The World Bank Economic Review, 38(1), 2024, 161–184 https://doi.org10.1093/wber/lhad019 Article Labor Market and Macroeconomic Dynamics in Latin Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 America amid COVID: The Role of Digital-Adoption Policies Alan Finkelstein Shapiro, Victoria Nuguer, and Santiago Novoa Gomez Abstract This paper analyzes how a policy that lowers firm digital-adoption costs shapes the labor-market and economic recovery from COVID-19 in Latin America (LA) using a framework with firm entry and unemployment, where salaried firms can adopt digital technologies and the employment and firm structure embodies key features of LA economies. Using Mexico as a case study, the model replicates the response of the labor market and output at the onset of the COVID recession and in its aftermath, including the dynamics of labor-force participation and informal employment. A policy-induced permanent reduction in the cost of adopting digital technologies at the trough of the recession bolsters the recovery of GDP, total employment, and labor income, and leads to a larger expansion in the share of formal employment compared to a no-policy scenario. In the long run, the economy exhibits a reduction in total employment but higher levels of GDP and labor income, greater average firm productivity, a larger formal employment share, and a marginally lower unemployment rate. Finally, as a side effect, the policy exacerbates the differential between formal and informal labor income, both as the economy recovers from the COVID recession and in the long run. JEL classification: E24, J23, J24, J64, O14 Keywords: COVID, business cycles, self-employment and informality, unemployment and labor force partici- pation, information and communications technologies (ICT) Alan Finkelstein Shapiro (corresponding author) is an associate professor in the Department of Economics at Tufts Univer- sity, Medford, United States; his email address is Alan.Finkelstein_Shapiro@tufts.edu. Victoria Nuguer is a senior research economist in the Department of Research and Chief Economist (RES) at the Inter-American Development Bank, Washing- ton, DC, United States; her email address is victorian@iadb.org. Santiago Novoa is a research fellow in the Department of Research and Chief Economist (RES) at the Inter-American Development Bank, Washington, DC, United States; his email address is snovoagomez@iadb.org. We thank the editor Norman Loayza, three anonymous reviewers, Andy Powell, Eduardo Cavallo, Arturo Galindo, Laura Ripani, Oliver Azuara Herrera, and David Kaplan for comments and feedback. This paper was originally written to support the analysis of the state of labor markets in Latin America and the Caribbean and the role of labor-market policies amid COVID-19 in Chapter 6 of the IADB Macro Report 2022. The section “Digital Adoption and the Labor Market Recovery in LAC amid COVID-19: New Firm Entry and Its Implications for Job Creation” of the report presents a brief summary of the paper’s main findings. This project received funding support from the Inter-American Development Bank. The views in this paper are solely the responsibility of the authors and should not be interpreted as representing the views of the Inter-American Development Bank, its Executive Board, or its Management. Any errors are our own. A supplementary online appendix is available with this article at The World Bank Economic Review website. C The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 162 Finkelstein Shapiro, Nuguer, and Novoa 1. Introduction In 2020, the COVID pandemic triggered an unprecedented recession around the world, with the Latin America and the Caribbean (LAC) region experiencing the most dramatic contraction in employment and economic activity. Since the onset of COVID, several studies have documented how the 2020 recession had a unique impact on Latin American labor markets compared to past recessions: the COVID shock Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 not only led to a dramatic reduction in labor-force participation but also in informal employment, which stands in stark contrast to the typical and well-known role of informality as a countercyclical employment and income buffer in LAC labor markets (ECLAC 2021; Leyva and Urrutia 2023). In addition, survey- based evidence suggests that firms in the region experienced a record collapse in sales, which not only led to firms exiting but also put a sharp dent in new firm creation. Critically, micro and small firms, which account for 50 percent of total employment and tend to be in high-contact sectors, were the most affected (ECLAC 2020a). Prior to COVID, digital-adoption rates by firms and households in the region had been expanding at a steady pace, with the latest available data showing that around 70 percent of the population in the region uses the internet. Moreover, alongside greater internet use, the share of the population making and receiving digital payments has been steadily growing. Recent evidence suggests that digital-adoption rates by firms increased sharply at the onset of COVID, both in LAC and in other developing and emerging economies. For example, using the World Bank Business Pulse Survey covering 60 economies, Cirera, Comin, and Cruz (2022) document that almost 50 percent of firms in the survey started using or increased the usage of digital platforms, while almost 30 percent devoted resources to digital solutions. Moreover, even though larger firms were better able to invest and use digital platforms during COVID, micro and small firms also saw a sharp increase in the probability of using digital solutions. Along similar lines, using a pulse survey covering more than 50 countries and 100,000 firms, Apedo-Amah et al. (2020) document that almost 35 percent of firms surveyed increased the use of digital platforms, while almost 20 percent spent resources on digital solutions. Turning to LAC, over the first half of 2020, new business websites registered and online business sales using digital platforms like PayU grew by more than 200 percent in Brazil, Chile, Colombia, and Mexico (ECLAC 2020b; Díaz de Astarloa et al. 2021). Importantly, several governments in the region actively adopted and promoted a host of measures and policies aimed specifically at supporting and facilitating greater firm digital adoption via greater accessibility and affordability (ECLAC 2020b; Díaz de Astarloa et al. 2021).1 A key objective of these policies is to leverage the use of digital technologies to reduce barriers to firm entry, the rationale being that lower barriers will support greater firm and employment creation, limit the adverse effects from the contraction in economic activity, and bolster a faster recovery from the COVID downturn by improving firm productivity via technology adoption.2 More broadly, policies that facilitate firm digital adoption have the potential to improve the productivity profile of firms and, in doing so, encourage greater firm and employment formality—a long-standing goal in the LAC region. While recent evidence suggests that digital technologies can improve firm-level outcomes, whether bolstering greater firm digital adoption has broader labor-market and macroeconomic consequences—both in the context of economic recoveries and in the long term—remains a key policy question. 1 In particular, in 16 LAC economies, there were on average 23 government measures aimed at bolstering access and affordability (ECLAC 2020b). Examples of these policies include free or cheaper access to online marketplaces; technical assistance aimed at improving firms’ digital presence, payments, and sales; and partnerships between banks and businesses supported by the government to improve access and expand online interactions and transactions. 2 As shown in Finkelstein Shapiro and Mandelman (2021), there is a negative relationship between barriers to firm entry and firm digital adoption. Examples of how digital adoption is associated with lower barriers to entry include the expansion of potential markets via online sales and promotion (see Díaz de Astarloa et al. (2021) for evidence from Mercado Libre), the reduction in effective registration and administrative costs associated with firm creation by using online filing systems, and the reduction in the effective costs of finding input suppliers and customers via digital platforms and markets. The World Bank Economic Review 163 This paper adapts the framework in Finkelstein Shapiro and Mandelman (2021) to analyze the labor- market and macroeconomic implications of digital-adoption policies in Latin America (LA) in the context of the COVID recession and recovery. The model features firm entry and exit, involuntary unemployment, and labor-force participation, and explicitly captures key defining characteristics of the firm and employ- ment structure of LA economies. Importantly, the model includes a margin whereby salaried firms can choose between adopting a regular technology that relies solely on salaried labor, or a technology that Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 combines salaried labor with information and communication technologies (ICT), thereby giving rise to endogenous digital adoption by firms. On the labor-market front, the model features self-employment and two categories of salaried workers—high-wage (or formal) workers and low-wage (or informal) workers. Salaried firms that use the regular technology only hire low-wage workers. In contrast, salaried firms that adopt ICT hire both high-wage and low-wage workers, where high-wage workers complement ICT and low-wage workers are imperfectly substitutable with the ICT-high-wage worker complement. Focusing on Mexico as a case study, the model incorporates a set of shocks to quantitatively replicate the increase in unemployment and the contraction in employment, labor-force participation, and output at the onset of the COVID-19 recession, as well as the subsequent (and ongoing) recovery. The model is then used to analyze how a policy-induced permanent reduction in the barriers to adopt digital technologies at the trough of the recession shapes labor-market and aggregate dynamics along the recovery path relative to a no-policy scenario. Fostering greater digital adoption in the aftermath of