Digital Adoption, Automation, and Labor Markets in Developing Countries1 Alan Finkelstein Shapiro2 2 Tufts 3 Federal
Federico Mandelman3
University
Reserve Bank of Atlanta
3rd Joint IMF-OECD-World Bank Conference on Structural Reforms
1
The views in this paper are solely the responsibility of the author and should not be interpreted as reflecting the views of the Federal Reserve Bank of Atlanta or the Board of Governors of the Federal Reserve System.
Motivation I Rising concern over consequences of digital adoption + automation for labor market outcomes in advanced economies (AEs)
I Greater firm digital adoption: precursor to adoption of automation I May disrupt labor markets, lead to job loss
I Concerns extend to developing countries (DCs) as well I 60 percent of jobs susceptible to automation (World Bank, 2016; Schlogl and Sumner, 2018)
I Distinct employment structure (high rates of self-employment) + pervasive barriers to salaried firm entry
I Regulations, red tape, poor infrastructure I Underdeveloped credit markets, costly market access I High entry costs
What We Do I Both DC characteristics make consequences of firm digital adoption for empl., job loss less clear-cut
I ⇑ digital adoption ⇓ salaried firm-entry costs, ⇑ firm entry, ⇑ salaried empl, ⇓ SE
I ⇑ digital adoption ⇓ salaried empl. via automation, ⇑ SE I Impact on employment structure, job loss, unemployment?
What We Do
I Characterize link between firm digital adoption in DCs and 1. Barriers to (salaried) firm entry 2. Self-employment (SE) rates 3. Unemployment rates
I Build quantitative framework with equilibrium unemployment and SE, salaried firm entry, and digital (ICT) adoption margins
Cost of Starting a Business (% of Income Per Capita)
100 80
corr = -0.54***
60 40 20 0 0.2
0.4
0.6
0.8
New Firm Density (New Firms Per 1000 Individuals)
Firm Digital Adoption in Developing Countries 7 6
4 3 2 1 0 0.2
1
0.3
0.8
0.25
corr = -0.76***
0.6 0.4 0.2
0.4
0.6
Business Digital Adoption (BDAI)
0.4
0.6
0.8
Business Digital Adoption (BDAI)
0.8
Unemployment Rate
Self-Employment Rate
Business Digital Adoption (BDAI)
0 0.2
corr = 0.45***
5
corr = 0.05
0.2 0.15 0.1 0.05 0 0.2
0.4
0.6
0.8
Business Digital Adoption (BDAI)
Notes: World Bank World Development Report 2016, Enterprise Surveys, Entrepreneurship Report, WDI. *** denote sig. at 1 percent level.
Self-Employment and Business Digital Adoption (2016) Dep. Var.: Self-Empl. Rate
Business Digital Adoption (BDAI)
(1)
(2)
(3)
(4)
(5)
-1.170∗∗∗ (-12.92)
-0.495∗∗∗ (-3.87) -0.129∗∗∗ (-7.47)
-0.353∗∗∗ (-3.17) -0.0752∗∗∗ (-3.26) -0.608∗∗∗ (-3.70) -0.327∗∗∗ (-2.70)
-0.337∗∗∗ (-2.93) -0.0641∗∗∗ (-2.76) -0.691∗∗∗ (-4.22) -0.368∗∗∗ (-3.10) 0.0355 (0.98) 0.00120 (1.47)
1.036∗∗∗ (19.20) 0.584 118
1.843∗∗∗ (16.47) 0.697 117
1.567∗∗∗ (11.19) 0.745 117
1.466∗∗∗ (10.73) 0.763 114
-0.393∗∗∗ (-3.16) -0.0739∗∗∗ (-3.20) -0.668∗∗∗ (-4.07) -0.365∗∗∗ (-3.05) 0.0371 (1.02) 0.00114 (1.35) 0.0349 (1.21) 1.588∗∗∗ (10.24) 0.764 114
Log Real GDP Per Capita Industrial Empl. Share Services Empl. Share Min. Wage/VA per Worker Severance Payment Government Effectiveness Constant Adjusted R 2 Observations
Notes: World Bank WDI, Doing Business Report, and World Development Report 2016. Notes: SE rate is computed as the number of self-employed individuals divided by the labor force in 2016. BDAI: Business Digital Adoption Index. Log Real GDP per capita expressed in PPP terms. Severance payment: pay for redundancy dismissal for worker with 5 years of tenure (salary weeks). t-statistics in parentheses. Standard errors are heteroskedasticity-robust. *** denotes sig. at 1 level.
