means to en-
service sectors serving primarily local or domestic markets. courage new industries without raising domestic prices. He
These firms will generally be smaller and significantly more anticipated economists’ contemporary understanding of
diverse in the types of support they require. The conduct of Pigovian subsidies as the most appropriately targeted in-
industrial policy will have to be modified accordingly. strument when targeting the internalization of externali-
Can productive development policies typically applied ties. He was keenly aware of the practical difficulties. “It is
to manufacturing also be appropriate for service sectors a familiar objection to [bounties], that they are difficult to
such as retail, hospitality, education, health care, or long- be managed and liable to frauds,” he wrote. It would “be
term care? We have less experience and evidence on the necessary to guard, with extraordinary circumspection, the
benefits of sectoral policy in these areas. But many service manner of dispensing them.” But such difficulties, he noted,
activities can benefit from complementary investments in were not prohibitive: “[They are not] sufficiently great to
new work practices, job-specific training, technologies that countervail the advantages of which [bounties] are produc-
complement and empower workers, better-tailored regula- tive, when rightly applied.”7 (Hamilton argued that the sub-
tions, and improved organizational culture. Public-private sidy to manufactures should be financed by import duties
initiatives that promote such investments can enhance labor on manufactures, whereas contemporary economic theory
productivity, enabling the provision of better jobs. See box 1 would favor a broad-based tax.)
for a discussion of an important case: long-term care. A Pigovian subsidy is the appropriate instrument when
information is plentiful and uncertainty is limited. The

figure 5
Manufacturing Trends Across Countries, 1990–2018

China Japan Malaysia Mexico
40
Real value-added share
30
25
Percent
20
15
Employment share
10
1990 1997 2004 2011 2018 1990 1997 2004 2011 2018 1990 1997 2004 2011 2018 1990 1997 2004 2011 2018
Singapore South Korea Taiwan Turkey

40
30
25
Percent
20
15
10
1990 1997 2004 2011 2018 1990 1997 2004 2011 2018 1990 1997 2004 2011 2018 1990 1997 2004 2011 2018
Source: Grogingen Growth and Development Centre (GGDC) 2021; de Vries et al. 2021.
Note: Data is sourced from GGDC’s Economic Transformation Database based on de Vries et al. (2021). Real value-
added shares reflect each country’s manufacturing share of total GDP in constant 2015 prices.
first-best response to a good-job externality is to subsidize clean air) due to technological or other imponderables. In
employers for creating good jobs. However, this assumes a classic article, Weitzman (1974) showed that quantity tar-
that policymakers confront a clear-cut externality of known gets may dominate price instruments (such as a Pigovian
magnitude and well-established elasticities of demand and subsidy) under these conditions. A subsidy minimizes the
supply. Things get murkier when there is uncertainty along costs of achieving a certain target, but creates the risk that
these dimensions. In a dynamic environment with multi- the target may be missed (because firms do not respond as
dimensional uncertainty, subsidies will generally fall short vigorously as anticipated). Quantitative targets, on the other
and be dominated by different policy tools.8 hand, achieve the requisite target (by assumption), but po-
A specific example of this arises in the case of envi- tentially produce greater economic cost than might have
ronmental externalities where there is uncertainty about been necessary. When the risks of missing the socially opti-
the costs and benefits of attaining a particular target (say, mal target—such as reducing air pollution by the mandated

box 1
Good Jobs in Long-Term Care?

Consider the care economy, and long-term care specifically. This is a sector where employment could increase rapidly in future
years as the population continues to age and demand for in-home or assisted living arrangements increases. Yet there is a short-
age of workers, due to low wages and other undesirable characteristics of the job (Poo and Berger 2022; Stevenson 2018).
Much of long-term care work is done in homes (through agencies that provide the caregivers or through self-employed
caregivers) or in assisted living or retirement communities where, unlike hospitals or nursing homes, regulations are weak. In
such settings, remuneration and work conditions have traditionally been very poor—characteristics that epitomize bad jobs.
Employees are mostly women and disproportionately are people of color. Long-term care workers are typically regarded as
performing low-skill jobs and are not viewed as “real members of the care team” (Osterman 2019). Home-care aides are unable
to undertake simple tasks (such as changing dressings or bandages, or applying nasal or eye drops), so that “it is difficult to in-
crease the productivity of aides in a way that might underwrite compensation gains” (Osterman 2019). In many ways, long-term
care is a test case for the idea that sectoral policies organized along the lines of industrial policy can boost good jobs.
As Osterman notes (2019), there are three ways in which jobs in long-term care can be improved. First, the government can
regulate and impose standards (such as high minimum wages). Second, reimbursement rates from Medicaid and Medicare can be
increased, in the hope that higher rates show up in increased wages. Third, the productivity of direct-care workers can be raised,
allowing the long-term care system to serve patients’ needs better and to reduce costs, generating room for better compensation.
While both of the first strategies might be useful, greater productivity is ultimately the most reliable source of better jobs.
Osterman suggests that it might be useful to increase productivity in long-term care through a strategy that is analogous to
the deployment of innovations in manufacturing pioneered by Japanese auto producers. This entails a combination of invest-
ing in worker skills; providing workers with greater voice, discretion, and autonomy; and giving them more responsibility for the
quality of the service. Bishop (2014) notes that, with respect to nursing homes, care workers that are empowered with greater
autonomy and decision-making can use their knowledge of residents and patients to customize their services and provide more
flexibility (e.g., in schedules, food, and treatment); the same approach can be applied to in-home care and long-term care facili-
ties as well. An important component of the strategy would be the introduction of new technologies that complement caregiv-
ers’ skills, such as digital tools that enable caregivers to collect real-time information, and to respond quickly and efficiently to
the needs of individual residents.
If long-term care is managed better in these ways, there would be real productivity benefits: lower turnover and burnout
on the part of care workers, reduced admissions to nursing homes and reduced hospitalization rates, better management of
chronic conditions, and quicker and smoother transitions out of acute-care facilities, in addition to improved customer satisfac-
tion (Osterman 2019).
These changes would require a willingness to experiment with novel work practices and a continuum of efforts—from R&D
and the introduction of new technologies for long-term care, on the one hand, to their local adoption, adaptation, and contextu-
alization in specific communities, on the other. Importantly, there is a complementarity between the provision of good jobs (with
high wages and job security) and the adoption and successful implementation of these practices by workers (Bishop 2014).
Workers who believe they are remunerated well and are treated with dignity and respect are more likely to respond positively to
additional responsibilities and to perform at a high level.
amount—outweigh the risks of imposing too large a cleanup direction of future technological change itself can become a
cost on producers, quantity targets are preferable to Pigov- target for policy. Moreover, learning about what works and
ian subsidies. In the good-jobs context, the analogous argu- what does not work becomes an integral part of the policy
ment would be that employment targets may be preferable process. Mechanisms of feedback from firms to public au-
if the risk of failing to generate a sufficient number of good thorities is critical to the regulatory apparatus. The relevant
jobs in a particular community dwarfs the risk of imposing policy space is of much higher dimensionality.
too high a burden on individual firms. Finally, an additional problem with standard regula-
Uncertainty also increases the dimensionality of the tory remedies in the present setting is that they postulate
policy space. In the standard conception of externalities, clear goals (or objective functions, in economics jargon). As
there is a single quantity (level of employment), with an uncertainty increases, it becomes difficult to specify in ad-
associated market price (wages), that is responsible for the vance not only the costs and benefits of regulation, but also
generation of the externality. The appropriate intervention its precise objectives. The government and its agencies will
consists of directly targeting that price (or quantity), and often have to go further and negotiate improvement targets
doing no more than that. But when there is uncertainty with individual firms or clusters of firms. What is a good job
about behavior, technology, and the effectiveness of differ- in a particular community? How many of those jobs can be
ent policies, optimal policies—in the second-best sense of reasonably created? How will technological and other firm-
the term—will extend over multiple margins of interven- level choices influence job creation? What are the comple-
tion and several different types of policy instruments. For mentary policy levers that are available? How can that set
example, policymakers may combine employment incen- of instruments be expanded? These are necessarily local,
tives with training, technology, and marketing assistance contextual questions. They can be answered, and periodi-
to firms; investments in infrastructure; and so on. The cally revised, only through a customized, iterative process