the recession bolsters the recovery of GDP, total employment, and labor income, and leads to a larger expansion in the share of for- mal employment compared to a scenario without policy. In the long term, the policy induces a reduction in total employment and labor-force participation that nonetheless results in higher levels of GDP and labor income, a larger formal employment share, and a marginally lower unemployment rate. This out- come stems from the change in the technological composition of firms and the associated improvement in average firm productivity induced by the policy. At the same time, the policy exacerbates the differential between formal and informal labor income, both as the economy recovers from the COVID recession and in the long run. This last finding points to a trade-off between improved macroeconomic outcomes and labor income inequality between formal and informal workers, and has broader implications in an environment where a significant share of households may not directly benefit from the policy’s positive impact on formal employment and labor income. This paper contributes to the literature on the effects of the COVID-19 pandemic in developing and emerging economies. Alon et al. (2020) analyze the effects of lockdowns, policies based on the age of the population, and school closures in a macro model with epidemics and incomplete markets that allows for differences in the economy’s fiscal and healthcare capacity, the population’s age structure, and the degree of informality. Their analysis highlights the relative effectiveness of age-based policies and school closures relative to lockdowns in a developing country context. Using a similar framework, Alon et al. (2021) show that the sharper decline in economic activity in emerging economies relative to lower-income and advanced economies can be explained by differences in public transfers and the share of employment in high-contact occupations. Closer to our focus on LA, Alfaro, Becerra, and Eslava (2021) show that microentrepreneurship, informality, and limited telework job opportunities play an important role in ex- plaining the collapse in employment in LA in 2020 by comparing Colombia to a labor-market structure that mimics the US labor market. Finally, Leyva and Urrutia (2023) provide a comprehensive overview of labor-market and output dynamics in Brazil, Chile, Colombia, Mexico, and Peru amid COVID, highlight- ing the unique behavior of informal employment and labor-force participation during COVID relative to previous recessions in the region. They then use a macro model with self-employment and search frictions in formal employment to show that shocks to informal employment and labor supply are essen- tial to explain the dynamics of the labor market at the onset of COVID and its aftermath, and analyze the role of labor-market policies in bolstering the recovery process. This paper complements these stud- ies by analyzing how policies that expand firm digital adoption shape not only the labor-market and 164 Finkelstein Shapiro, Nuguer, and Novoa Figure 1. Labor-Market and Macroeconomic Dynamics in Mexico, 2020–2021 Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Source: Saint Louis Federal Reserve Economic Database (FRED) and National Institute of Statistics and Geography (INEGI). Note: All series are seasonally adjusted. LFP denotes labor force participation. economic recovery in the aftermath of COVID, but also long-run employment and macro outcomes in an environment that captures the distinct employment and firm structure of LA economies. The rest of the paper is structured as follows. Section 2 provides a brief overview of labor-market and macroeconomic dynamics in Mexico at the onset of the COVID recession and in its aftermath. Sec- tion 3 summarizes the key features of the model. Section 4 presents the main results from the quantitative experiments and discusses the main economic mechanisms. Section 5 concludes. 2. Overview: Labor-Market and Macroeconomic Dynamics in Mexico amid COVID As noted in Section 1, Leyva and Urrutia (2023) document the dynamics of labor markets in five major LA economies since the onset of COVID. For completeness, this section presents similar facts to those in their work with a focus on Mexico, which is chosen as a case study because, out of all the major LA economies, Mexico implemented the fewest and most limited set of policies to counteract the adverse effects of COVID on economic activity, thereby making the analysis of digital-adoption policies in the context of the COVID economic recovery more transparent. Figure 1 plots the dynamics of real GDP, total employment, self-employment, the share of informal employment in total employment, the unemployment rate, and the labor-force participation rate for the The World Bank Economic Review 165 period 2020Q1 through 2021Q3 (the latest quarter of available data for these variables). As the figure sug- gests, real GDP, total employment, and labor-force participation all experienced a dramatic contraction in 2020Q2—which marked the onset of COVID—and 2020Q3, with GDP and total employment falling by almost 18 percent relative to 2020Q1, and with labor-force participation falling by almost 15 percentage points relative to 2020Q1. As noted in Leyva and Urrutia (2023), in stark contrast to the typical counter- cyclicality of informal employment in Mexico and in LA more generally, both self-employment and the Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 share of informal employment (comprised of self-employment and informal salaried employment) expe- rienced contractions, with self-employment falling by 30 percent relative to 2020Q1. At the same time, the unemployment rate rose by more than 1.5 percentage points relative to 2020Q1. While this may seem low compared to the increase in unemployment rates in other countries amid COVID, a 1.5 percentage- point increase is substantial compared to the expansion in unemployment in pre-COVID recessions in Mexico. While real GDP and labor-force participation continue to remain below their pre-COVID levels by 2021Q3, total employment went back to its pre-COVID level in 2021Q2, roughly a year after its initial contraction. Importantly, the recovery in total employment has been driven primarily by the recovery of informal employment. As a result, the recovery from the original collapse in economic activity in 2020Q2 has been characterized by a change in the composition of total employment towards greater informality. In particular, as is evident from fig. 1, the share of informal employment in total employ- ment surpassed its pre-COVID level after the second half of 2020 and has continued to grow. Given the well-known negative link between informality and productivity, this change in the composition of employment towards greater informality has relevant implications for the medium- and long-term pro- ductivity profile of the economy if the share of informal employment remains above its pre-COVID level. With this background in mind, the next section describes a framework that can shed light on how a policy that permanently reduces the barriers to digital adoption by firms affects the recovery process from the COVID recession, and shapes long-term labor-market and macroeconomic outcomes. 3. Model Summary Finkelstein Shapiro and Mandelman (2021) build a macroeconomic framework that features involuntary unemployment via search and matching frictions, endogenous labor-force participation, and salaried- firm entry and exit. In the model, households not only send household members to search for jobs at salaried firms, but can also send their members to work in self-employment. Based on their idiosyn- cratic productivity level upon entry, salaried firms can use a regular technology that uses salaried la- bor subject to search and matching frictions or, after paying a fixed cost, choose to adopt a pro- duction technology that combines information and communication technologies (ICT) with salaried labor. The production process of firms using ICT is such that ICT complements a sub-segment of salaried workers, thereby creating an ICT-salaried labor composite that is imperfectly substitutable with a dif- ferent sub-segment of salaried employment within the firm. The category of salaried employment that complements ICT can be interpreted as high-skilled formal salaried employment, while the category that is substitutable with the ICT-salaried labor composite can be interpreted as low-skilled formal salaried employment. Finally, the category of workers employed by firms that do not use ICT can be interpreted as low-skilled informal employment. As such, the framework features both self-employment and differ- ent categories of salaried employment that are meant to capture formal and informal salaried work in a tractable way. This model is adopted to conduct a quantitative analysis of digital-adoption policies in an LA labor market in the context of COVID. The rest of this section presents a brief summary of the production 166 Finkelstein Shapiro, Nuguer, and Novoa structure, the labor market, and the household structure, pointing out the main modifications and ad- ditions made to the model—mainly, changes to the timing of the labor-market matching process and the inclusion of a set of shocks that allow the model to replicate the response of key labor-market and macroeconomic variables to the pandemic—and refer the reader to the model in Finkelstein Shapiro and Mandelman (2021) for details on the economic environment. Total Output. Total output Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 φy φy −1 φy −1 φy −1 φ φ Yt = Ys,t y + Ye,t y is comprised of total output from salaried firms Ys,t and total output from self-employed individuals Ye,t , where total salaried and self-employment outputs are imperfectly substitutable, with parameter φ y > 1 dictating the degree of substitutability. The relative prices of Ys,t and Ye,t are given by ps,t and pe,t , respectively. Salaried Firms: Production Technologies. Salaried firms operate in a monopolistically competitive en- vironment. Firm entry into the salaried sector is endogenous and subject to a sunk entry cost fe > 0. Upon entry, firms draw their idiosyncratic productivity level a from a common distribution G(a), and