MODEL SKETCH
Model Structure: Salaried Firms I Economy is comprised of households (include SE) and salaried firms I Prospective new firms must incur sunk cost fe to start operating I Right after entry, draw idiosync. prod. a from distribution G (a) I 2 available technologies: I Regular r ⇒ uses salaried labor nrr I ICT i ⇒ uses salaried labor nri and nii + ICT capital ki I nii and ki are complements but nri and (ki , nii )-composite are imperfect substitutes
I Firms with a > ai pay fixed cost fi and use i technology
Salaried Firms (Continued) I With N salaried firms, fraction of i firms Ni /N = model-BDAI I Firms post vacancies vr , vi at costs ψr , ψi to attract workers I Key assumption (motivated by stylized facts): fe and fi are positively related
I fi = λf fe where 0 < λf < 1 I Reductions in fi , which all else equal increase Ni /N, are associated with reductions in fe I I I I
Firm digital adoption can help firms overcome barriers to entry Digital form-filling (e-governance) reduces red tape E-banking facilitates payments, credit access Access to e-commerce via digital platforms, expanded market access (goods and inputs)
Households and Self-Employment I Make labor-force-participation (LFP) decisions, including over SE I Differential (utility) costs associated with sending members to search for salaried r employment, i employment, or SE I Captures in reduced-form way relative ease of finding SE, r employment vs. i employment
I Choice over SE LFP = SE firm creation I SE produce using own labor ne
I Own salaried firms, make decisions over salaried-firm creation
I Total output is a CES composite of total salaried-firm output and SE output Model Details
QUANTITATIVE ANALYSIS
Quantitative Analysis I Baseline economy: economy with lowest BDAI = 0.20 in country sample
I Main calibration targets for analysis: I Predicted value of SE rate at baseline BDAI = 0.20 I Predicted value of unemployment rate at baseline BDAI = 0.20 I Predicted value of cost of starting business at baseline BDAI = 0.20
I Calibration of total LFP disutility parameters delivers asymmetric LFP costs across empl. categories
I Reduce fi (and, per our assumption, fe as well) to generate a BDAI range from 0.20 to 0.85 Functional Forms
Parameter Values
Cost of Starting a Business (% of Income Per Capita)
100 80
corr = -0.54***
60 40 20 0 0.2
0.4
0.6
0.8
New Firm Density (New Firms Per 1000 Individuals)
Firm Digital Adoption in Developing Countries 7 6
4 3 2 1 0 0.2
1
0.3
0.8
0.25
corr = -0.76***
0.6 0.4 0.2
0.4
0.6
Business Digital Adoption (BDAI)
0.4
0.6
0.8
Business Digital Adoption (BDAI)
0.8
Unemployment Rate
Self-Employment Rate
Business Digital Adoption (BDAI)
0 0.2
corr = 0.45***
5
corr = 0.05
0.2 0.15 0.1 0.05 0 0.2
0.4
0.6
0.8
Business Digital Adoption (BDAI)
Notes: World Bank World Development Report 2016, Enterprise Surveys, Entrepreneurship Report, WDI. *** denote sig. at 1 percent level.
Digital Adoption and Labor Markets: Data vs. Model Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Quantitative Analysis: Baseline Model I Model successfully generates I Negative relationship between firm digital adoption and SE, and absence of link between digital adoption and unemployment I Absence of link between digital adoption and unemployment
I Model also generates negative link between salaried-firm entry costs and firm digital adoption
I What happens if we break link between firm-creation costs fe and cost of tech. adoption fi ?
Delinking Digital Adoption Costs from fe : Changes in fi Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction, No Change in f e
Data: Linear Trend
Quantitative Analysis: Baseline Model
I If change in fi alone does not generate changes in unemployment rates across digital adoption range either, what drives disconnect between digital adoption and unemployment in the data?
I Asymmetric labor-force-participation (utility) costs between SE and salaried employment!