table 1
Traditional Industrial Policy and Proposed Approach

Traditional Industrial Policy Proposed Approach
Externalities targeted R&D, innovation, learning externalities; coordination failures in Good-job externalities in local economic development and
investment direction of innovation
Sectors Manufacturing, tradable sectors Largely services (in addition to manufacturing)
Firms Large, globally competitive firms Small and medium-size firms
Assumptions about the government Governments can identify market failures ex ante and are Widely dispersed knowledge about location and magnitude
sufficiently insulated from capture of market failures; governments face stubstantial uncertainty;
endogenous state capacity
Types of incentives Tax, credit subsidies A portfolio of business services, including marketing,
management & tech assistance, customized training,
infrastructure, seed capital/loans for directed technologies
Application of incentives Fixed schedule of incentives, except for incentive packages for Customized to firms’ needs and adapted to context
large firms which may be negotiated
Selection criteria Pre-specified Voluntary buy-in and participation
Conditionality Hard; rigid ex-ante criteria Soft; provisional, open-ended and evolving
Relationship with recipients Arms’ length Collaborative, iterative; active project management
of strategic interaction between public agencies and private communities. Under this conception, the government is not
firms. This process does not quite fit the familiar, principal- presumed to know where the market failures are beforehand
agent framework of rulemaking which assumes that goals and, therefore, does not determine ex ante what the spe-
and social benefits must be known in advance if public ac- cific policy instruments are. Industrial strategy consists of
tion is to be effective and accountable.9 a collaborative process of discovery involving business and
All these problems are particularly severe in the case of agencies of the state, where the objective is to identify the
services and when working with small and medium-sized constraints and opportunities over time, and to design in-
enterprises with very heterogenous needs. Under extreme terventions appropriately. As learning takes place, policies
uncertainty neither the policymaker nor employers have are revised, refined, and sometimes reversed.
reliable information on the possibilities and costs of creat- This kind of industrial policy diverges sharply from stan-
ing good jobs, and have only vague conjectures regarding dard conceptions of top-down, arms’ length, ex ante regula-
the possibilities that may open upon further investigation. tion that is built in to the Pigovian subsidy model. It reflects
Incentivizing desired private sector responses under these ideas that have developed over the past couple of decades
conditions requires the creation of an information exchange into a new conception of industrial policy (e.g., Evans 1995;
regime that ties ongoing specification of goals—here, good- Fernández-Arias et al. 2016; Ghezzi 2017; Hausmann, Rodrik,
job creation—to continuing exploration of new solutions. It and Sabel 2008; Rodrik 2007, 2008; Sabel 2007). It bears close
would be based on provisional goals, iterative benchmarks, similarity to experimental governance, as explored by Victor
collaborative decision-making, monitoring, and revision and Sabel (2022) for climate policy. Some of the key differenc-
of goals and instruments in light of new information. The es from the traditional model are summarized schematically
contours of public action would be shaped over time by in table 1, and will be discussed further below.
the needs and requirements of different types of firms and

The Proposal
I have emphasized in the previous discussion that public The conventional narrative around the labor market
action must address the good-job externalities associated implications of new technologies goes something like this:
with employment, production, investment, and technology “New technologies make rapidly increasing demands on
choices that firms make, and that such action must move skills needed on the job, and workers need to adjust through
beyond manufacturing and proceed in the face of deep increased education and continuous training.” This perspec-
uncertainty. These considerations highlight two key areas tive treats the direction of technological change—whether
for public action. At the local level, we must encourage the it augments or replaces labor—as essentially exogenous and
proliferation and expansion of a new type of collaborative out of our control. It is workers and society at large that
business development arrangements that explicitly target have to adjust to technological change—not the other way
good-job creation within communities. At the federal level, around. This is a curiously one-sided view. As the late An-
we must launch an R&D and innovation program, along thony Atkinson emphasized, the direction of technological
the Defense Advanced Research Projects Agency (DARPA) change is not autonomous and cannot be left to firms and
and ARPA-E(nergy) models, that focus on developing new innovators alone (Atkinson 2015). This argument has been
technologies that are labor-friendly (i.e., complementary to picked up more recently by Daron Acemoglu (2019; see also
workers with ordinary skills) rather than those that are la- Korinek 2019, and Rodrik and Stantcheva 2020).
bor-displacing: ARPA-W.10 The direction of technological change responds to eco-
Since ARPA-type organizing principles can be general- nomic incentives and prevailing social norms. First, and
ized and applied to local initiatives as well, I will begin with most directly, government-funded and government-directed
a discussion of the federally-focused program. As I hope will innovation programs make decisions about what kind of in-
become clear, the two sets of initiatives are different in scale novations to promote. Defense-related and green technolo-
and scope, but they are connected both by their objective— gies are clear examples, promoted by DARPA and the much
expanding the supply of good jobs—and by a new approach smaller ARPA-E, respectively. Employment-friendly technol-
to industrial policy that is collaborative and iterative rather ogies—those that augment rather than replace labor—could
than top-down and prescriptive. The successful undertak- be part of those priorities, though they are not at present.
ing of these tasks does not rely on unrealistic assumptions Second, private sector innovation incentives can be skewed
on policy capabilities. These initiatives can be built on ar- because of prevailing financing methods or policies. Venture
rangements that are tested and that already exist at the local capital, for example, naturally seeks areas where the returns
and national levels. can be capitalized relatively quickly by investors. This may
exclude innovations where the gains are longer term or are
A Federal Innovation Initiative reaped by society at large (Lerner and Nanda 2020).
Third, prevailing tax regimes shape innovation incen-
to Promote Employment- tives. Most advanced economies subsidize capital formation
(through depreciation allowances and other incentives) and tax
Friendly Technologies labor (through personal income taxes and labor charges). An
unintended consequence of the tax system is to induce firms
Initiatives to incentivize firms to create good jobs and pro- to economize on labor by investing in machinery, to an extent
vide them with the complementary inputs to facilitate doing that may be socially suboptimal (Acemoglu, Manera, and Re-
so, in the manner I will discuss in the next subsection, will strepo 2020). Fourth, global competition also alters innovation
not be very effective if technological progress continues to incentives. Increased competitive pressure from labor-abun-
displace workers with middle skills and education. To en- dant, low-wage countries has accelerated labor-saving innova-
sure that technology helps rather than hampers inclusive tion in the advanced countries, as I mentioned earlier.
prosperity, a key plank of a good-jobs strategy must be a Fifth, beyond economic incentives, there are informal
parallel national effort to steer technology in a more labor- norms that guide innovators’ decisions. The high-tech com-
friendly direction. This could be viewed as a moonshot mis- munity often operates under a shared set of values and ex-
sion, in Mariana Mazzucato’s (2021) sense of the term.11 pectations with respect to what is a desirable direction for
technological change. Groupthink is aggravated by the very

high concentration of venture capital funding in a small national innovation effort along the ARPA model with an
number of firms and cities (such as San Francisco, Boston, emphasis on the development of labor-friendly technologies
and New York City). Such norms might be amenable to such as those discussed above—technologies that comple-
change as society begins to attach specific value to employment rather than displace workers who have an intermedi-
ment-friendly technologies. The growing ecological con- ate range of skills and education. Like DARPA and ARPA-E,
sciousness of households and firms provides an apt analogy. this new innovation program—ARPA-W—would target the
Finally, the direction of technological change also de- development of new technologies at the frontier of science
pends on the balance of power between employers and em- and technology, where solutions are unclear and many ef-
ployees. When workers have a say in the workplace, manage- forts will necessarily fail. But it would focus on technologies
ment has to get buy-in from them before major technologies that are best suited to complement human labor and ingenu-
are deployed and work is restructured. This can result in a ity, and not on technologies that are potentially critical for
modern version of Luddism—aversion to any kind of innova- national security. The appropriate quantitative targets and
tion that appears to threaten jobs. But it can also be a useful benchmarks for such an effort are yet to be developed, and
counterweight to adverse incentives in the system encourag- would have to be part of the initiative. But an initial check-
ing too much automation or the adoption of what Acemoglu, list might prioritize technologies that complement and aug-
Manera, and Restrepo (2020) call “so-so technologies.” ment what workers with intermediate-range skills can do,
In short, the direction of technological change, in ad- expand the range of tasks they can perform, increase their
dition to its rate, depends on a wide range of factors, many ability to customize services to specific needs and types of
of which could be influenced by societal and governmental customer demand, and increase their (labor) share of value
decision-making. It might be possible to direct technology added in production.12 Many of these yardsticks can be mea-
so that it better serves the existing workforce’s needs, in ad- sured and monitored. The overarching objective would be to
dition to preparing the workforce to match the requirements allow workers to do what they cannot presently do, instead
of technology. of displacing them by taking over the tasks that they already
Acemoglu and Restrepo (2018) have argued that it is perform.13
possible to countervail present technological trends and To see the operational implications of such an approach,
push innovation in a direction that creates new, labor-ab- it might be useful to review the governance mechanisms of
sorbing tasks. They cite three areas for potential application. the DARPA model. DARPA’s projects are managed by pro-
First, they suggest artificial intelligence (AI) could be used gram directors—recruited from academia, industry, or
in education in order to create more-specialized tasks for elsewhere in government for three- to five-year terms with
teachers, personalize instruction for students, and increase the possibility of renewal—who play a key role in the col-
effectiveness of schooling in the process. Indeed, computer- laborative setting and revision of goals. DARPA treats goals
assisted learning software that automatically adjusts content at every stage in the organization of research as provisional,
to students’ needs is one of the most promising avenues for or subject to revision in light of experience. Program direc-
productivity enhancement in education (Biasi, Deming, and tors are initially hired largely on the basis of their promise in
Moser 2022). Pairing such software with the human connec- giving direction to an emergent area of investigation. They
tion and support that only teachers can provide could pro- develop program goals in consultation with scientists and
duce greater demand for customized instruction as well as businesses. If approved, a program would consist of a num-
higher-quality education. ber of different projects over several years. Concrete propos-
Second, they note a similar potential in health care. AI als, many of them submitted by partnerships among several
tools can significantly enhance the diagnostic and treatment organizations, are developed and executed in the same it-
capabilities of nurses, physicians’ aides, care workers, and erative manner, with goals open to recurrent challenge and
other medical technicians, allowing the less-skilled prac- revision. Critically, program directors and award recipients
titioners to perform tasks that only physicians with many agree on precise milestones that will structure relationship
more years of professional education have traditionally un- between the agency and awardee partners. Milestones are
dertaken. Third, Acemoglu and Restrepo (2018) mention the not rigid: they are altered or deleted in a large proportion
use of augmented and virtual reality technologies in manu- of the projects, signifying a change in the direction of the
facturing, enabling humans and robots to work together in research investigation.
performing precision tasks, rather than the latter replacing Awardees must provide quarterly progress reports,
the former. Such technologies are based on smaller, more- which the agency rates with a traffic light system: red for
nimble robots that also enable greater customization of projects that missed a critical milestone and are at risk of
production in response to specific customer needs. Indeed, failing, yellow for projects that missed a milestone but can
companies such as BMW and Mercedes are building their be expected to recover, and green for projects that are on
automation plans around human work, which they have track to reach their goals. Project budgets can be increased
found allows both greater reliability and more customiza- (if interim results are especially promising) or decreased
tion in production. (when key milestones are persistently missed). Red rat-
These considerations suggest some broad directions ings—or the anticipation of them—set off an intensification
for policy. In particular, one can imagine the launch of a of the monitoring process (routine site visits, conference