firms de- cide which production technology they adopt. Each firm’s realized level of a remains unchanged until the firm exits the market with exogenous probability 0 < δ < 1. Firms with productivity a below an endogenous threshold ai,t rely on a regular (r) production technol- ogy that uses salaried workers, denoted by nr r,t (a ), making them r firms. In turn, firms with productivity a above ai,t rely on a production technology that combines ICT capital ki,t (a) with two categories of salaried workers, one that complements ICT capital, denoted by nii,t (a ), and one that is imperfectly sub- stitutable with the ICT-salaried labor composite, denoted by nir,t (a ). To adopt the ICT-based technology, each firm that decides to do so must incur a fixed cost fi > 0—the cost of digital adoption—making them i firms. Given the complementarity between nii,t (a ) and ICT capital, nii,t (a ) can be interpreted as skilled labor while nir,t (a ) can be interpreted as unskilled labor. The variable nr r,t (a ) can be interpreted as unskilled labor as well. The terms nir,t (a ) and nr r,t ( a ) are perfectly substitutable, the only difference being whether they work at i or r firms. Salaried Firms: Individual-Firm Profits, Optimal Pricing, and Technology Choice. Each salaried firm maximizes profits by choosing inputs and its real output price optimally subject to a standard downward- sloping demand function for its output, where the effective real marginal cost of inputs for a given firm depends on the production technology the firm uses. Denoting the real price of firm j ∈ {i, r} by ρ j,t (a) and its effective real marginal cost by (mcj,t /a), optimal pricing yields a standard expression, where the real price is a markup over marginal cost: ρ j,t (a) = (ε /(ε − 1))(mcj,t /a), where ε > 1 is the elas- ticity of substitution between individual salaried-output categories. Finally, denoting individual profits for firm j ∈ {i, r} by dj,t (a), a salaried firm is indifferent between technologies when di,t (ai,t ) = dr,t (ai,t ), where fi is a component of di,t (a), so that changes in the fixed cost fi affect firms’ choices over digital adoption. Evolution of Salaried Firms, Salaried Firm Categories, and Average Profits. Let Ne,t be the number of new salaried entrants and Nt be the number of salaried firms that are active in period t. Recalling that salaried firms exit the market with probability 0 < δ < 1, the number of active salaried firms is given by Nt = (1 − δ )[Nt − 1 + Ne,t − 1 ]. In turn, given the optimal choice over production technolo- gies, the number of firms using the r and i technologies are defined as Nr,t = G(ai,t )Nt and Ni,t = [1 − G(ai,t )]Nt , respectively. The object G(a ) = [1 − (amin /a )k p ] follows a Pareto distribution, where the shape parameter kp > ε − 1. Finally, average profits for a given salaried firm are given by d ˜t ≡ (Nr,t /Nt )d˜r,t + ˜ ˜ ˜ (Ni,t /Nt )di,t where dr,t and di,t are evaluated at their respective average idiosyncratic productivity levels. The World Bank Economic Review 167 Matching Processes for Salaried Employment. The salaried labor market is subject to search and match- ing frictions. The matching functions for r and i salaried employment follow Den Haan, Ramey, and Watson (2000). Specifically, they are constant-returns-to-scale and given by zmr ,t (sr,t vr,t ) m(sr,t , vr,t ) = 1/ξ , sξ ξ r,t + vr,t Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 (si,t vi,t ) m(si,t , vi,t ) = 1/ξ , sξ i,t + viξ,t where ξ > 0 and zmr ,t is a matching efficiency shock associated with r employment. For each salaried category j ∈ {r, i}, sj,t denotes the measure of salaried searchers and vj,t denotes the aggregate number of vacancies in employment category j. Defining market tightness as θ j,t = vj,t /sj,t , the job-filling and job- finding probabilities for salaried category j are given by q(θ j,t ) = m(sj,t , vj,t )/vj,t and f(θ j,t ) = m(sj,t , vj,t )/sj,t , respectively. ICT Capital Accumulation, Salaried Job Creation, and Salaried Production Processes. ICT capi- tal follows a standard accumulation process whereby the existing stock of ICT capital ki,t depre- ciates at an exogenous rate 0 < δ i < 1 and real resources are spent on ICT capital investment invt in order to bolster the future stock of ICT capital ki,t + 1 . The profit maximization problem of i firms is subject to the evolution of ICT capital, and delivers a standard Euler equation for ICT capital. Turning to salaried job creation, in the presence of search and matching frictions, searching for salaried workers requires spending ψ > 0 per posted vacancy in each category of salaried employment. From an aggregate standpoint, the evolution of salaried employment from the perspective of firms is given by nr,t = (1 − zρs ,t ρs )nr,t −1 + vr,t q(θr,t ), nii,t = (1 − zρs ,t ρs )nii,t −1 + vi,t q(θi,t ), where nir,t = ωt nr,t is the total measure of r workers working in i firms, nr r,t = (1 − ωt )nr,t is the total mea- sure of r workers working in r firms, and vr,t and vi,t denote aggregate vacancies for salaried employment category j ∈ {i, r}. Here, ωt represents the endogenous share of r employment that is employed by i firms, q(θ j,t ) is the job-filling probability in salaried employment category j ∈ {i, r}, 0 < ρ s < 1 is the exogenous job separation probability, and zρs ,t is a salaried job destruction shock. The profit maximization prob- lems of r and i firms are subject to the relevant perceived laws of motion for salaried employment, and deliver standard job creation conditions for each category of salaried employment. The aggregate produc- tion function of r firms is linear in labor nr r,t and can be expressed as H (nr,t ) = zt nr,t while the aggregate r r production function of i firms can be written as 1 λi λi λi αk kλ λk F nii,t , nir,t , ki,t = zt zi (1 − φi ) nir,t + φi k i,t + (1 − αk )(nii,t )λk , where 0 < φ i , α k < 1, λi , λk < 1. The variable zt represents exogenous aggregate productivity and zi > 0 represents the exogenous sectoral productivity of i firms. 168 Finkelstein Shapiro, Nuguer, and Novoa With this information in mind, firms’ choices over job creation, the allocation of r workers across firm categories, and ICT capital can be expressed as ψ r i ψ = (1 − ωt ) mcr,t Hnrr ,t − wr,t + ωt mci,t Fnir ,t − wr,t + Et t +1|t (1 − zρs ,t +1 ρs ) , q(θr,t ) q(θr,t +1 ) ψ ψ = mci,t Fnii ,t − wii,t + Et t +1|t (1 − zρs ,t +1 ρs ) , Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 q(θi,t ) q(θi,t +1 ) i r mci,t Fnir ,t − wr,t = mcr,t Hnr r ,t − wr,t , and 1 = Et t +1|t mci,t +1 Fki ,t +1 + (1 − δi ) , where t + 1|t denotes the household’s stochastic discount factor. Households and Self-Employment. Households are the ultimate owners of all firms. They consume and send their members to search for employment opportunities, which introduces a labor-force participation margin. They also make decisions over the creation of new salaried firms subject to the evolution of salaried firms, taking individual salaried-firm profits as given. To capture an important component of the labor-market structure in LA, in addition to sending members to search for salaried positions in categories r and i, the household can also send its members to work in self-employment. A measure ne,t of self- employed individuals use their own labor to operate owner-only firms. Thus, households receive income from salaried-firm profits d ˜t Nt ; wage income from employed salaried workers in the two categories wi,t ni , i,t i i r r wr,t nr,t , and wr,t nr,t ; and income from production by self-employed individuals pe,t zt ze ne,t , where ze denotes exogenous self-employment productivity. These resources are used to finance consumption expenditures ct and the cost of salaried-firm creation fe Ne,t . The perceived evolution of, respectively, i salaried employment, r salaried employment, and self- employment that households are subject to are given by nr,t = (1 − zρs ,t ρs )nr,t −1 + sr,t f (θr,t ), ni,t = (1 − zρs ,t ρs )ni,t −1 + si,t f (θi,t ), ne,t = (1 − zρe ,t ρe )ne,t −1 + se,t φe , where sj,t denotes the measure of searchers in category j ∈ {i, r, e}, 0 < φ e ≤ 1 is the exogenous probability that a household member searching for a self-employment opportunity finds one, 0 < ρ e < 1 is the ex- ogenous self-employment exit probability, and zρe ,t is a self-employment destruction shock. Thus, sectoral labor-force participation is defined as lfpe,t = ne,t + (1 − φe )se,t , lfpi,t = ni,t + (1 − f (θi,t ))si,t , lfpr,t = nr,t + (1 − f (θr,t ))sr,t . The labor-force participation rate in the economy is therefore given by lfpt = lfpe,t + lfpi,t + lfpr,t and the unemployment rate can be defined as (1 − φe )se,t + (1 − f (θi,t ))si,t + (1 − f (θr,t ))sr,t urt ≡ . lfpt The household maximizes ⎡ 1 ⎤ ∞ zκr ,t κ j (lfp j,t )1+ χ 1 c1−σc zκi ,t κi (lfpi,t )1+ χ E0 β ⎣ t t − − ⎦, 1 − σc 1+ 1 χ 1+ 1 χ t =0 j∈{r,e} The World Bank Economic Review 169 subject to its budget constraint, the evolution of salaried firms defined earlier, and the perceived evolution of each category of employment, where σ c , κ j , κ i , χ > 0, zκr ,t is a shock to the disutility of labor-force participation of r workers and self-employed individuals, and zκi ,t is a shock to the disutility of labor-force participation of i workers. The solution to the household’s problem delivers a labor-force participation condition for each of the three categories of employment (e, i, and r): Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 1 1 zκr ,t κr (lfpr,t ) χ 1 r i 1 zκr ,t +1 κr (lfpr,t +1 ) χ −σc = wr,t (1 − ωt ) + wr,t ωt + Et t +1|t (1 − zρs ,t +1 ρs ) −1 −σc , ct f (θr,t ) f (θr,t +1 ) ct +1 1 1 zκi ,t κi (lfpi,t ) χ 1 1 zκi ,t +1 κi (lfpi,t +1 ) χ −σc = wii,t + Et t +1|t (1 − zρs ,t +1 ρs ) −1 −σc , ct f (θi,t ) f (θi,t +1 ) ct +1 1 1 zκr ,t κe (lfpe,t ) χ 1 1 zκr ,t +1 κe (lfpe,t +1 ) χ −σc = pe,t zt ze + Et t +1|t (1 − zρe ,t +1 ρe ) −1 −σc , ct φe φe ct +1 as well as a firm creation condition, fe = (1 − δ )Et t +1|t d˜t +1 + fe , −σc −σc ˜t were defined earlier. where t +1|t = β ct +1 /ct and average salaried-firm profits d Wage Determination, Symmetric Equilibrium, and Market Clearing. Real wages are determined via bilateral Nash bargaining between firms and salaried workers, where 0 < ν < 1 is the bargaining power of workers (see supplementary online appendix S2 for the value equations associated with salaried firms and the household). Finally, in the presence of labor search frictions, endogenous firm entry, and costly technology adoption, the economy’s resource constraint is given by Yt = ct + (ki,t +1 − (1 − δi )ki,t ) + ψ (vr,t + vi,t ) + fe Ne,t + fi Ni,t . Supplementary online appendix S3 presents the list of equilibrium conditions. Barriers to Firm Entry, Digital-Adoption Costs, and Digital Adoption. Following the empirical evidence and discussion in Finkelstein Shapiro and Mandelman (2021), there is a positive link between the sunk cost of salaried-firm creation and the cost of digital adoption. Specifically, fe = λf fi with λf > 1. This assumption implies that any changes in digital-adoption costs, which directly affect the number of firms adopting ICT in the model, will also change firm entry costs and vice versa. More broadly, this assumption implies that the model is consistent with the negative link between barriers to firm entry and firm digital adoption in LAC (see fig. S1.1 in the supplementary online appendix). Mapping of Employment and Firm Categories in the Model to the Data. For the purposes of mapping the categories of employment and salaried firms in the model to the data, salaried workers in r firms represent informal salaried workers, while salaried workers in i firms (regardless of whether they com- plement ICT or are imperfectly substitutable with the ICT composite) are formal salaried workers. Then total informal employment is given by the sum of informal salaried workers and self-employed workers. In turn, the informal employment share is defined as the total informal employment divided by the total (formal and informal) employment. Thus, r firms represent informal salaried firms while i firms represent formal salaried firms. 4. Quantitative Analysis Models with endogenous firm entry exhibit a love-for-variety component that is not present in empirical measurements of the CPI (Bilbiie, Ghironi, and Melitz 2012). Amid salaried-firm creation, this implies that model-based quantity variables that are compared to their empirical counterparts need to be adjusted to purge the variety effect. With this in mind, denote by om,t a quantity variable in the model that is inclusive 170 Finkelstein Shapiro, Nuguer, and Novoa of the variety effect. Then, given the aggregation of salaried and self-employment output categories, the model-based quantity variable od,t = t om,t , where 1 1−φy 1−φy t = Nt 1−ε + 1 is comparable to its empirical counterpart.3 In the quantitative analysis, all model-based quantity variables Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 are expressed such that they are comparable to their counterparts in the data. 4.1. Calibration of Baseline Economy The model is calibrated to Mexico. As such, the calibration uses several parameter values from the emerg- ing economy literature and related studies that also consider Mexico as their case study. The calibration of the model is based on a scenario without policy interventions. Having matched the dynamics of the COVID recession and ongoing recovery, the model is then used to analyze how a policy that increases firm digital adoption by permanently reducing the barriers to adoption and that is implemented at the trough of the recession shapes the recovery path of the economy. Parameters from the Literature. A period in the model represents a quarter. The subjective discount factor is β = 0.985, the depreciation rate of ICT capital at δ i = 0.025, and the risk aversion parameter is σ c = 2. The exogenous salaried-firm exit rate is δ = 0.025 based on available data on average firm exit rates, and the minimum level of idiosyncratic productivity is normalized to amin = 1. The variables ε and kp are set to 4 and 4.2 respectively, which satisfy ε − 1 < kp and are consistent with average markups in the literature. The exogenous productivity of the self-employed is normalized to ze = 1. Turning to the labor market, the bargaining power of salaried workers is ν = 0.5, which is a common value in the search-and-matching literature. The separation probabilities are ρ e = 0.044 and ρ s = 0.04, and φ e = 0.20, where the value for φ e is consistent with the probability of transitioning to self-employment in Mexico (Bosch and Maloney 2008). Similar to Finkelstein Shapiro and Mandelman (2021), the baseline elasticity of labor supply on the extensive margin is χ = 0.26, and φ y = 5 is chosen as a baseline. Based on Eden and Gaggl (2018), λi = 0.9, φ i = 0.47, and λk = 0.3, implying imperfect substitutability between the ICT–labor composite and r employment. Calibrated Parameters. The values of the remaining parameters—λf , α k , ξ , κ e , κ i , κ r , ψ , fi , and zi — are chosen to match select first-moment targets using Mexican data. Targets associated with the labor market are based on data for 2020Q1—that is, the quarter prior to the onset of the pandemic recession. Variables at lower frequencies are based on the latest available data prior to 2020. In particular, the model replicates a cost of creating a business of 18 percent of GDP per capita (per World Bank Enterprise Survey data), a cost of posting vacancies equivalent to roughly 3.5 percent of the average wage (in line with the average cost of hiring; see Levy (2007)), a share of self-employment in total employment of 22.7 percent, a share of employment with tertiary education (represented by nii ) in total employment of 16 percent, a labor-force participation rate of 61.3 percent, an unemployment rate of 4 percent (per ENOE data), a share of salaried firms that use digital technologies of 41.5 percent (per business digital-adoption data from the World Bank), a share of total expenditures on ICT (capital and adoption) of 1.4 percent of GDP, and a share of expenditures on ICT adoption costs in the total cost of using ICT of 10 percent (per the ITU-D ICT Database). The calibrated parameter values are as follows: λf = 0.00061947, α k = 0.0456, ξ = 0.8963, κ e = 419.6129, κ i = 7, 166.6, κ r = 40.3102, ψ = 0.1213, fi = 0.00028, and zi = 1.9686. Calibration of Shocks: Onset of COVID Recession and Post-Recession Path. To capture the contraction at the onset of the COVID recession and the subsequent post-recession path, six shocks are introduced: a 3 See Cacciatore et al. (2016) for a similar adjustment to model-based quantity variables in the context of a small open economy with firm entry, and supplementary online appendix S4 for more details. Note that in the absence of salaried- firm entry and exit, t collapses to a constant which, without loss of generality, can be normalized to 1 so that od,t = om,t . The World Bank Economic Review 171 shock to aggregate productivity (reflected in exogenous changes in zt ), a shock to the disutility of labor- force participation of r workers and self-employed individuals (reflected in exogenous changes in zκr ,t ), a shock to the disutility of labor-force participation of i workers (reflected in a exogenous changes in zκi ,t ), a shock to the matching efficiency of r employment (reflected in exogenous changes in zmr ,t ), a shock to the self-employment separation probability (reflected in exogenous changes in zρe ,t ), and a shock to the salaried employment separation probability (reflected in exogenous changes in zρs ,t ). The shocks follow Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 independent AR(1) processes: z zt = (1 − z )z + z zt −1 − t, κr zκr ,t = (1 − κ )zκr + κ zκr ,t −1 + t , κi zκi ,t = (1 − κ )zκi + κ zκi ,t −1 + t , mr zmr ,t = (1 − m )zmr + m zmr ,t −1 − t , ρe zρe ,t = (1 − ρe )zρe + ρe zρe ,t −1 + t , ρs zρs ,t = (1 − ρs )zρs + ρs zρs ,t −1 + t , j where z = zκr = zκi = zmr = zρe = zρs = 1, 0 < z , κ , m , ρe , ρs < 1, and t ∼ N (0, σ j ) for j ∈ {z, κ r , κ i , mr , ρ e , ρ s }. These six shocks embody several relevant aspects that characterized the onset of the COVID pandemic and its aftermath. First, by its very nature, the pandemic significantly hindered firms’ and individuals’ ability to produce a given amount of output with the same inputs (capital and/or labor), a fact that would be captured by movements in exogenous aggregate productivity z. At the same time, participating in the labor market—either by working or by searching for employment—at the onset of COVID was accom- panied by greater risk of exposure and illness. As such, the pandemic increased the cost of labor-force participation, which is reflected in changes in zκr and zκi , where individuals’ ability to participate in the labor market may differ depending on the employment category—hence the differential shocks to par- ticipation in r employment or self-employment and participation in i employment. In turn, by restricting mobility, the pandemic also hindered the matching process between firms and potential workers in high- contact occupations that do not rely on, or cannot be performed with, digital technologies—in the model, these occupations belong to the r employment category. This particular pandemic-related disturbance is embodied in changes in matching efficiency in those particular occupations and reflected in a changes in zmr . Finally, the lockdowns and reduction in economic activity associated with the onset of COVID generated significant self-employment job destruction and salaried employment job destruction, which would be reflected in changes in zρe ,t and zρs ,t , respectively. All told, the shocks embody the broad fea- tures of the COVID recession, with shocks that affect participation in the labor market and the matching process between workers and firms being particularly relevant for capturing the dynamics of the labor market. To discipline the quantitative analysis, the shock processes are parameterized to jointly match the dynamics of six time series: real GDP, total employment, self-employment, the informal employment share, the unemployment rate, and the labor-force participation rate from the onset of the COVID recession until 2021Q3. Given the unprecedented size of the COVID shock, the highly non-linear nature of the COVID recession, and the richness of the model, the model is solved using a perfect-foresight solution that makes use of the full non-linear version of the model (Juillard 1996). Figure 