I Asymmetry embodies switching costs between SE and salaried empl. in reduced-form way
I To see this, assume that LFP cost parameter is the same for SE and salaried labor used in r technology (WLOG)
Cost of Starting a Business (% of Income Per Capita)
Symmetric LFP Costs Between SE and Salaried Empl. 100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Self-Employment Rate
Digital Adoption by Firms 1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Unemployment Rate
0.3
0.2
0.1
0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Model Prediction e = r
Data: Linear Trend
Testable Model Implications Model suggests the following: controlling for the cost of firm creation, changes in cost of digital adoption have no impact on SE rate
Dep. Var.: BDAI
(1)
(2)
Cost of Fixed Broadband
-0.454∗∗∗
Adjusted R 2 Observations
0.621∗∗∗ (30.69) 0.169 103
-0.196∗ (-1.77) -0.003∗∗∗ (-4.27) 0.640∗∗∗ (4.06) 0.330 103
Dep. Var.: Self-Empl. Rate
(1)
(2)
Cost of Fixed Broadband
0.614∗∗∗
0.248 (1.15) 0.004∗∗∗ (3.57) 0.265∗∗∗ (7.09) 0.274 103
(-4.96) Cost of Firm Creation Constant
(3.65) Cost of Firm Creation Constant Adjusted R 2 Observations
0.291∗∗∗ (8.68) 0.134 103
Notes: t-statistics in parentheses. *** and * denote sig. at 1 and 10 percent levels, respectively.
Conclusion I Rising concern over consequences of digital adoption, automation for labor market outcomes, but little work on developing countries (DCs)
I Document robust negative link between firm digital adoption and self-employment (SE) rates in DCs, but no link with unemployment
I Build framework with equilibrium unemployment, SE, salaried firm creation, and digital (ICT) adoption decisions
I Model highlights link between tech.-adoption costs and firm entry costs (switching costs between SE and salaried empl.) for rationalizing relationship between digital adoption and SE (unemployment) in DCs
APPENDIX
Business Digital Adoption and Routine Task Intensity 0.5
0 -0.5 -1 corr = 0.73*** -1.5 0.2
0.4
0.6
Latest Routine Task Intensity (RTI)
Initial Routine Task Intensity (RTI)
0.5
0 -0.5 -1 corr = 0.60*** -1.5 0.2
0.8
Business Digital Adoption (BDAI) 2014
0.8
0.5
0 -0.5 -1 corr = 0.76*** 0.4
0.6
Business Digital Adoption (BDAI) 2016
0.8
Latest Routine Task Intensity (RTI)
Initial Routine Task Intensity (RTI)
0.6
Business Digital Adoption (BDAI) 2014
0.5
-1.5 0.2
0.4
0 -0.5 -1 corr = 0.65*** -1.5 0.2
0.4
0.6
0.8
Business Digital Adoption (BDAI) 2016
Sources: World Bank World Development Report 2016 and Das and Hilgenstock (2018). Notes: the Business Digital Adoption Index (DAI) corresponds to the index in 2014 in the upper two panels and 2016 in the lower two panels. The Initial and Latest Routine Task Intensity (RTI) measures are from Das and Hilgenstock (2018). *** denotes significance at the 1 percent level.
Model Details
Total Output A perfectly-competitive firm aggregates total salaried-firm output Ys ,t and self-employment output Ye ,t according to
φy −1 φy
Yt = Ys,t
φy −1 φy
+ Ye,t
φyφy−1 , φy > 1
Denote by ps ,t the relative price of aggregate salaried output, and by pe ,t the relative price of total self-employment output Can show that the (normalized) aggregate price index is h i 1 1− φ 1−φ 1−φy 1 = ps ,t y + pe ,t y , Ys ,t = (ps ,t )−φy Yt , and Ye ,t = (pe ,t )−φy Yt Back