calls, meetings, and written analysis of problems and pos- say, ‘This is what a partnership means.’ In Greenville, South
sible solutions by which program directors keep abreast of Carolina, the public school system includes an ‘Elementary
project developments). If recovery efforts fail, the program School of Engineering,’ in a poor neighborhood. The city
director sends an “at risk” letter warning of the possibility of runs the school; local industries including GE, BMW, and
termination, usually between one and two quarters before a Michelin send in engineers to teach and supervise science
project is in fact ended. fairs, at the companies’ expense. In little Holland, Michi-
The DARPA example illustrates in detail the institu- gan, a large family-owned scrap-recycling company works
tional mechanisms through which public entities can ori- with the state correctional system to hire ex-convicts who
ent, coordinate, and discipline collaborative investigation in would otherwise have trouble reentering the work force. In
an environment of extreme uncertainty. These mechanisms Fresno, California, a collaboration among the city, county,
could be readily adapted to the ARPA-W proposed here.14 and state governments, the local colleges and universities,
Indeed, the DARPA model can be seen as proof of concept and several tech start-ups trains high school dropouts and
that the new style of industrial policy discussed in this es- other unemployed people in computer skills” (Fallows and
say is not only feasible, but is already in existence. Moreover, Fallows 2018).
while a DARPA-style innovation program may seem quite The specifics differ, but in each of these cases these part-
distant from the local and regional industrial incentives nerships succeeded in engineering complementary invest-
I will discuss next, the broad governance principles that ments in the workforce, infrastructure, business incentives,
shape DARPA’s operation are in fact not too dissimilar from public spaces, and the quality of life. Indeed, by forming
those that prevail in successful public-private collaborative coalitions of businesses, workers, and agencies to collabo-
arrangements at the local level. I turn to those next. rate towards shared goals, they were experiments that bore
significant similarity to the type of industrial policy I have
Local and Regional Industrial described above, even if they were designed less self-con-
sciously. Their overseers were practical men and women re-
Policies sponding to immediate challenges. “Had a politician labeled
[this strategy] ‘picking winners’ or ‘industrial policy,’ it
When the husband-and-wife team of James and Deborah would have been stillborn,” remarks James Fallows. But “as
Fallows (Fallows and Fallows 2018) flew around the country a series of ‘public-private partnerships,’ it is a source of civic
in their single-engine prop airplane to survey the national pride” (Fallows and Fallows 2018).
economic landscape in the aftermath of the China shock In a similar vein, Tim Bartik (2020), a long-time observer
and the global financial crisis, they discovered a range of and analyst of local business incentive programs, has argued
local experiments in economic development that had many that policies that work best to enhance local employment are
of the features of the new industrial policies described those that focus on the specific needs of firms rather than on
above. The cities they visited had been ravaged by a variety subsidies or tax incentives. He finds that “public services to
of shocks: “For Greenville and the surrounding communi- businesses, such as customized job training or business ad-
ties of the Upstate [South Carolina], it had been the rapid vice or infrastructure,” can be more effective and much less
loss of the textile companies that even in the early 1990s costly per job actually created than subsidies and tax incen-
had been major employers. For St. Marys and its environs tives. I argue that a similar approach is a more efficient meth-
in the coastal Georgia pinelands, it was the sudden disap- od of creating good jobs. Because they are customized, such
pearance of the region’s dominant industry—followed, for- programs also need state and local agencies to work closely
tunately and thanks mainly to Georgia’s political pull, by with firms and a range of other partners in the community
the arrival of the U.S. Navy. For northeastern Mississippi, in order to be devised and delivered appropriately.
the departure of low-wage workshops like the toilet-seat Many of the examples that Fallows and Bartik describe
factory and blue jeans plant compounded economic and come from manufacturing. But the general principles also
social problems that were decades—even centuries—in the apply to service sectors on which productive jobs programs
making.” But each of these areas had found ways of creat- will need to focus. Economic transformation requires set-
ing public-private partnerships that rebuilt downtowns, ting goals; partnering with firms, workers, training facili-
created parks and cultural centers, reinvigorated training ties, and community groups; mobilizing resources and in-
programs, and promoted new businesses and employment puts; and collaborating in a cumulative, iterative process
creation. “The phrase ‘public-private partnership’ refers to toward those goals.
something real,” they wrote, somewhat to their surprise. More concretely, we can envisage the enactment of
James Fallows continued, “Through the years, I had heard federal framing legislation that allocates funds and pro-
about ‘public-private partnerships’ but had thought of this mulgates general principles for the operation of local good-
as just another slogan. If it meant anything at all, it was jobs program along the lines sketched below. The approach
probably a euphemism for sweetheart deals between big would be similar to the American Rescue Plan “challenges,”
government and big business—the ‘public-private partner- where $3 billion was allocated to a small number of “chal-
ship’ to build the latest fighter plane, for instance. In suc- lenge” programs, with regional and good-jobs challenges
cessful towns, people can point to something specific and being the two largest ones. Under these challenges, local

groupings are invited to submit bids to receive funding from cooperative production of the information that would al-
Washington. As proposed here, the legislation would em- low more-refined (and more-realistic) goals and standards
power existing state and local agencies as well as new local to emerge over time. Penalties—exclusion from public ben-
partnerships formed for this specific purpose to act as or- efits—would be imposed only for failure to report or for
ganizational hubs for federal-supported initiatives. The call fraudulent reports, or for persistent failure to achieve results
would be open to local officeholders (mayors, say) as well as whose feasibility is demonstrated by the attainments of oth-
ad hoc coalitions of civic organizations or public-private co- er firms in like positions. As new information is generated,
alitions. An important objective here would be to incentiv- the range and type of policy instruments—the public inputs
ize and mobilize local groups to take the lead, via the carrot provided by government agencies—might also be adjusted
of federal financial resources. Fostering a certain degree of accordingly. The process would proceed incrementally and
competition among different groups of civic patriots to be require monitoring, to ensure the required revisions in goals
selected to receive financial support from the federal gov- and instruments are undertaken.
ernment might act as a useful counterweight to the inertia A benefit of these voluntary arrangements over the
of existing power structures. Participation by worker and medium term might be the development of an inventory of
community organizations would be essential both to pre- good practices—a repertory of contextualization measures
vent firm capture and to ensure appropriate coordination variously suited to a wide range of settings—that can even-
among service providers down the line. tually guide application of the good-jobs strategy to a larger
These local counterparts (policymakers) might start by set of firms, cutting the costs and increasing the chances
establishing an ambitious, open-ended outcome, such as the for early successes of broader coverage. Put differently, the
creation by a certain date of X number of good jobs. Good initial, selective projects might serve as a pilot program for
jobs can be operationalized provisionally using any of the a new system of regulation that extends the formal obliga-
metrics mentioned previously, or through indicators select- tions of firms. As such, these arrangements could come to
ed locally. If selected by the federal program, policymakers resemble the European regulatory model, with a uniform
would announce a portfolio of federally-supported public requirement of participation but responses highly differenti-
services and inputs—ranging from customized training and ated by locale.
tax incentives to infrastructure and technology assistance— A deliberate quid pro quo underlies the policymakers’
to which firms willing to contribute to that outcome would relationship with the private sector in the type of industri-
have access. The policymakers would also designate a full- al policy I have just sketched out. Governments need firms
time economic development official and supporting staff to internalize good-jobs externalities in their employment,
tasked with managing the program. The official would play training, investment, and technological choices. Firms, for
the role of the program director in ARPA-type agencies. their part, need access to a stable, skilled workforce; to re-
Ideally, the official would have the high-level political sup- liable horizontal and vertical networks of suppliers; and to
port needed to work with diverse local agencies and coor- a variety of public inputs ranging from infrastructure and
dinate their activities so as to respond to opportunities and technology to contractual and property rights enforcement.
participating firms’ needs in a timely fashion.15 Meanwhile deep uncertainty precludes simple solutions
Firms that receive public assistance would be obligated such as Pigovian subsidies. Instead of open-ended tax incen-
to make plans to achieve the goals and to regularly report tives or subsidies, the conduct of industrial policy must then
their results. They would be subject to soft conditionalities. rely on the provision of customized public inputs through
In return for public services, firms would be asked to make collaborative, iterative dialog with firms, and with soft con-
provisional commitments on specific quantities of good jobs ditionality on employment quantity and quality.
they will create at different qualification levels (i.e., low-sal- Such a framework might seem idealized and impracti-
aried employees, medium-salaried employees, etc.). Firms cal, at least in the US context. Nevertheless, there is evidence
would be encouraged to pool proposals when they make not only that similar programs already exist, but also that
use of common inputs—as would be the case for workers they work much better than throwing tax incentives at in-
with particular skills or infrastructure. Other condition- vestors. The United States already has a variety of programs
alities might be included as well. A firm might be asked to at the federal, state, and local levels—DARPA-type initia-
work with local suppliers to improve upstream management tives, manufacturing institutes, local business incentive/de-
or technological capabilities. Or a firm that is consider- velopment programs, and sectoral workforce training pro-
ing outsourcing part of its production to a foreign country grams—that exhibit attributes of the design principles just
might be asked to delay doing so for a number of years, in discussed.16 Many of these are improvised responses by local
case productivity improvements at home render those plans development officials and civic organizations in response
unnecessary. The firm could be encouraged to arrange for to the challenges of economic decline. Fallows and Fallows
additional training for some of its employees or implement (2019) present many illustrations and case studies.
new complementary technology to make workers more While much of the action with this kind of industrial
productive. policy takes place at the local level, the federal government
Penalties for missing targets would aim not to de- also plays an important role. First, through the framing leg-
ter infraction of clear rules, but rather to incentivize the islation the federal government would relax local budget