2 shows the realized paths of zt , zκr ,t , zκi ,t , zmr ,t , zρe ,t , and zρs ,t . In turn, fig. 3 compares the model-based paths of real GDP, total employment, self- employment, the share of informal employment, the unemployment rate, and the labor-force participation rate to their empirical counterparts. Per fig. 2, the onset of the COVID recession in the model stems from reductions in aggregate produc- tivity z and in the efficiency of the matching process associated with r employment zmr , and by an increase 172 Finkelstein Shapiro, Nuguer, and Novoa Figure 2. Model Shocks Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Source: Authors’ model simulations. Note: The figure shows the path of the shocks to aggregate productivity z, the disutility of labor-force participation of r workers and self-employed individuals zκr , the disutility of labor-force participation of i workers zκi , the matching efficiency of r employment zmr , the self-employment separation probability zρe , and the salaried employment separation probability zρs that match the dynamics of real GDP, total employment, self-employment, the informal employment share, the unemployment rate, and the labor-force participation rate from 2020Q1 to 2021Q3. in both the disutility of participation across employment categories and in the probabilities of job sepa- ration in salaried and self-employment. The shock series also suggest that the aftermath of the COVID recession is characterized by a rebound in aggregate productivity, persistently low matching efficiency for r workers, persistent shocks that prevent labor-force participation and salaried job destruction probabil- ities from returning to their pre-COVID levels, and a somewhat rapid normalization in self-employment job destruction. While these shocks are able to replicate the data relatively well, compared to the data, the model generates a larger contraction in the share of informal employment and underestimates the ex- tent of the reduction in labor-force participation.4 These limitations notwithstanding and from a broader perspective, the model is able to capture key aspects of the COVID recession. 4 The model cannot quantitatively match the contraction in labor-force participation without generating a contraction in GDP that is significantly greater than what is observed in the data. One potential explanation behind this limitation is the absence of household heterogeneity. The World Bank Economic Review 173 Figure 3. Onset and Aftermath of COVID Recession in Data vs. Targeted Variables in Benchmark Model Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Source: Authors’ model simulations and comparison with data. Note: The data series are obtained from the Saint Louis Federal Reserve Economic Database (FRED) and National Institute of Statistics and Geography (INEGI). 4.2. Firm Digital-Adoption Policies To understand the role of firm digital adoption, consider a policy that increases the share of i firms in the economy, Ni /N, by 1 percentage point in the steady state. This increase in Ni /N is engineered via a permanent reduction in the fixed cost of digital adoption fi . Recalling that fi and the sunk cost of salaried- firm creation fe are linked—specifically, fe = λf fi with λf > 1—the reduction in fi also reduces the barriers to entry for salaried firms. This experiment is similar in nature to the implementation of structural reforms that tackle barriers to entry in models with endogenous firm entry (see, for example, Cacciatore and Fiori (2016), or Cacciatore et al. (2016), Cacciatore et al. (2021), among others). Increasing the share of i firms in the economy by 1 percentage point relative to its baseline level entails an 18-percent reduction in fi from its baseline value.5 The policy increases the steady-state share of ICT investment in GDP from a baseline of 0.78 percent to 0.83 percent of GDP. 5 For reference, per International Telecommunication Union (ITU) data, the cost of a mobile-broadband basket with a monthly data allowance of 1 GB in Mexico fell from 3.32 percent of gross national income per capita in 2008 to 0.52 percent in 2019—an average reduction of 7.7 percent per year. In the quantitative experiments, the long-run reduction in the cost of digital adoption as a share of income per capita is roughly 18 percent. 174 Finkelstein Shapiro, Nuguer, and Novoa Table 1. Steady-State Changes in Response to Greater Firm Digital Adoption Variable Percent change relative to no-policy baseline Salaried firms (N) 32.37 Firms using ICT (Ni ) 35.58 Ave. firm productivity 0.13 Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Total output 0.51 Consumption 0.29 i worker real wage 3.20 r worker real wage 2.39 Total self-employment income −4.61 Total informal labor income −3.73 Total labor income 0.12 Perc.-pt. change relative to no-policy baseline Self-employment rate −0.30 Informal employment share −0.64 i employment share 0.10 Unemployment rate −0.07 LFP rate −1.06 Share of firms using ICT (Ni /N) 1.00 Source: Authors’ model-based calculations. Note: Perc.-pt. change denotes percentage-point change. ICT denotes information and communications technologies. Total informal labor income is defined as the sum of total informal salaried labor income and total self-employment income. Total labor income is defined as the sum of informal salaried labor income, formal salaried labor income, and total self-employment income. 4.2.1. Long-Run Effects Summary of Results. Table 1 shows the long-run changes of select variables in response to the policy by considering how the steady state of the economy changes with lower firm digital-adoption costs. First, note that both the number of firms that use ICT, Ni , and the total number of salaried firms, N, increase. Therefore, the 1-percentage-point increase in Ni /N stems from a larger increase in Ni relative to N. This change in the composition of salaried firms towards those that use ICT is reflected in an increase, albeit moderate, in average salaried-firm productivity. Second, in response to the policy, output and consumption increase by roughly 0.50 and 0.30 percent, respectively. The policy also leads to greater real wages across salaried worker categories, which increase by roughly 3 percent on average, with a somewhat larger percentage increase in the wage of i salaried workers relative to r workers. Ultimately, the increase in real wages, coupled with the reallocation of employment towards i firms, contributes to greater total labor income—the sum of total labor income from formal workers, informal salaried workers, and self- employed individuals—even as self-employment and self-employment income fall. Interestingly, the policy generates a decline in labor-force participation of roughly 1 percentage point, as well as a decline in the level of total employment of roughly 1.65 percent (not shown). However, by increasing the number of firms using ICT and the total number of salaried firms, the policy also shifts the composition of employment towards greater labor formality. This is reflected in a decline in the shares of self-employment and informal employment and an increase in the share of i employment. While not shown, the levels of self-employment and informal employment also fall. Thus, the decline in informality takes place in both absolute terms and relative to total employment. On net, given the change in labor- force participation and the reallocation of workers across employment categories, the unemployment rate exhibits a reduction, albeit small. The response of labor-market variables to the policy hinges on how the equilibrium change in consumption affects participation decisions, and therefore on the degree of household risk aversion. The World Bank Economic Review 175 Finally, given that under the policy, the wage of i workers grows more than the wage of r workers, total self-employment income falls, and the levels of self-employment and informal salaried employment fall while the levels of formal employment increase, the policy generates an increase in total formal labor income and a reduction in total informal labor income (the sum of total self-employment income and total informal salaried income). This translates into an increase in labor income inequality between i workers and r workers, and between i workers and self-employed individuals. In a context with a single Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 representative household and full consumption insurance—a standard assumption in search models—the asymmetric effect of the policy on total formal and informal labor income is not an issue since household members pool their resources to consume the same amount ex post. However, in a more realistic context with household heterogeneity, where some households may not have household members who can work in i firms, greater labor income inequality between employment categories may translate into greater consumption inequality between households. Economic Mechanisms. To better understand the results in table 1, especially those associated with the labor market, note that the policy-induced reduction in fi (and therefore in fe as well) reduces the marginal cost of creating new salaried firms and the marginal cost of adopting the ICT production technology. As a result, more salaried firms enter, bolstering the demand for r and i labor, improving labor-market conditions for salaried workers and, in doing so, real wages (via greater labor-market tightness, which is a component of the real wage in both salaried-firm categories). At the same time, the reduction in barriers to entry and ICT adoption makes self-employment less attractive, leading to a reduction in self-employment. The latter contributes to the reduction in the informal employment share. Despite the increase in real wages, which all else equal encourages more search for salaried jobs, the equilibrium measure of individuals searching for salaried jobs actually declines with the policy. This result traces back to the strength of the income effect, where the latter stems from the positive impact of the policy on salaried wages and on consumption. As such, its quantitative impact is therefore influenced by the household’s degree of relative risk aversion. To see this more clearly, without loss of