Salaried Production Unbounded number of salaried-firm entrants face sunk entry cost fe Immediately after entry, firms draw their idiosyncratic productivity a from common distribution G (a) Firms with a > ai ,t pay fixed cost fi and use inputs produced with ICT-labor (i) technology Assume fi = λf fe where 0 < λf < 1 Firms with a < ai ,t use inputs produced with regular (labor-only, r ) technology Measure of i and r firms: Ni ,t = [1 − G (ai ,t )] Nt , Nr ,t = G (ai ,t )Nt where Nt is measure of salaried firms
Salaried Production Individual-firm profits from producing using r and i technologies:
h mcr ,t i dr ,t (a) = ρr ,t (a) − yr ,t (a) a h mci,t i di,t (a) = ρi,t (a) − yi,t (a) − fi a where ρj ,t (a) is firm a’s real price, mcj ,t is the price of j intermediate goods, and yj ,t (a) is firm a’s output for j ∈ {r , i } Total profits for firm a: dt (a) = di ,t (a) + dr ,t (a) Demand function for firm a’s output yj ,t (a) for j ∈ {i, r }:
yj,t (a) = (ρj,t (a)/ps,t )−ε Ys,t
Salaried Production: Optimal Pricing and Tech. Choice Optimal real price for firm a using tech. j ∈ {i, r }:
ρj,t (a) =
ε mcj,t ε−1 a
Idiosyncratic productivity threshold ai ,t :
di,t (ai,t ) = dr ,t (ai,t ) Average idiosyncratic productivity levels:
Z ∞ ε−1 1 1 ε −1 aei,t = a dG (a) 1 − G (ai,t ) ai,t Z a ε−1 1 i,t 1 ε −1 aer ,t = a dG (a) G (ai,t ) amin
Salaried Production
Average salaried firm profits:
det =
Nr ,t Nt
der ,t +
Ni,t Nt
dei,t
where der ,t ≡ dr ,t (aer ,t ), dei ,t ≡ di ,t (aei ,t ), ρer ,t ≡ ρr ,t (aer ,t ), ρei ,t ≡ ρi ,t (aei ,t ), yer ,t ≡ yr ,t (aer ,t ) and yei ,t ≡ yi ,t (aei ,t )
Total salaried firm output: Ys ,t =
R
ζ ∈Z
ys , t ( ζ )
ε −1 ε
dζ
ε ε −1
, ε > 1 where
each firm produces differentiated output variety ζ, firm output ys ,t (ζ )
Intermediate Goods Choose vr ,t , vi ,t , nr ,t +1 and nii ,t +1 , fraction ωt of nr ,t allocated to production of i intermediate goods, and ITC capital ki ,t +1 to maximize E0 ∑t∞=0 Ξt |0 Πs ,t subject to i ,k ) Πs,t = [mcr ,t zr ,t H (nrr ,t ) − wrr,t nrr ,t − ψr vr ,t ] + [mci,t zi,t F (nri ,t , ni,t i,t i i i i −wi,t ni,t − wr ,t nr ,t − (ki,t +1 − (1 − δi )ki,t ) − ψi vi,t ]
nr ,t +1 = (1 − ρs ) [nr ,t + vr ,t q (θr ,t )] i i ni,t +1 = (1 − ρs ) ni,t + vi,t q ( θi,t ) nrr ,t = (1 − ωt )nr ,t nri ,t = ωt nr ,t
Intermediate Goods: Optimality Conditions
Standard ICT capital Euler equation
1 = Et Ξt +1|t [mci,t +1 zi,t +1 Fki ,t +1 + (1 − δi )]
Optimal decision over allocation of r workers across production of r and i intermediate goods, ωt
mci,t zi,t Fnri ,t − wri ,t = mcr ,t zr ,t Hnrr ,t − wrr,t
Intermediate Goods: Optimality Conditions
Standard job creation conditions
(1 − ωt +1 )[mcr ,t +1 zr ,t +1 Hnrr ,t +1 ψr r = (1 − ρs )Et Ξt +1|t −wr ,t +1 ] + ωt +1 [mci,t +1 zi,t +1 Fnri ,t +1 q (θr ,t ) ψ −wri ,t +1 ] + q (θr ,tr +1 ) and
ψi ψi i = (1 − ρs )Et Ξt +1|t mci,t +1 zi,t +1 Fni ,t +1 − wi,t + +1 i q (θi,t ) q (θi,t +1 )
Households and Self-Employment Choose ct , searchers se ,t , sr ,t , and si ,t , and desired empl. ne ,t +1 , nr ,t +1 , and nii ,t +1 , and Ne ,t and Nt +1 to maximize
E0
∞
∑ βt [u(ct ) − h(lfpe,t , lfpi,t , lfpr ,t )] s.t.