constraints by providing grants to localities that engage in direction might help, too. For example, experience accumu-
these types of programs. This already happens to some ex- lated locally with good-jobs policies might help ARPA-W
tent such as under the American Rescue Plan challenge pro- refine its operationalization of labor-friendly technologies.
grams. But the scale of funding for this kind of industrial In Rodrik and Sabel (2022) we proposed a more ambi-
policy is small, both in terms of the magnitude of American tious national agenda for industrial policies that builds on
Rescue Plan itself ($1.9 trillion) and in relation to other pol- local efforts. That agenda, which is broader and somewhat
icy initiatives under consideration in Congress, which focus different than the one described above, has four steps, which
largely on innovation, manufacturing, and strategic indus- I summarize here.
tries.17 Currently the bulk of public resources are spent on First, the federal government would commit through
business tax incentives.18 The annual spending on these in- legislation or other means to prioritize good jobs as a fun-
centives, mostly at the state and local levels, stands at more damental challenge that threatens the foundations of our
than $46 billion, while the corresponding figure for custom- democracy—in the same way that the climate change chal-
ized training and business extension services is a meager lenge endangers our physical universe—and requires con-
$1 billion (Bartik 2019). There is significant scope for ramp- certed cooperation between regulators, service providers,
ing up and reallocating federal fiscal resources toward the and private actors. The federal government would mandate
right kind of programs. regulators with relevant authority to put in place informa-
Second, there is a federal role in establishing a national tion-generating regimes that allow for standard setting and
platform for local experiments in industrial policy. These revision. The same legislation would create an interagency
experiments have remained largely under the radar screen body to periodically review and prompt improvement of
of economic policymakers and analysts. There is no system- regulatory responses, resolve coordination problems arising
atic evidence base to facilitate information exchange among from them, and provide funds and authority for voluntary
local policymakers and learning for analysts. A national programs in anticipation of an eventual, step-wise extension
platform for local industrial policy efforts might help bridge of regulatory reach.
these informational gaps, disseminate local learning and ex- In the second phase, agencies with authority over areas
periences, prevent adverse spillovers across localities (as in directly affecting job abundance and quality—vocational
the case of tax competition), help scale up industrial policy training, agricultural and manufacturing extension, stan-
efforts, and promote better practices through systematic re- dard setting, and the like—would introduce innovation-in-
porting and evidence. ducing and contextualizing governance mechanisms where
Third, existing federal programs in related policy ar- these are not already in place, anticipating the need for sup-
eas need to be better aligned with local efforts. The federal port services to help vulnerable actors comply with increas-
government maintains a vast infrastructure of initiatives ingly demanding requirements. The requirements could
in workforce development, technical advice, credit, and as- take different forms, including specific employment quan-
sistance to small firms. These initiatives can provide valu- tity targets and/or standards.
able resources at the local level. But even though the federal In the third step, where current regulatory author-
government generally encourages experimentation in these ity does not reach, the government would create volunteer,
areas, it tends to do so within organizational silos that op- public-private programs to advance the frontiers of technol-
erate on one domain of policy at a time. Even when the si- ogy and organization, and to provide support services (in-
los are designed so as to give control over policy design to cluding perhaps subsidies) to help firms move beyond their
the local level, they do so in a single domain. This gener- current low-productivity/low-skill position. These programs
ally runs counter to the open-ended approach advocated would combine services to workers as well as to firms, and
here and makes coordination across these domains more they would be customized to the needs of particular sectors
difficult. Overcoming the resulting local fragmentation in and locales. They would adhere to the design principles of
policy requires a parallel effort at coordinating the federal innovation-inducing governance; their performance would
initiatives.19 be accordingly reviewed, and their goals adjusted, by the re-
Fourth, the federal government can establish interme- sponsible agency, and then, if problems persist, by the inter-
diate institutions that sit in between local development ef- agency body.
forts and the national technological mission that is focused Finally, and assuming voluntary arrangements are suc-
on labor-friendly technologies, as described in the previ- cessful, the scope of these practices would be gradually made
ous section. The purpose of these intermediate institutions obligatory for nonparticipating firms, starting requirements
would be to facilitate the local deployment and diffusion of for submitting credible plans for improving the quality and
the fruits of national R&D efforts. Working as incubators, quantity of jobs together with competitive position by better
demonstration projects, or technology extension services, organization, use of skill and technology, where appropriate
such arrangements might help local development officials in coordination with other firms and institutions. Penalty
and employers implement innovations that enable produc- defaults might be imposed on laggard firms that, despite the
tive jobs. They might disseminate, for example, new AI availability of support services, persistently fail to comply.
techniques in education, health care, or long-term care that
enhance labor skills. Information flowing in the opposite

Questions and Concerns
The traditional image of industrial policy that economists carefully about the design of industrial policy, it is a lot less
carry in their head goes something like this: A group of decisive than it might seem at first sight.
planners selects certain sectors (e.g., supersonic transport, To begin, the “governments cannot pick winners” argu-
advanced electronics, or biotech) for promotion on the basis ment is largely irrelevant. In the presence of uncertainty, it
of some ex-ante analysis about the productivity- and growth- is inevitable that some projects backed by the government
promoting properties of these activities. Then they devise will fail. In this respect, the government is no different from
financial incentives to encourage investment and innova- the private sector. The relevant question is whether enough
tion in those sectors. These incentives might take the form of the projects backed by the government will succeed and
of cheap credit, loan guarantees, tax incentives, provision produce the social surplus to pay for the failures (and more).
of infrastructure, or more-direct subsidies. The programs What matters is how the overall portfolio does, and not how
might also contain a degree of conditionality: state agencies individual projects do. This is, of course, a point that every
would establish certain performance requirements (physical investor operating in a high-uncertainty environment, such
investment, employment, exports, etc.), and the incentives as venture capital, understands well. It is not unusual for
would be phased out for firms/sectors that do not meet the just a few investments in the portfolio to produce very high
performance requirements. This is how the industrial poli- returns, while hundreds of others fail. A similar result has
cies of successful practitioners such as China, Japan, South been demonstrated in programs ranging from Fundación
Korea, and Taiwan are supposed to have operated. Chile (in Chile) to the US Department of Energy (DoE)
Critics of industrial policy marshal their objections green technology loan programs, once they are evaluated as
against the background of this image. The objections take a portfolio (Rodrik 2014).
two forms: inadequate knowledge and political capture. An important implication is that, under an optimal
Governments cannot pick winners, since they lack the req- strategy to maximize social surplus, a public investment
uisite information about which activities are the actual re- portfolio will necessarily include projects that turn out as
positories of positive externalities. And even if they had ad- failures ex post. It would be a sign of inadequate ambition—
equate information, once governments get into the business suboptimal effort—if all individual projects were to succeed.
of picking winners, they inevitably become hostage to pri- As Thomas Watson, the founder of IBM, is said to have ad-
vate special interests, with lobbying and rent-seeking divert- vised his managers, “If you want to succeed, double your er-
ing industrial policy from its true objective. The East Asian ror rate” (Rodrik 2014).
cases are regarded, grudgingly if at all, as exceptions to the The critical criterion therefore is not to avoid mistakes
rule. The critics argue that other countries are unlikely to in picking winners, but to ensure that public agencies have
replicate the East Asian experience—because they are de- the ability to let losers fail. This is where the information-
mocracies, lack “hard states” that are effective and can bend generating mechanisms, iterative reviews, milestones, and
firms to their will, have lower bureaucratic capacity, are so on of “new” industrial policies come in. They are meant
more diverse societies, or any number of other differences. to develop and fortify the capacity of policymakers to learn
In reality, the practice of industrial policy in East Asia what is working and what is not, and to revise their supports
was a lot less top-down than in the traditional image and in light of the information. Capabilities to administer such
much more collaborative in the manner described previous- programs effectively do not fall from the sky ready-made.
ly. The sociologist Peter Evans (1995) has called that model State capacity is endogenous: it is built over time through
“embedded autonomy”: state agencies had the independence practice.
from private firms to exercise discipline when needed, but Ultimately, it is an empirical matter whether industrial
they also had plenty of dynamic interaction and collabora- policies succeed. Most contemporary programs of industrial
tion with them to have access to the information necessary support have some mechanisms of review whereby firms
to design and revise support policies as needed. that are not performing in line with program expectations
Nevertheless, the critics’ points should make us wary of are cut off from assistance. Sometimes firms can fail in spec-
how much industrial policy can accomplish or the damage it tacular fashion. A much-publicized example is Solyndra, a
can cause—especially if the skepticism is grounded in actual solar cell company that was the beneficiary of $535 million
evidence. But while the criticism is useful to make us think in DoE loan guarantees before going bankrupt in 2011. The