generality, consider the optimal participation decision for i workers in steady state, which can be written as 1 κi (lfpi ) χ (1 − β (1 − ρs )) + β (1 − ρs ) = wii . c−σc f (θi ) Given the value of σ c in the baseline calibration, which is consistent with those in the emerging econ- omy business cycle literature, the quantitative change in the equilibrium job-finding probability and in consumption—which embodies the income effect—are such that sectoral salaried search and labor-force participation decline.6 The equilibrium response of searchers and participation across salaried employ- ment categories, coupled with the decline in participation by the self-employed, ultimately leads to an equilibrium reduction in the total labor-force participation rate. The reduction in the measure of searchers across employment categories and the increase in job-finding probabilities stemming from greater vacancy postings by salaried firms leads to a reduction, albeit small, in the unemployment rate. Under the baseline calibration, the average productivity of i firms is endogenously greater relative to the productivity of r firms. The increase in the number of firms that adopt the ICT technology narrows this average productivity differential, but by increasing the share of i firms in the economy and there- fore changing the composition of salaried firms towards those that are endogenously more productive, it increases average firm productivity at the economy-wide level. The resulting change in the composition of salaried firms offsets the reduction in self-employment production and is strong enough to ultimately bolster total output. 6 This result is also present when it comes to the labor-force participation decision associated with jobs in r firms. A simple counterfactual experiment with household risk neutrality confirms the quantitative relevance of the income effect on labor-force participation across salaried employment categories. 176 Finkelstein Shapiro, Nuguer, and Novoa Figure 4. Labor-Market and Macroeconomic Dynamics amid COVID: Benchmark vs. Policy Scenario Source: Authors’ model simulations. Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Note: Perc.-pt. denotes percentage-point. LFP denotes labor-force participation. ICT denotes information and communications technologies. Model-based formal i ni + w i ni ). Model-based informal salaried labor income is defined as w r nr . Model-based self-employment income is defined as salaried labor income is defined as (wr r i i r r r nr + p z z n ). Total labor income is the sum of informal salaried labor income, formal salaried labor income, pe,t zt ze ne,t . Informal labor income is given by (wr r e,t t e e,t and total self-employment income. All relevant variables are expressed in data-consistent terms. 4.2.2. Model Response to COVID Recession and Recovery Process under Policy Summary of Results: No Policy vs. Digital-Adoption Policy Comparison. Figure 4 shows the responses of GDP, the unemployment rate, the labor-force participation rate, the total employment, the informal salaried employment share, the self-employment share, the informal employment share, the formal and informal labor income, the new salaried firms, the salaried firms, and the share of firms using ICT tech- nology to the set of shocks that replicate the dynamics of the Mexican economy under COVID. The solid line shows the benchmark (no-policy) scenario while the dash-dotted line shows the response under the digital-adoption policy. As the figure suggests, the policy is able to offset the sharp reduction in salaried-firm creation that would otherwise take place, leading to a steady expansion in the measure of salaried firms. The fact that The World Bank Economic Review 177 the policy lowers the barriers to digital adoption is reflected in a larger increase in the share of salaried firms that adopt digital technologies relative to the baseline (no-policy) scenario. By propping up the measure and share of salaried firms that adopt digital technologies, the policy also limits the extent to which formal employment contracts (not shown) and changes the composition of employment towards greater labor formality. The policy has negligible quantitative effects on the self-employment share, but not on the share of informal salaried employment, which falls by significantly more compared to the Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 baseline scenario. The response of informal salaried employment ends up being the main driver of the larger contraction in the informal employment share, where the latter is a broader measure of labor informality that incorporates both informal salaried employment and self-employment. The asymmetric effect of the policy on formal and informal employment translates into a more subdued reduction in formal labor income but a larger fall in informal labor income. In this sense, the policy exacerbates labor income inequality between formal and informal workers and favors those who are employed in firms that use digital technologies, even as the policy bolsters the overall measure of salaried firms (both formal and informal). This point is important given that even in the absence of the policy, informal labor income exhibits a significantly larger contraction relative to formal labor income, and the digital-adoption policy makes this contraction even greater. Moreover, by causing a larger reduction in informal employment relative to the baseline scenario, the policy ends up generating a larger increase in the unemployment rate but the policy-induced expansion in unemployment is relatively short-lived. Finally, the policy initially limits the extent of the contraction in total employment and GDP at the onset of the recession. While these effects are quantitatively small, they become increasingly larger roughly three-quarters after the onset of the recession and ultimately generate non-trivial differences with respect to the baseline scenario. Given the unique magnitude of the COVID recession (note the y-axes in fig. 4), fig. 4 cannot fully capture the quantitative extent of the policy’s positive impact on total employment and aggregate eco- nomic activity. To better appreciate the quantitative effects of the policy, fig. 5 plots the percentage-point difference between the response of the economy under the policy and the baseline (no-policy) economy for select variables of interest. The figure makes clear that compared to the baseline (no-policy) scenario, the policy is successful in limiting the extent of the contraction in total employment, total labor income, and GDP, while generating a short-lived expansion in unemployment that traces back to the policy’s impact on the share of infor- mal employment. At the same time, fig. 5 highlights the improvement in total labor income vis-a-vis the baseline scenario stemming from the policy, as well as the policy’s significant adverse impact on informal- employment labor income. The differential response of total labor income and informal labor income suggests that formal labor income unambiguously benefits from the policy, and implies a widening of la- bor income differentials between formal workers and informal workers. Since the framework in this paper follows the search literature by assuming perfect consumption insurance at the household level, what ul- timately matters for the household is the response of total labor income. However, the policy-induced widening of labor income differentials becomes relevant in a more realistic context where a significant share of households rely primarily on labor income from informal employment and therefore would not directly benefit from relative improvements in formal labor income due to the policy. Economic Mechanisms. In the benchmark economy, the shocks initially lead to a sharp decline in the number of new salaried firms, which translates into a persistent contraction in the number of salaried firms. Given the simultaneous adverse shocks to aggregate productivity, salaried job destruction, and the matching process for r workers, salaried firms initially reduce the number of job vacancies associated with r employment but increase the number of job vacancies associated with i employment. As the economy rebounds after the initial set of shocks, new salaried-firm creation recovers sharply, thereby limiting the decline in the number of salaried firms that originally took place. The recovery in new salaried firms pushes salaried firms to expand their workforce after the initial contraction. At the same time, given the shocks that allow the model to replicate the COVID recession, self-employment also 178 Finkelstein Shapiro, Nuguer, and Novoa Figure 5. Labor-Market and Macroeconomic Dynamics amid COVID: Benchmark vs. Policy Scenario Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Source: Authors’ model simulations. Note: The figure shows the difference between the response of the economy under the policy and the no-policy scenarios. Perc.