t =0
i ni + w r nr + w i ni + d et Nt ct + fe Ne,t + fi Ni,t = wi,t r ,t r ,t r ,t r ,t i,t +pe,t ze,t ne,t + Πs,t + Πy ,t
nj,t +1 = (1 − ρs ) [nj,t + sj,t f (θj,t )] ne,t +1 = (1 − ρe ) [ne,t + se,t φe ] Nt +1 = (1 − δ) [Nt + Ne,t ] for j ∈ {i, r }, where nrr ,t = (1 − ωt )nr ,t and nri ,t = ωt nr ,t and LFP lfpe ,t = ne ,t + se ,t , lfpi ,t = nii ,t + si ,t , and lfpr ,t = nr ,t + sr ,t
Household Optimality Conditions Salaried-firm creation condition:
h i fe = (1 − δ)Et Ξt +1|t det +1 + fe Participation decisions: hlfpr ,t 1 u0 (ct ) f (θr ,t )
= (1 − ρs )Et Ξt +1|t [wrr,t +1 (1 − ωt +1 ) + wri ,t +1 ωt +1 ] h lfp +(1 − ρs )Et Ξt +1|t f (θr1,t +1 ) − 1 u0 (crt,t++11)
hlfpi,t hlfpi,t +1 1 1 i = (1 − ρs )Et Ξt +1|t wi,t +1 + −1 u0 (ct ) f (θi,t ) f (θi,t +1 ) u 0 ( ct + 1 ) hlfpe,t 1 hlfpe,t +1 1 = (1 − ρe )Et Ξt +1|t pe,t +1 ze,t +1 + −1 u0 (ct ) φe φe u 0 ( ct + 1 )
Nash Wages and Unemployment Bilateral Nash bargaining between firms and salaried workers:
wrr,t wri ,t i wi,t
= ν mcr ,t zr ,t H
nrr ,t
h0 (lfpr ,t ) u0 (ct )
+ (1 − ν )
h0 (lfpr ,t ) u0 (ct )
h0 (lfpi,t ) u0 (ct )
= ν mci,t zi,t Fnri ,t + (1 − ν) = ν mci,t zi,t Fni
i,t
+ (1 − ν )
where 0 < ν < 1 is the bargaining power of workers
Total LFP is lfpt = lfpe ,t + lfpi ,t + lfpr ,t so that the unemployment rate is
urt ≡ (se,t + si,t + sr ,t )/lfpt
Market Clearing
Market clearing for each category of salaried output:
yer ,t = Nr ,t aer ,t yei,t i zi,t F (nri ,t , ni,t , ki,t ) = Ni,t aei,t zr ,t H (nrr ,t )
The resource constraint of the economy is
Yt = ct + (ki,t +1 − (1 − δi )ki,t ) + ψr vr ,t + ψi vi,t + fe Ne,t + fi Ni,t
Matching Process Details Matching function for employment category j ∈ {r , i }:
m (sj,t , vj,t ) = sj,t vj,t /(sj,t + vj,t )1/ξ , ξ
ξ
where ξ > 0, sj ,t are searchers in employment category j, and vj ,t are vacancies in that same category
Then, the job-finding and job-filling probabilities are defined as
f (θj,t ) = vj,t /(sj,t + vj,t )1/ξ ξ
ξ
q (θj,t ) = sj,t /(sj,t + vj,t )1/ξ ξ
where market tightness is θj ,t ≡ vj ,t /sj ,t
ξ
Functional Forms c 1−σc
Utility over consumption: u(ct ) = 1t−σc Disutility from LFP:
1
"
(κe (lfpe,t ) + κi (lfpi,t ) + κr (lfpr ,t ))1+ χ h(lfpe,t , lfpi,t , lfpr ,t ) = 1 + χ1 "
φy −1 φy
Total Output: Yt = Ys ,t
φy −1 φy
#
+ Ye ,t
#
φy φy −1
where Ye ,t = zt ne ,t
Production by i firms:
h iλ /λ 1/λi λi λk i λk i k F (., ., .) = (1 − φi ) nri ,t + φi αk ki,t + (1 − αk )(ni,t ) where 0 < φi , αk < 1 and λi , λk < 1 Back
Calibration Details I Parameters from literature: σc = 2, β = 0.985, δi = 0.025, ε = 6, δ = 0.025, kp = 6.5, ν = 0.5, amin = 1, ze = zr = 1. ρe = 0.03, ρs = 0.05, φy = 5, λk = 0.3, λi = 0.9, φi = 0.47, and χ = 0.26
I Calibrated parameters: αk = 0.1353, ξ = 0.3611, κe = 1.2862, κi = 2.7537, κr = 1.8358, φe = 0.50, ψ = 0.0538, fi = 0.0069, zi = 2.0362, and λf = 0.00392 Back
Additional Quantitative Results
Economic Mechanisms: Benchmark Model 0.2
ni/LFP
0.3
i
0.06
0.4
0.25 0.4
nr/LFP
0.08
nrr/LFP
ne /LFP
0.1 0.6
0.2
0.04 0.4
0.6
0.1 0.2
0.8
0.4
0.8
0.2
0.4
0.6
0.8
Ni/N 1
0.07
0.6
0.8
0.8 0.6 0.2
2
0.4 0.2
0.4
Ni/N
0.8
0.2
0.54
0.48
0.52 0.6
4.5 4 5 0.4
0.8
0.2
3.5 0.4
0.03 0.02 0.8
N
80
Ni
0.01 0.005
60 40 20 0.2
0.4
0.6
Ni/N
0.6
0.8
Ni/N
100
sr/LFP
0.04
Ni/N
0.6
Ni/N 0.05
0.6
6
1.5
Ni/N
0.4
5
1.6
0.2
0.02
0.8
1.4
0.8
0.015
0.6
7
Ji
Ave. i Prod.