company had been visited and praised by President Barack inference. Statistical issues with mismeasurement, omitted
Obama the year before and was showcased as a symbol of variables, and possible reverse causality plague many of the
economic growth based on green technologies. Solyndra’s early studies of industrial policies (Rodrik 2012). Recent
technology relied on non-silicon inputs; even though the studies have produced more-meaningful results, paying
technology worked, the commercial case for the company careful attention to these issues. The results paint a very het-
relied on silicon prices remaining high. Once silicon price erogeneous picture.
tumbled after 2008, the company had no chance to compete In the United States, enterprise zones, which are gen-
with conventional silicon-based photovoltaic cells. It looked erally nondiscretionary and provide tax incentives for all
like the government had backed the wrong horse. The bank- firms operating in a designated geographical area, do not
ruptcy produced significant political embarrassment and a generally have a good record in promoting investment and
search for culprits. Solyndra’s offices were searched by FBI employment. Wessel (2021) provides a recent exposé of
agents, and the company’s top executives were hauled before the most recent version of these zones, called Opportunity
Congress (where they took the Fifth Amendment). Zones; Wessel documents how these zones have been open
Solyndra is a case study of what can go wrong when prog- to abuse and manipulation and are mainly a vehicle for en-
ress—or lack thereof—is not sufficiently scrutinized. Silicon riching well-connected investors.
prices had already begun their precipitous decline before the But subsidy programs with greater discretion, monitor-
loan guarantee was approved, which should have raised some ing, and conditionality seem to do better. In their review of
alarm bells. As Solyndra’s financial difficulties mounted, DoE place-based policies, Neumark and Simpson (2015) find that
officials justified the losses by arguing that this was common the investment and employment response to such policies
in all start-ups. The company had been selected early on as tend to be positive. They write, “The fact that plants that re-
a showcase for the Obama Administration’s efforts and was ceive subsidy offers have their applications pass through an
pushed through the approval process in record time. The ad- initial scrutiny process, and that the targeted outcomes are
ministration invested substantial political capital in the com- often heavily monitored and that payment of the subsidy is
pany’s success, making a potential cut in support difficult to contingent on the job and/or investment targets being met,
contemplate.20 Moreover, Solyndra’s principal private investor may explain why these policies appear more successful in
was an important fundraiser for President Obama and had achieving their stated goals than, for example, enterprise
access to the White House—making this perhaps an example zone programs.” Slattery and Zidar (2020) also provide evi-
of political capture. The company spent huge amounts, for a dence on the employment effects of firm-specific local tax
small start-up, on political lobbying (Rodrik 2014). incentives. Criscuolo et al. (2019) and Cingano et al. (2022)
Even with this bad apple in the cart, one cannot con- provide valuable perspectives on firm subsidies in different
clude that the DoE’s loan guarantee program was a failure as national contexts, Britain and Italy, respectively. In both
a whole. The program supported scores of companies, many countries subsidies appear to have spurred employment, and
of which weaned themselves off public support and became interestingly the effects were stronger for smaller firms.
commercially successful. Industrial policy critics who point In the end, the success of industrial policy depends on
to Solyndra as their Exhibit A rarely mention that Tesla a variety of contingent factors. Political capture cannot be
was also a beneficiary—receiving a $465 million DoE loan ruled out—though in truth the same is true of practically
guarantee in 2009. The company repaid its loan early and any policy. It is difficult to generalize either on ex-ante theo-
became the world’s premier company for electric autos and retical grounds or as an empirical matter. As in most areas
solar batteries. of government policy, the details matter greatly.
Systematic empirical evaluation of industrial poli-
cies has traditionally been hampered by problems of causal

Conclusion
There has never been a shortage of the practice of indus- emphasized, what is required is not just jobs, but bringing
trial policy in the United States. What is different today is good jobs to those who would otherwise be unemployed or
that the need for industrial policy is widely acknowledged employed in less-productive positions.22 An explicit good-
by both sides of the political spectrum. Industrial policy jobs focus, along with better coordination of these different
has risen at the top of the national agenda self-consciously streams of work, would ensure both that training practices
rather than being conducted surreptitiously. This provides are appropriate and that firms receiving public assistance
an opportunity to reexamine its goals and principles. create the right kind of jobs for those who need them the
When the inadequacy of good middle-class jobs is driv- most.
en by secular trends such as technology and globalization, The kind of programs I have advocated in this piece re-
traditional social policy remedies are no longer adequate. main largely a sideshow in present discussions. The regional
Income supports and social insurance do not address the and good-jobs challenges in the American Rescue Plan lack
underlying malady. And preparing young workers for the ambition in scope and magnitude. The House and Senate
labor market and reskilling older workers for newer occupa- versions of the more expansive industrial policy bills cur-
tions can only go so far when firms are not supplying an ad- rently under consideration are heavily focused on manu-
equate quantity of good jobs for middle-skill workers. What facturing, supply-chain resilience, high-tech industries per-
is required are policies that intervene directly in the produc- ceived to be critical to national security, and competition
tive sphere of the economy with the goal of expanding the with China. The Biden Administration’s industrial policy
supply of such jobs. Good jobs require good firms. blueprint focuses mostly on accelerating the energy transi-
Hence, creating good jobs requires interventions on tion (Council of Economic Advisers [CEA] 2021). While
both the supply and the demand sides of the labor market. the need to stimulate quality employment is mentioned fre-
On the one hand, workers must acquire the requisite skills quently in all these initiatives, the maintained presumption
to become productive on the job. This is the province of is that promoting these other objectives will also result in
traditional workforce development and skills-training pro- the creation of adequate numbers of good jobs.
grams. On the other, firms and entrepreneurs must be pro- I have argued here that this outcome cannot be taken
vided with the right ecosystem to nurture their development for granted. In the absence of programs targeted specifical-
and expansion so they can generate quality employment. ly on the supply of good jobs and on technologies that are
A fundamental weakness of prevailing local economic friendly to workers, labor market problems will continue,
development practices is that programs to retain and attract with significant costs to the social and political fabric of the
firms operate at some distance from—and often in tension nation, even if the primary objectives of this newfound en-
with—workforce training programs.21 As Tim Bartik has thusiasm for industrial policy are attained.

Endnotes
1. In terms of the taxonomy of policies presented in Rodrik and 10. U.S. legislators have shown appetite for ARPA-style agencies.
Stantcheva (2021a), the policies covered in this policy proposal In March 2022, the Advanced Research Projects Agency for
fall in the middle cell of the matrix. Health (ARPA-H) was set up at the National Institutes of Health
2. Sometimes it may be easier to define a good job by what it is to “improve the U.S. government’s ability to speed biomedical
not: a bad job. See Kantor, Weise, and Ashford (2021) for an and health research” (National Institutes of Health, n.d.).
exposé on Amazon’s employment practices that subject workers 11. The rest of this section draws heavily on Rodrik and Stantcheva
to arbitrary dismissal, close monitoring that strips them of (2021b).
autonomy and agency, and being shortchanged on pay and 12. Acemoglu (2021) asks, “How do you distinguish an AI
benefits. automation application from one that leads to new tasks and
3. In The Good Jobs Strategy, Zeynep Ton (2014) advocates a range activities for humans? For government policy to redirect
of employment policies such as higher wages and benefits that research, these guidelines need to be in place before the
she argues could help employers as well as employees. The research is undertaken and technologies are adopted. This
argument, encapsulated in her subtitle, is that smart companies calls for a better measurement framework—a tall order, but
can boost profits by investing in their employers. Though such not a hopeless task. Existing theoretical and empirical work on
opportunities clearly exist, it is not clear we can rely on firms’ the effects of automation and new tasks shows that they have
own bottom-line incentives for high-road employment practices. very distinct effects on the labor share of value added (meaning
For a review of the literature and a skeptical take, see Osterman how much of the value added created by a firm or industry goes
(2018). to labor). Greater automation reduces the labor share, while
4. The government subsidy might need to be financed by raising new tasks increase it. Measuring the sum of the work-related
taxes elsewhere in the economy, which will create its own consequences of new AI technologies via their impact on the
inefficiencies (due to the deadweight loss of taxation). Because labor share is therefore one promising avenue. Based on this
of these inefficiencies, it will not be optimal in general to fully measurement framework, policy can support technologies that
internalize the externality: the desirable Pigovian subsidy will tend to increase the labor share rather than those boosting
fall short of the gap between social and private benefits. profits at the expense of labor.”
5. Austin, Glaeser, and Summers (2018) consider three sources of 13. Often the distinction is easy enough to make. When Elon
economic externalities from nonemployment: fiscal costs on Musk started to build a fully automated car factory for Tesla’s
the state through the tax-transfer system, costs imposed on the Model 3 in 2016, he was clear that his objective was to enable
family, and spillovers that encourage nonemployment by others essentially worker-less production. Complete automation
in the community. They reckon these costs range 0.21–0.36 would allow the factory to operate beyond human speed: “Raw
times the wage of low-income workers. See also Bartik (2019) for materials would go in one end and finished cars would roll
a broad discussion of economic and social costs associated with out the other. In between, robots would do everything, a very
employment losses. high speed—speeds too dangerous to risk around frail human
bodies” (DeBord 2017). Interestingly, Elon Musk’s plans failed
6. See Owens, Rossi-Hansberg, and Sarte (2020) for an application
and he had to improvise a new factory built around human
to revitalizing the city of Detroit.
workers. He conceded (on Twitter) that excessive automation
7. Hamilton (1791) claimed he knew how these practical difficulties was a mistake. By contrast, BMW and Mercedes have built
could be overcome, but did not have the space to discuss them— their automation plans around human work, seeking both
the kind of license every author would love to afford himself: greater reliability and more customization in production. The
“The requisite precautions have been thought of; but to enter McKinsey Global Institute (2018) reports, “[BMW] says that
into the detail would swell this report, already voluminous, to combining people with machines on its automotive assembly
a size too inconvenient. If the principle shall not be deemed lines increases the flexibility to build multiple models in
inadmissible the means of avoiding an abuse of it will not be smaller batches and thus respond to shifting customer
likely to present insurmountable obstacles.” demands more quickly.” In new BMW and Mercedes-Benz
8. This discussion is based and draws on Rodrik and Sabel (2022). factories, lightweight robots (called cobots) do not have to be
9. Modern theories of R&D support do allow for asymmetric physically separated from workers and other humans, and
information, as in Akcigit, Hanley, and Stantcheva (2022). But machines perform complementary tasks. Mercedes’s S-class
the range of uncertainty that is allowed is quite limited and these sedans are built by workers who customize cars on the fly using
frameworks preclude other methods of information revelation hand-held tablets, with the automated work being performed
of the type considered below. by the lightweight robots (Wilson and Dougherty 2018).
14. Azoulay et al. (2018) argue that the ARPA model is particularly