-point diff. denotes percentage-point difference. ICT denotes information and communications technologies. LFP denotes labor-force participation. Model-based formal salaried labor income is defined as (wri ni + w i ni ). Model-based informal salaried labor income is defined as w r nr . Model-based self-employment income is defined as p z z n . Informal labor income r i i r r e,t t e e,t is given by (wr r nr + p z z n ). Total labor income is the sum of informal salaried labor income, formal salaried labor income, and total self-employment income. r e,t t e e,t All relevant variables are expressed in data-consistent terms. exhibits a rebound after its initial sharp contraction. The behavior of informal salaried employment and self-employment has direct implications for labor-market conditions and contributes to both bringing the unemployment rate back down from its peak and bolstering informal labor income. By reducing barriers to salaried-firm entry and digital adoption at the trough of the recession, the policy effectively offsets the sharp contraction in new salaried-firm creation that otherwise takes place, leading not only to an equilibrium rise in the number of salaried firms, but also to a faster recovery. Since the policy changes the composition of total employment towards greater formality (recall table 1 and figs 4 and 5), salaried and total (salaried and self-employed) informal labor income remain more subdued compared to the benchmark economy. However, by limiting the contraction in formal employment, the policy not only limits the contraction in total formal salaried income, but also bolsters the recovery of formal income by fostering greater ICT adoption and improving the productivity profile of firms. On net, the response of total formal salaried income as the economy recovers more than offsets the contraction in informal income, which explains the earlier recovery in total labor income. The World Bank Economic Review 179 It is worth noting once again that the contraction in informal labor income in the absence of policy is significantly larger than the contraction in formal labor income, and that the policy exacerbates the difference in the dynamic responses of labor income in the two employment categories. As a result, the policy not only increases income inequality between formal and informal workers in the steady state, but also increases income inequality between these two groups as the economy recovers from the shocks that rationalize the COVID recession. Therefore, considering the response of macroeconomic aggregates Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 alone, such as GDP, may not provide a complete picture of the effects of the policy in a context where some households may not benefit from the policy’s positive effect on formal labor income. 4.2.3. Post-Shock Labor-Market and Macroeconomic Dynamics under Policy: Individual Shocks As described in Section 4.1, the model replicates the onset of the COVID recession and the subsequent path of the economy with six shocks: a shock to aggregate productivity, a shock to the disutility of labor- force participation of r workers and self-employment workers, a shock to the disutility of labor-force participation of i workers, a shock to the matching efficiency of r employment,and shocks to the salaried employment and self-employment job separation probabilities. Figures S5.1 through S5.6 of the sup- plementary online appendix present the responses of the economy in the baseline (no-policy) and the digital-adoption-policy scenarios to each individual shock. The following paragraphs briefly summarize the main takeaways from considering each shock in isolation. Shock to Aggregate Productivity (Supplementary Online Appendix Fig. S5.1). The shock generates a short-lived contraction in GDP, total employment, labor income across employment categories, the measure of new salaried firms and the measure of firms that use ICT, labor-force participation, and un- employment, as well as an expansion in both the share of self-employment and the share of informal employment. This last finding is consistent with the well-known countercyclicality of self-employment in typical recessions in emerging economies, including Mexico. The behavior of informal employment and labor-force participation explains the short-lived reduction in unemployment. Similar to the results when all shocks are active, by bolstering the share of salaried firms that adopt digital technologies, the policy leads to an increase in salaried firms, limits the extent of the contraction in GDP, and leads to an earlier recovery. The policy generates a sharp expansion in total employment, as well as a decline in the share of informal employment, which are ultimately reflected in an increase in the labor-force participation and unemployment rates. The policy also exacerbates the differential between formal and informal labor income. Finally, in contrast to the results under all shocks, the policy does have quantitative effects on the share of self-employment in total employment. Shock to the Disutility of Labor-Force Participation of r-Employment and Self-Employment Categories (Supplementary Online Appendix Fig. S5.2). The shock generates a persistent contraction in GDP, total employment, labor income across employment categories, the measure of new salaried firms and the measure of firms that use ICT, labor-force participation, and unemployment, as well as an expansion in the share of self-employment, but not in the share of informal employment. Given the sectoral nature of the shock, even though the measure of salaried firms falls, the share of firms that use digital technologies expands. The policy leads to a sharp increase in new salaried firms that limits the extent of the contraction in GDP and total employment and the expansion in the share of self-employment; it bolsters formal labor income and further reduces the share of self-employment and informal employment, leading to a sharper reduction in informal labor income. Shock to the Disutility of Labor-Force Participation of i-Employment Category (Supplementary On- line Appendix Fig. S5.3). The shock generates a persistent contraction in GDP, total employment, labor income across employment categories, the measure of firms that use ICT, labor-force participation, and unemployment, as well as expansions in the share of self-employment and in the share of informal em- ployment. A key difference relative to the results under all shocks is that informal labor income increases 180 Finkelstein Shapiro, Nuguer, and Novoa sharply, mainly due to the expansion in informal salaried and self-employment. The unemployment rate increases, mainly due to the adverse effects of the shock on i employment. The policy generates more subdued responses in all variables, mainly by bolstering the measure of new salaried firms. Similar to the results under all shocks, the policy leads to a sharper expansion in the unemployment rate, mainly by limiting the extent to which self-employment and informal employment respond to the shock. Shock to r-Employment Matching Efficiency (Supplementary Online Appendix Fig. S5.4). The shock Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 has similar qualitative effects to those stemming from a shock to the disutility of labor-force participation of r-employment and self-employment: it reduces GDP, total employment, labor income, new salaried firms, the informal employment share, and the labor-force participation and unemployment rates, and increases the share of self-employment and the share of firms that adopt digital technologies. The policy has similar effects to those observed when all shocks are active. Shock to Self-Employment Job Destruction (Supplementary Online Appendix Fig. S5.5). The shock generates a deep but short-lived contraction in GDP, total employment, informal labor income, and the measure of new salaried firms, and a sharp reduction in the shares of self-employment and informal employment. The reduction in self-employment plays a key role in driving the significant expansion in the unemployment rate. The policy does not have any meaningful effects on the self-employment share, but does affect salaried labor by bolstering the measure of new salaried firms. This, in turn, contributes to smaller contractions in total employment and GDP. As was the case in a scenario where all shocks are active, the policy generates an asymmetric response in formal and informal labor income. However, the policy does not alter the response of unemployment. Shock to Salaried Job Destruction(Supplementary Online Appendix Fig. S5.6). In stark contrast with the response of the economy to the other shocks, a shock to salaried job destruction increases GDP and total employment. This is a standard result in search-and-matching models, where job destruction shocks alone push firms to replenish their workforce by hiring new workers. At the same time, given the nature of the shock, the expansion in GDP goes hand in hand with an expansion in the unemployment rate. The policy amplifies the response of most variables to the shock, leading to a sharper expansion in GDP and total employment. Summary and Main Takeaways from Individual Shocks. A unique characteristic of the COVID reces- sion was the sharp decline in the share of self-employment and informal employment, where these shares tend to expand during typical recessions. The shock to self-employment job destruction plays a critical role in the model’s ability to replicate the reduction in informality shares during COVID. At the same time, the shocks to labor-force participation and the matching process for r-employment are critical to generat- ing the sharp reductions in GDP and labor-force participation that characterized the recession. Regardless of the type of shock, though, in the short term and relative to an environment without policy, a digital- adoption policy that increases the share of salaried firms that use ICT contributes to an improvement in total employment, formal labor income, new salaried firms, and GDP; a reduction in the share of informal employment and informal labor income; greater labor-force participation; greater unemployment; and a widening differential between formal and informal labor income. The effects of the policy are qualitatively similar in the short term and in the long term for GDP and all labor-market variables, except for the unemployment rate, the level of total employment, and the labor-force participation rate. Indeed, recalling that in the long run the policy reduces total employment, the labor-force participation rate, and the unemployment rate, in the short run, the policy causes the unemployment rate to briefly overshoot the no-policy scenario, total employment to fall by less than such a scenario, and the labor-force participation rate to be higher relative to the no-policy scenario. However, the policy leads to an improvement in total labor income and GDP, with asymmetric effects between formal labor income and informal labor income in both the short and the long run. The World Bank Economic Review 181 Table 2. Steady-State Changes: Lower Digital-Adoption Costs vs. Greater ICT Investment Variable Percent change relative to no-policy Percent change relative to no-policy baseline, lower digital-adoption costs baseline, greater ICT investment Salaried firms (N) 32.37 0.35 Firms using ICT (Ni ) 35.58 2.77 Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Ave. firm productivity 0.13 0.12 Total output 0.51 0.25 Consumption 0.29 0.10 i worker real wage 3.20 0.83 r worker real wage 2.39 0.04 Self-employed real income −4.61 −0.00 Total informal labor income −3.73 −1.32 Total labor income 0.12 0.15 Perc.