i
0.5
p( r)
0.52
0.4
0.4
Ni/N
1.7
0.56
0.2
0.6
Ni/N
0.58
0.2
vr
4
0.4
Jrr
0.4
0.6
6
vi
0.08 0.075
0.2
p( )
0.6
Ni/N ICT Capital
Unempl. Rate
Ni/N
si/LFP
0.1 0.05
0.2 0.2
0.15
0.8
120 100 80 60 40 20 0.2
0.4
0.6
Ni/N
0.8
Delinking Digital Adoption Costs from fe 0.14
0.016
0.01
0.78
ni/LFP
0.785
0.12
i
r
0.02
nr/LFP
0.03
0.79
nr/LFP
ne /LFP
0.795
0.014 0.012
0.1 0.01
0.2
0.4
0.6
0.8
0.2
0.4
N /N
0.6
0.8
0.2
0.8
i
0.055 0.9
0.0654 0.0652
0.26 0.05
0.4
0.6
0.8
0.045
0.25
0.7
0.245
0.04 0.6
0.2
0.255
0.8
0.2
0.24
0.035 0.4
Ni/N
0.6
0.8
0.2
0.4
Ni/N
0.6
0.8
Ni/N
1.7
0.4545
0.54
0.454
0.5395
0.4535
0.539
3.48 6.46
1.6
6.44
Ji
0.5405
p( r)
0.455
Ave. i Prod.
0.541
3.46 6.42
1.5 6.4 3.44 6.38
1.4
0.453
0.2
0.4
0.6
0.8
0.2
0.4
Ni/N 10
0.6
0.8
0.2
0.4
Ni/N
0.6
0.8
Ni/N
-3
2.5 0.016
0.015
2
2.6
N
Ni
0.0155
1.2
sr/LFP
si/LFP
2.7 1.4
1.5
0.4
0.6
N /N i
0.8
0.5 0.2
2.5 2.4
1
1 0.2
vr
0.0656
2.3 0.4
0.6
N /N i
0.8
0.2
0.4
0.6
N /N i
0.8
Jrr
0.0658
vi
ICT Capital
Unempl. Rate
0.6
N /N
i
0.066
p( i)
0.4
N /N
i
Linking fi to Vacancy Costs Instead of fe Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Exogenous LFP: Data vs. Model Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Benchmark Model with Higher Baseline fe Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Benchmark Model with Resource Search Costs Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Benchmark Model: Linear and Log Trends in Data Cost of Starting a Business (% of Income Per Capita)
100 80 60 40 20 0 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.7
0.8
0.7
0.8
Digital Adoption by Firms
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 0.2
0.3
0.4
0.5
0.6
Digital Adoption by Firms Data
Model Prediction
Data: Linear Trend
Data: Log Trend
Delinking Digital Adoption from fe : Changes in fe Digital Adoption by Firms
1 0.8 0.6 0.4 0.2 0 10
20
30
40
50
60
70
80
90
100
110
80
90
100
110
80
90
100
110
Cost of Starting a Business (% of Income Per Capita)
Self-Employment Rate
1 0.8 0.6 0.4 0.2 0 10
20
30
40
50
60
70
Cost of Starting a Business (% of Income Per Capita)
Unemployment Rate
0.3 0.25 0.2 0.15 0.1 0.05 0 10
20
30
40
50
60
70
Cost of Starting a Business (% of Income Per Capita) Data
Model Prediction
Data: Linear Trend