suitable for public challenges with the following three features: (2020): “In 2014, states spent between $5 and $216 per capita
a clear mission, an S-shaped technology curve (building on on incentives for firms in the form of firm-specific subsidies
technologies that exist in nascent form but that would greatly and general tax credits, which mostly target investment, job
benefit from further exploration), and significant frictions creation, and research and development. Collectively, these
in the markets for ideas and technology. This seems to fit the incentives amounted to nearly 40 percent of state corporate
ARPA-W suggested here. tax revenues for the typical state, but in some states, incentive
15. As one long-term practitioner put it at a meeting at Harvard spending exceeded corporate tax revenues. States with higher
University, “We are at the center of everything, but control per capita incentives tend to have higher state corporate tax
nothing.” High-level political support and buy-in from the rates. Recipients of firm-specific incentives are usually large
local community are therefore important to ensure things can establishments in manufacturing, technology, and high-skilled
get done. service industries, and the average discretionary subsidy is
$160 million for 1,500 promised jobs.”
16. DARPA-type regimes were discussed above. Sectoral training
programs, with their emphasis on dual-client (employees and 19. I am grateful to Gordon Hanson for the ideas in this paragraph.
employers) wraparound services, have also been extensively 20. The provisional loan commitment to Solyndra in March 2009
discussed in the labor literature. was marked by joint appearances by Secretary Steven Chu
17. The Build Back Better Regional Challenge and Good Jobs of the US Department of Energy, and Vice President Joseph
Challenge programs come closest to the spirit of the local Biden. And, as noted, President Obama himself showed up
industrial policies advocated here. The Regional Challenge is at the company’s headquarters in California in May 2010 to
for “transformational investments to develop and strengthen publicly celebrate Solyndra’s apparent success.
regional industry clusters across the country, all while 21. For example, local economic development agencies may
embracing equitable economic growth, creating good-paying prioritize creating jobs in high-tech or highly skill-intensive
jobs, and enhancing U.S. global competitiveness.” The Good firms with little employment-generating capacity over creating
Jobs Challenge promotes “collaborative skills training systems” jobs for graduates of workforce training programs.
and “aims to get Americans back to work by building and 22. Connecting people in disadvantaged communities with good
strengthening systems and partnerships that bring together jobs requires, beyond training, complementary investments
employers who have hiring needs with other key entities to in transportation, child care, and services that increase job
train workers with in-demand skills that lead to good-paying retention. See Bartik (2022) for a broad discussion. Particularly
jobs. These two programs have been allocated $1 billion and interesting is the model of employer resource networks; that
$500 million, respectively, under the American Rescue Plan. model provides success coaches to small- and medium-size
These programs are described at US Economic Development enterprises to help employees keep their jobs, thus effectively
Administration (2021). linking the workers with the relevant services (Bartik 2022).
18. Tax incentives are the predominant form of industrial and
regional policy in the U.S. In the words of Slattery and Zidar

References
Acemoglu, Daron. 2019. “The Wrong Kind of AI? Artificial Intel- Azoulay, Pierre, Erica Fuchs, Anna Goldstein, and Michael Ke-
ligence and the Future of Labor Demand.” Cambridge arney. 2018. “Funding Breakthrough Research: Promises
Journal of Regions, Economy and Society, 2019. and Challenges of the ‘ARPA Model.’” Working Paper
———. 2021. “AI’s Future Doesn’t Have to Be Dystopian.” Boston 24674, National Bureau of Economic Research, Cam-
Review, May 20, 2021. bridge, MA.
Acemoglu, Daron, Andrea Manera, and Pascual Restrepo. 2020. Ballard-Rosa, Cameron, Amalie Jensen, and Kenneth F. Scheve.
“Does the U.S. Tax Code Favor Automation?” BPEA 2018. “Economic Decline, Social Identity, and Authori-
Conference Drafts. Brookings Papers on Economic Activity tarian Values in the United States.” Harvard University,
(Spring), Brookings Institution, Washington, DC. Cambridge, MA.
Acemoglu, Daron, and Pascual Restrepo. 2018. “Artificial Intel- Bartik, Tim. 2019. “Should Place-Based Jobs Policies Be Used to
ligence, Automation and Work.” Working Paper 24196, Help Distressed Communities?” W.E. Upjohn Institute
National Bureau of Economic Research, Cambridge, MA. for Employment Research, Kalamazoo, MI.
———. 2019. “Automation and New Tasks: How Technology ———. 2020. “Bringing Jobs to People: Improving Local Econom-
Displaces and Reinstates Labor.” Journal of Economic ic Development Policies.” Aspen Institute, Washington,
Perspectives 33 (2): 3–30. DC.
Akcigit, Ufuk, Douglas Hanley, and Stefanie Stantcheva. 2022. ———. 2022. “How State Governments Can Target Job Oppor-
“Optimal Taxation and R&D Policies.” Econometric Soci- tunities to Distressed Places.” Draft, March 2022. W.E.
ety 90 (2): 645–84. Upjohn Institute for Employment Research, Kalamazoo,
Andrews, Michael J., Aaron Chatterji, Josh Lerner, and Scott MI.
Stern, eds. 2022. The Role of Innovation and Entrepre- Biasi, Barbara, David Deming, and Petra Moser. 2022. “Education
neurship in Economic Growth. Chicago, IL: University of and Innovation.” In Andrews et al., The Role of Innovation
Chicago Press. and Entrepreneurship in Economic Growth, 537–558.
Atkinson, Anthony B. 2015. Inequality: What Can Be Done? Cam- Bishop, Christine E. 2014. “High-Performance Workplace Prac-
bridge, MA: Harvard University Press. tices in Nursing Homes: An Economic Perspective.” The
Austin, Benjamin, Edward Glaeser, and Lawrence Summers. 2018. Gerontologist 54 I (S1): S46–S52.
“Jobs for the Heartland: Place-Based Policies in 21st-Cen- Bureau of Economic Analysis (BEA). 2022. “Gross Domestic
tury America.” Brookings Papers on Economic Activity, Product, Corporate Profits, and GDP by Industry, Fourth
(Spring), Brookings Institution, Washington, DC. Quarter and Year 2021.” Bureau of Economic Analysis,
Autor, David H., and David Dorn. 2013. “The Growth of Low-Skill Suitland, MD.
Service Jobs and the Polarization of the US Labor Mar- Bureau of Labor Statistics (BLS). 2022. “Current Population Sur-
ket.” American Economic Review 103 (5): 1553–97. vey.” US Department of Labor, Bureau of Labor Statistics,
Autor, David, David Dorn, and Gordon Hanson. 2019. “When Washington, DC.
Work Disappears: Manufacturing Decline and the Falling Case, Anne, and Angus Deaton. 2020. Deaths of Despair and the
Marriage-Market Value of Men.” American Economic Future of Capitalism, Princeton, NJ: Princeton University
Review: Insights 1 (2): 161–78. Press.
———. 2021. “On the Persistence of the China Shock.” Working Census Bureau. 1980-2019. “American Community Survey.” US
Paper 29401, National Bureau of Economic Research, Census Bureau, Suitland, MD.
Cambridge, MA. Cerrato, Andrea, Federico Maria Ferrara, and Francesco Ruggieri.
Autor, David, David Dorn, Gordon Hanson, and Kaveh Majlesi. 2018. “Why Does Import Competition Favor Repub-
2017. “Importing Political Polarization? The Electoral licans? Localized Trade Shocks, Voting Behavior, and
Consequences of Rising Trade Exposure.” Working Paper Scapegoating in the U.S.” Available at SSRN: https://doi.
22637, National Bureau of Economic Research, Cam- org/10.2139/ssrn.3147169
bridge, MA. Cingano, Federico, Filippo Palomba, Paolo Pinotti, and Enrico
Autor, David, David Mindell, and Elisabeth Reynolds. 2020. The Rettore. 2022. “Making Subsidies Work: Rules vs. Discre-
Work of the Future: Building Better Jobs in an Age of Intel- tion.” Working Paper 1364, Bank of Italy. Available at
ligent Machines. Cambridge, MA: The MIT Press. SSRN: https://ssrn.com/abstract=4090693