-pt. change relative to no-policy baseline Perc.-pt. change relative to no-policy baseline Self-employment rate −0.30 −0.01 Informal employment share −0.64 −0.38 i employment share 0.10 0.03 Unemployment rate −0.07 −0.00 LFP rate −1.06 −0.02 Share of firms using ICT (Ni /N) 1.00 1.00 Source: Authors’ model-based calculations. Note: Perc.-pt. change denotes percentage-point change. ICT denotes information and communications technologies. Total labor income is defined as the sum of informal salaried labor income, formal salaried labor income, and total self-employment income. 4.2.4. Alternative Policy: Subsidy to ICT Investment Expenditures vs. Lower Digital-Adoption Costs To highlight the policy role of reducing digital-adoption costs, this section compares the benchmark find- ings to those of a policy experiment that introduces a subsidy on ICT capital investment expenditures. Specifically, recalling that the evolution of ICT capital is given by ki,t + 1 = (1 − δ i )ki,t + invt , let 0 < τ i,t < 1 be the subsidy rate on ICT investment expenditures invt . The optimal ICT capital accumulation decision is (1 − τi,t ) = Et t +1|t mci,t +1 Fki ,t +1 + (1 − τi,t +1 )(1 − δi ) . For comparability, the subsidy is covered with lump-sum taxes and tailored to achieve the same objective as the digital-adoption policy—mainly, an increase in the share of firms that use ICT of 1 percent in the steady state. This policy increases the baseline steady-state ratio of ICT investment to GDP from 0.78 percent of GDP to 0.90 percent of GDP (recall that the original policy, which focuses on lowering digital- adoption costs, increases the steady-state ICT investment–GDP ratio from a baseline of 0.78 percent to 0.83 percent of GDP). Table 2 compares the steady-state outcomes under the original policy to those under the alternative policy that subsidizes ICT investment. Qualitatively, the two policies have the same effect on firm digital adoption, labor-market outcomes, labor income by employment category, and macroeconomic aggregates. This result is expected given that the policy operates through the same channels as those described earlier. However, quantitatively, a subsidy to ICT investment generates much smaller positive changes in the share of firms that adopt digital technologies, labor income by employment category, consumption, and output. The main reason for the quantitative differences traces back to the effect of each policy on overall salaried-firm creation: for the same percentage-point increase in the share of firms that use ICT, the digital- adoption policy generates a much greater increase in the overall measure of salaried firms, where the latter contribute to the greater increase in employment and GDP. 182 Finkelstein Shapiro, Nuguer, and Novoa Figure 6. Labor-Market and Macroeconomic Dynamics amid COVID: Benchmark vs. Subsidy to ICT Investment Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Source: Authors’ model simulations. Note: The figure shows the difference between the response of the economy under the policy (Digital-Adoption Policy or ICT Investment Subsidy) and the no-policy scenario. Perc.-point diff. denotes percentage-point difference. ICT denotes information and communications technologies. LFP denotes labor-force participation. Total labor income is the sum of informal labor income (wr r nr + p z z n ), and formal labor income (w i ni + w i ni ), and is expressed in data-consistent terms. r e,t t e e,t r r i i Turning to the effects of the policy amid the COVID recession, fig. 6 presents the counterpart of fig. 5 for the policy that subsidizes ICT investment expenditures and compares the results of this alternative policy to those from the benchmark digital-adoption policy. As the figure suggests, subsidizing ICT investment to achieve the same increase in the share of firms that use ICT has a much weaker positive short-term impact on GDP, total labor income, and formal employment compared to a policy that reduces barriers to firm digital adoption via lower digital-adoption costs. The subsidy to ICT investment also has a weaker short- term effect on the unemployment rate and labor-force participation, effectively having no meaningful consequences for unemployment. All told, our quantitative analysis suggests that supporting greater firm digital adoption via lower barriers to the adoption of ICT technologies can be more impactful on total employment and The World Bank Economic Review 183 GDP in the short and long terms compared to a policy that subsidizes the cost of ICT capital expenditures. 5. Conclusion Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 In 2020 and as a result of the COVID pandemic, Latin America (LA) experienced the most dramatic contraction in employment and economic activity relative to other regions around the world. The adoption and use of digital technologies by firms was an important adjustment margin in response to the adverse economic effects of the COVID shock. Using a search-and matching-framework with firm entry and exit, where salaried firms can choose to adopt digital technologies and the labor-market and firm structure is consistent with the LA context, this paper analyzes how policies that bolster firm digital adoption shape the labor-market and economic recovery from COVID-19. Using Mexico as a case study in the region, the model can quantitatively cap- ture the dynamics of the labor market and output during COVID, including the behavior of labor-force participation and informal employment. A permanent reduction in the barriers to adopt digital technologies introduced at the trough of the recession contributes to an earlier recovery in GDP, total employment, and labor income, and to a larger expansion in the share of formal employment compared to a scenario without policy. Even though this policy induces a reduction in total employment in the long term, it improves the average productivity pro- file of firms, and leads to greater levels of GDP and labor income, a larger share of formal employment, and marginally lower unemployment. Therefore, fostering greater firm digital adoption can not only sup- port earlier labor-market and economic recoveries, but also leads to improved long-term macroeconomic outcomes. At the same time, the policy exacerbates the labor income differentials between formal and informal workers and points to a trade-off between improved macroeconomic outcomes and labor in- come inequality. This trade-off may be particularly relevant in environments where a significant share of households rely primarily on informal labor income. Data Availability Model codes are available on request. References Alfaro, L., O. Becerra, and M. Eslava. 2021. “EMEs and COVID-19: Shutting Down in a World of Informal and Tiny Firms.” NBER Working Paper 27360. Alon, T. M. Kim D. Lagakos and M. VanVuren. 2020. “How Should Policy Responses to the COVID-19 Pandemic Differ in the Developing World?” NBER Working Paper 27273. ———. 2021. “Macroeconomic Effects of COVID-19 Across the World Income Distribution.” Mimeo. Apedo-Amah, M. C., B. Avdiu, X. Cirera, M. Cruz, E. Davies, A. Grover, L. Iacovone et al. 2020. “Unmasking the Impact of COVID-19 on Businesses: Firm Level Evidence from across the World.” World Bank Policy Research Working Paper 9434. Bilbiie, F. O., F. Ghironi, and M. J. Melitz. 2012. “Endogenous Entry, Product Variety, and Business Cycles.” Journal of Political Economy 120(2): 304–45. Bosch, M., and W. Maloney. 2008. “Cyclical Movements in Unemployment and Informality in Developing Countries.” IZA Discussion Paper No. 3514. Cacciatore, M., R. Duval, G. Fiori, and F. Ghironi. 2016. “Short-Term Pain for Long-Term Gain: Market Deregulation and Monetary Policy in Small Open Economies.” Journal of International Money and Finance 68: 358–85. ———. 2021. “Market Reforms at the Zero Lower Bound.” Journal of Money, Credit and Banking 53(4): 745–77. 184 Finkelstein Shapiro, Nuguer, and Novoa Cacciatore, M., and G. Fiori. 2016. “The Macroeconomic Effects of Goods and Labor Markets Deregulation.” Review of Economic Dynamics 20: 1–24. Cirera, X., D. Comin, and M. Cruz. 2022. Bridging the Technological Divide: Technology Adoption by Firms in Developing Countries. World Bank. https://doi.org/10.1596/978-1-4648-1826-4. Den, Haan, W. J., G. Ramey, and J. Watson. 2000. “ Job Destruction and Propagation of Shocks.” American Economic Review 90(3): 482–98. Downloaded from https://academic.oup.com/wber/article/38/1/161/7249989 by Joint Bank/Fund Library user on 02 February 2024 Díaz de Astarloa, B., N. Mulder, S. Corcuera-Santamaría, W. Weck, L. Barreiros, R. Contreras Huerta, A. Patiño, and G. Marcee. 2021. “Post Pandemic COVID-19 Economic Recovery: Enabling Latin America and the Caribbean to Better Harness E-Commerce and Digital Trade.” Inter-American Development Bank Discussion Paper No. IDB- DP-00881. Economic Commission for Latin America and the Caribbean (ECLAC). 2020a. “Sectors and Businesses Facing COVID-19: Emergency and Reactivation.” Special Report COVID-19 No. 4, United Nations. ———. 2020b. “Universalizing Access to Digital Technologies to Address the Consequences of COVID-19.” Special Report COVID-19 No. 7, United Nations. ———. 2021. Economic Survey of Latin America and the Caribbean. United Nations. Eden, M., and P. Gaggl. 2018. “On the Welfare Implications of Automation.” Review of Economic Dynamics 29: 15–43. Finkelstein Shapiro, A., and F. S. Mandelman. 2021. “Digital Adoption, Automation, and Labor Markets in Developing Countries.” Journal of Development Economics 151: 102656. Juillard, M. 1996. “Dynare: A Program for the Resolution and Simulation of Dynamic Models with Forward Variables through the Use of a Relaxation Algorithm.” ICEPREMAP Working Papers (Couverture Orange) 9602. Lama, R., G. Leyva, and C. Urrutia 2022. “Labor Market Policies and Business Cycles in Emerging Economies.” IMF Economic Review, Vol. 70, pp. 300–337. Levy, S. 2007. “Can Social Programs Reduce Productivity and Growth? A Hypothesis for Mexico.” IPC WP37. Leyva, G., and C. Urrutia. 2023. “Informal Labor Markets in Times of Pandemic: Evidence for Latin America and Policy Options.” Review of Economic Dynamics 47: 158–85.