Colantone, Italo, and Piero Stanig. 2016. “Global Competition and Friedman, Benjamin M. 2005. The Moral Consequences of Eco-
Brexit.” Unpublished paper, Bocconi University, Milan, nomic Growth. New York, NY: Alfred A. Knopf.
Italy. Fuchs, Erica R. H., Christophe Combemale, Kate S. Whitefoot,
———. 2017. “The Trade Origins of Economic Nationalism: and Britta Glennon. 2022. “The ‘Weighty’ Manufactur-
Import Competition and Voting Behavior in Western ing Sector: Transforming Raw Materials into Physical
Europe.” BAFFI CAREFIN Centre Research Paper Series Goods.” In Andrews et al., The Role of Innovation and
49 (January). Available at SSRN: https://papers.ssrn.com/ Entrepreneurship in Economic Growth, 31–94.
sol3/papers.cfm?abstract_id=2904105 Gallup. 2020. “Lumina Foundation, Gates Foundation and
———. 2018. “The Economic Determinants of the ‘Cultural Omidyar Network Tell Full Story of U.S. Job Quality.”
Backlash’: Globalization and Attitudes in Western Eu- Gallup, Washington, DC.
rope.” BAFFI CAREFIN Centre Research Paper 2018-91. Ghezzi, Piero. 2017. “Mesas Ejecutivas in Peru: Lessons for
Available at SSRN: https://papers.ssrn.com/sol3/papers. Productive Development Policies.” Global Policy 8 (3):
cfm?abstract_id=3267139 369–80.
Council of Economic Advisers (CEA). 2021. “Innovation, Invest- Good Jobs Institute. n.d. “Good Jobs Scorecard.” Good Jobs Insti-
ment, and Inclusion: Accelerating the Energy Transition tute, Cambridge, MA.
and Creating Good Jobs.” White Paper, CEA, White Grogingen Growth and Development Centre (GGDC).
House, Washington, DC. 2014. “10-Sector Database.” GGDC, Grogingen, the
Criscuolo, Chiara, Ralf Martin, Henry G. Overman, and John Netherlands.
Van Reenen. 2019. “Some Causal Effects of an Industrial ———. 2021. “Economic Transition Database.” GGDC, Grogin-
Policy.” American Economic Review 109 (1): 48–85. gen, the Netherlands.
Dal Bò, Ernesto, Frederico Finan, Olle Folke, Torsten Persson, Guiso, Luigi, Helios Herrera, Massimo Morelli, and Tommaso
and Johanna Rickn. 2018. “Economic Losers and Political Sonno. 2017. “Demand and Supply of Populism.” Dis-
Winners: Sweden’s Radical Right.” Institute for Interna- cussion Paper 11871, Center for Economic and Policy
tional Economic Studies, Stockholm, Sweden. Research, Washington, DC.
de Vries, Gaaitzen, Linda Arfelt, Dorothea Drees, Mareike Gode- Hamilton, Alexander. 1791. “Alexander Hamilton’s Final Version
mann, Calumn Hamilton, Bente Jessen-Thiesen, Ahmet of the Report on the Subject of Manufactures.” Presented
Ihsan Kaya, Hagen Kruse, Emmanuel Mensah and Pieter to the House of Representatives, Philadelphia, December
Woltjer. 2021. “The Economic Transformation Data- 5, 1791. Available online at https://founders.archives.gov/
base (ETD): Content, Sources, and Methods.” WIDER documents/Hamilton/01-10-02-0001-0007
Technical Note 2/2021.’ DOI: 10.35188/UNU-WIDER/ Hausmann, Ricardo, Dani Rodrik, and Charles Sabel. 2008.
WTN/2021-2 “Reconfiguring Industrial Policy: A Framework with an
DeBord, Matthew. 2017. “Tesla’s Future Is Completely Inhuman— Application to South Africa.” Working Paper RWP08-31,
and We Shouldn’t Be Surprised.” Insider, May 20, 2017. Harvard Kennedy School, Cambridge, MA.
Dorn, David. 2009. “Essays on Inequality, Spatial Interaction, and Jewish Vocational Service (JVS). n.d. “The History of JVS’ Job
the Demand for Skills.” Dissertation University of St. Gal- Quality Index.” JVS, Boston, MA.
len no. 3613, September 2009. Kantor, Jodi, Karen Weise, and Grace Ashford. 2021. “Power and
Eurofound. 2017. Income Inequalities and Employment Patterns in Peril: 5 Takeaways on Amazon’s Employment Machine.”
Europe Before and After the Great Recession. Luxembourg New York Times, June 15, 2021.
City, Luxembourg: Publications Office of the European Korinek, Anton. 2019. “Labor in the Age of Automation and Arti-
Union. ficial Intelligence.” Policy Brief, Economics for Inclusive
Evans, Peter B. 1995. Embedded Autonomy: States and Industrial Prosperity, Harvard Kennedy School, Cambridge, MA.
Transformation. Princeton, NJ: Princeton University Lawrence, Robert Z. 2022. Behind the Curve: Can Manufacturing
Press, 1995. Still Provide Inclusive Growth? Washington, DC: Peterson
Fallows, James, and Deborah Fallows. 2018. Our Towns: A Institute for International Economics.
100,000-Mile Journey into the Heart of America. New Lerner, Josh, and Ramana Nanda. 2020. “Venture Capital’s Role
York, NY: Pantheon. in Financing Innovation: What We Know and How Much
Families & Workers Fund (FWF). 2022. “Families and Workers We Still Need to Learn.” Journal of Economic Perspectives
Fund Launches Collaboration with the U.S. Department 34 (3): 237–61.
of Labor to Measure and Support Good Jobs.” FWF, Mazzucato, Mariana. 2021. Mission Economy: A Moonshot Guide
Washington, DC. to Changing Capitalism. New York, NY: Harper Business.
Fernández-Arias, Eduardo, Charles Sabel, Ernesto H. Stein, and McKinsey Global Institute (MGI). 2018. “Skill Shift: Automation
Alberto Trejos. 2016. Two to Tango: Public-Private Col- and the Future of the Workforce.” MGI, Washington, DC.
laboration for Productive Development Policies. Washing- National Institutes of Health. n.d. “Advanced Research Projects
ton, DC: Inter-American Development Bank. Agency for Health (ARPA-H).” Accessed August 4th,
Frieden, Jeffry. 2006. Global Capitalism: Its Fall and Rise in the 2022, at https://www.nih.gov/arpa-h.
Twentieth Century. New York, NY: W. W. Norton.

Neumark, David, and Helen Simpson. 2015. “Place-Based Poli- ———. 2021a. “A Policy Matrix for Inclusive Prosperity.” Work-
cies.” In Handbook of Regional and Urban Economics, vol. ing Paper 28736, National Bureau of Economic Research,
5, edited by Gilles Duranton, J. Vernon Henderson and Cambridge, MA.
William C. Strange, pp. 1197–287. Amsterdam, Nether- ———. 2021b. “Fixing Capitalism’s Good Jobs Problem.” Oxford
lands: Elsevier. Review of Economic Policy 37 (4): 824–37.
Organisation for Economic Co-operation and Development Sabel, Charles. 2007. “Bootstrapping Development: Rethinking
(OECD). 2019. The Future of Work: OECD Employment the Role of Public Intervention in Promoting Growth.”
Outlook 2019. Paris: OECD Publishing. In On Capitalism, edited by Victor Lee and Richard Swed-
———. n.d. “Employment Database: Job Quality.” Accessed May berg, 305–41. Stanford, CA: Stanford University Press.
3rd, 2022, at https://www.oecd.org/els/emp/employment- Sitaraman, Ganesh. 2019. “The Collapse of Neoliberalism.” The
database-jobquality.htm New Republic, December 23, 2019.
Osterman, Paul. 2018. “In Search of the High Road: Meaning and Slattery, Cailin, and Owen Zidar. 2020. “Evaluating State and Lo-
Evidence.” ILR Review 71 (1): 3–34. cal Business Incentives.” Journal of Economic Perspectives
———. 2019. “Improving Job Quality in Long Term Care.” In 34 (2): 90–118.
Creating Good Jobs: An Industry-Based Strategy, edited Stevenson, Matthew. 2018. “Demand for Healthcare Workers Will
by Paul Osterman, 115–144. Cambridge, MA: The MIT Outpace Supply by 2025: An Analysis of the US Health-
Press. care Labor Market.” Mercer, New York, NY.
Owens, Raymond III, Esteban Rossi-Hansberg, and Pierre-Daniel Strain, Michael R., and Stan A. Veuger. 2019. “Economic Shocks
Sarte. 2020. “Rethinking Detroit.” American Economic and Clinging.” Working Paper, American Enterprise
Journal: Economic Policy 12 (2): 258–305. Institute, Washington, DC.
Poo, Ai-jen, and Ilana Berger. 2022. “We Need More Home Care Timmer, Marcel P., Gaaitzen de Vries, and Klaas de Vries 2015.
Workers. But First We Have to Pay Them Enough.” New “Patterns of Structural Change in Developing Countries.”
York Times, March 30, 2022. In Routledge Handbook of Industry and Development,
Rodrik, Dani. 2007. “Industrial Policies for the Twenty-First Cen- edited by John Weiss and Michael Tribe, 65–83. Oxford-
tury.” In One Economics, Many Recipes: Globalization, In- shire, UK: Routledge.
stitutions, and Economic Growth, edited by Dani Rodrik, Ton, Zeynep. 2014. The Good Jobs Strategy: How the Smartest
99–152. Princeton, NJ: Princeton University Press, 2007. Companies Invest in Employees to Lower Costs & Boost
———. 2008. “Normalizing Industrial Policy.” Working Paper 3, Profits. Boston, MA: New Harvest.
Commission on Growth and Development, Washington, US Economic Development Administration (EDA). 2021. “EDA
DC. American Rescue Plan Programs.” EDA, Washington,
———. 2012. “Why We Learn Nothing from Regressing Eco- DC.
nomic Growth on Policies.” Seoul Journal of Economics 25 Victor, David G., and Charles F. Sabel. 2022. Fixing the Climate:
(2): 137–51. Strategies for an Uncertain World. Princeton, NJ: Princ-
———. 2014. “Green Industrial Policy.” Oxford Review of Econom- eton University Press.
ic Policy 30 (3): 469–91. Weitzman, Martin. 1974. “Prices vs. Quantities.” Review of Eco-
———. 2017. “Rescuing Economics from Neoliberalism.” Boston nomic Studies 41 (4): 477–91.
Review, November 6, 2017. Wessel, David. 2021. Only the Rich Can Play: How Washington
———. 2021. “Why Does Globalization Fuel Populism? Econom- Works in the New Gilded Age. New York, NY: Public
ics, Culture, and the Rise of Right-Wing Populism.” An- Affairs.
nual Review of Economics 13: 133–70. Wilson, William Julius. 1996. When Work Disappears: The World
Rodrik, Dani, and Charles F. Sabel. 2022. “Building a Good Jobs of the New Urban Poor. New York, NY: Alfred A. Knopf.
Economy.” In A Political Economy of Justice, edited by Wilson, H. James, and Paul R. Daugherty. 2018. “Collaborative In-
Danielle Allen, Yochai Benkler, Leah Downey, Rebecca telligence: Humans and AI Are Joining Forces.” Harvard
Henderson, and Josh Simons, chap. 2. Chicago, IL: Uni- Business Review (July–August).
versity of Chicago Press. World Inequality Lab. 2021. “World Inequality Database.” Ac-
Rodrik, Dani, and Stefanie Stantcheva. 2020. “Economic Inequal- cessed May 10th, 2022, at https://wid.world/data/
ity and Insecurity: Policies for an Inclusive Economy.”
Report for President Macron’s Blanchard-Tirole Commis-
sion on Major Future Economic Challenges, Republic of
France.

ADVISORY COUNCIL
STEPHANIE AARONSON ROBERT GREENSTEIN NANCY L. ROSE

Vice President and Director, Visiting Fellow, The Hamilton Project, Charles P. Kindleberger Professor of
Economic Studies; Economic Studies, The Brookings Institution; Applied Economics,
Senior Fellow, Economic Studies, Founder and President, Massachusetts Institute of Technology
The Brookings Institution Center on Budget and Policy Priorities
DAVID M. RUBENSTEIN
GEORGE A. AKERLOF MICHAEL GREENSTONE Co-Founder and Co-Chairman,
University Professor, Milton Friedman Professor in The Carlyle Group
Georgetown University Economics and the College,
Director of the Becker Friedman Institute for ROBERT E. RUBIN
ROGER C. ALTMAN Research in Economics, and Former U.S. Treasury Secretary;
Founder and Senior Chairman, Director of the Energy Policy Institute, Co-Chair Emeritus,
Evercore University of Chicago Council on Foreign Relations
KAREN L. ANDERSON GLENN H. HUTCHINS LESLIE B. SAMUELS

Senior Director of Policy & Communications, Chairman, North Island and Senior Counsel,
Becker Friedman Institute for North Island Ventures Cleary Gottlieb Steen and Hamilton LLP
Research in Economics,
The University of Chicago LAWRENCE F. KATZ SHERYL SANDBERG
Elisabeth Allison Professor of Economics, Chief Operating Officer,
ALAN S. BLINDER Harvard University Facebook
Gordon S. Rentschler Memorial Professor of
Economics and Public Affairs, MELISSA S. KEARNEY DIANE WHITMORE SCHANZENBACH
Princeton University; Neil Moskowitz Professor of Economics, Margaret Walker Alexander Professor and
Nonresident Senior Fellow, University of Maryland; Director, The Institute for Policy Research,
The Brookings Institution Director, Aspen Economic Strategy Group; Northwestern University;
Nonresident Senior Fellow, Nonresident Senior Fellow,
STEVEN A. DENNING The Brookings Institution The Brookings Institution
Chairman,
General Atlantic LILI LYNTON STEPHEN SCHERR
Founding Partner, Chief Executive Officer,
JOHN M. DEUTCH Boulud Restaurant Group Goldman Sachs Bank USA
Institute Professor,
Massachusetts Institute of Technology HOWARD S. MARKS RALPH L. SCHLOSSTEIN
Co-Chairman, Chairman Emeritus,
CHRISTOPHER EDLEY, JR. Oaktree Capital Management, L.P. Evercore
Co-Founder and President Emeritus,
The Opportunity Institute KRISTON MCINTOSH ERIC SCHMIDT
Managing Director, Co-Founder, Schmidt Futures;
BLAIR W. EFFRON Hamilton Place Strategies Former CEO and Chairman, Google
Partner,
Centerview Partners LLC ERIC MINDICH ERIC SCHWARTZ
Founder, Chairman and CEO,
DOUGLAS W. ELMENDORF Everblue Management 76 West Holdings
Dean and Don K. Price Professor of
Public Policy, DAMBISA MOYO THOMAS F. STEYER
Harvard Kennedy School Co-Principal, Business Leader and Philanthropist;
Versaca Investments Founder,
JUDY FEDER NextGen America
Professor and Former Dean, SUZANNE NORA JOHNSON
McCourt School of Public Policy, Former Vice Chairman, MICHAEL R. STRAIN
Georgetown University Goldman Sachs Group, Inc.; Director of Economic Policy Studies and
Co-Chair, Arthur F. Burns Scholar in Political Economy,
JASON FURMAN The Brookings Institution American Enterprise Institute
Aetna Professor of the Practice of
Economic Policy, Harvard University; PETER ORSZAG LAWRENCE H. SUMMERS
Senior Fellow, Peterson Institute for CEO, Financial Advisory, Charles W. Eliot University Professor,
International Economics; Lazard Freres & Co LLC Harvard University
Senior Counselor, The Hamilton Project
RICHARD PERRY ALLEN THORPE
MARK T. GALLOGLY Managing Partner & Chief Executive Officer, Partner,
Cofounder, Perry Capital Hellman & Friedman
Three Cairns Group.
PENNY PRITZKER LAURA D’ANDREA TYSON
TED GAYER Chairman and Founder, PSP Partners; Distinguished Professor of the Graduate School,
President, 38th Secretary of Commerce University of California, Berkeley
Niskanen Center
MEEGHAN PRUNTY DAVID WEIL
TIMOTHY F. GEITHNER Principal, Co-President,
President, Warburg Pincus; PE Strategic Partners Charles and Lynn Schusterman Family
Senior Counselor, The Hamilton Project Philanthropies
ROBERT D. REISCHAUER
Distinguished Institute Fellow and
WENDY EDELBERG
President Emeritus,
Director
Urban Institute
Industrial policies have been with us for a long time, but often they have been carried out surreptitiously
and without clear motivation. The recent revival of discussions around industrial policy provides a welcome
opportunity for self-consciously crafting an improved set of policies. A modern approach to industrial policy
must respond to new circumstances. It must target “good-jobs externalities,” in addition to the traditional
learning, technological, and national security considerations. Relatedly, industrial policy’s traditional focus
on manufacturing and globally competitive industries has to be broadened to service sectors and smaller
and medium-sized firms. And the practice of industrial policy will need to rely less on traditional top-down
policy instruments—such as subsidies and tax incentives for firms—and more on collaborative, iterative
interaction whereby public agencies supply a portfolio of customized public services in exchange for firms
undertaking soft commitments on the quantity and quality of employment. With these objectives in mind,
this paper develops two types of specific initiatives: one at the local level and the other at the federal level.
The local approach builds on existing development and business assistance programs that take the form of
collaborative partnerships between local development agencies, firms, and other stakeholders aiming to
revitalize local communities and create good jobs. The federal initiative is an Advanced Research Projects
Agency (ARPA) focused on the promotion of employment-friendly technologies: ARPA-W(orkers).
Private Employment in Goods and Services Sectors, 1939–2022

120
Service-providing sector
100
80
Millions of workers
+59.0m
60
40
20 −4.3m
Goods-producing sector
0
1939 1944 1950 1955 1961 1966 1972 1977 1983 1988 1994 1999 2005 2010 2016 2022
Source: Bureau of Labor Statistics 2022.

Note: Gray bars indicate recession periods.
1775 Massachusetts Ave., NW

Washington, DC 20036
(202) 797-6484
Printed on recycled paper. W W W. H A M I LTO N P R O J E C T.O R G

Rodrik - An Industrial Policy For Good Jobs

Uploaded by

Copyright:

Available Formats

Rodrik - An Industrial Policy For Good Jobs

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rodrik - An Industrial Policy For Good Jobs

Uploaded by

Copyright:

Available Formats

POLICY PROPOSAL | SEPTEMBER 2022

An Industrial Policy for Good Jobs

THE HAMILTON PROJECT MISSION STATEMENT

ii The Hamilton Project • Brookings

The Hamilton Project • Brookings iii

An Industrial Policy for Good Jobs 1

2 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 3

Employment Growth by Occupation and Annual Pay, 1980 to 2019

4 The Hamilton Project • Brookings

Middle Class Pretax Income Shares in Select Countries, 1980–2021

Source: World Inequality Lab 2021.

An Industrial Policy for Good Jobs 5

6 The Hamilton Project • Brookings

U.S. Manufacturing Shares of Total Employment and GDP, 1939–2020

An Industrial Policy for Good Jobs 7

Private Employment in Goods and Services Sectors, 1939–2022

Source: Bureau of Labor Statistics 2022.

output as a whole). But there does not seem to be a single

8 The Hamilton Project • Brookings

Manufacturing Trends Across Countries, 1990–2018

Singapore South Korea Taiwan Turkey

An Industrial Policy for Good Jobs 9

Good Jobs in Long-Term Care?

10 The Hamilton Project • Brookings

Traditional Industrial Policy and Proposed Approach

An Industrial Policy for Good Jobs 11

12 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 13

14 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 15

16 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 17

18 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 19

20 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 21

22 The Hamilton Project • Brookings

An Industrial Policy for Good Jobs 23

24 The Hamilton Project • Brookings

STEPHANIE AARONSON ROBERT GREENSTEIN NANCY L. ROSE

KAREN L. ANDERSON GLENN H. HUTCHINS LESLIE B. SAMUELS

Private Employment in Goods and Services Sectors, 1939–2022

Source: Bureau of Labor Statistics 2022.

1775 Massachusetts Ave., NW

Printed on recycled paper. W W W. H A M I LTO N P R O J E C T.O R G

You might also like