backward vs. forward integration of firms in gvcs

Peter H. Egger Katharina Erhardt Gerard Masllorens
30/10/2020
CEPR Bwd vs. Fwd Integration of Firms in GVCs 30/10/2020 1 / 36
Overview
2 Model
3 Data
4 Results
Motivation
Production, trade and investment are increasingly organised within so-called global value chains (GVCs).
Different stages of the production process are located across national borders.
Organizational structure of global production networks is of crucial importance.
Any given firm can be integrated in the value chain in three forms:
Backward integration: e.g., a manufacturing firm integrating a commodity producer. Forward integration: e.g., a manufacturing firm integrating a wholesale firm. No integration: input demand and supply at arm’s length.
Motivation
The literature on topic has focused on the decision between no integration and backward integration.
Nevertheless both data and anecdotal evidence suggest that forward integration is very prevalent.
In ORBIS dataset covering 2007-2013 we see:
Unique shareholders: 571,636 Unique subsidiaries: 999,531 Firm-to-firm links in average year: 12,229,737 Backward integration for 52% of links (subsidiary is in a top-5 supplying country-sector). Forward integration for 52% of links (subsidiary is in a top-5 buying country-sector). Clearly there is some but not a complete overlap between top5- buying and supplying.
Literature
Seminal property rights framework for the boundaries of the firm by Grossman and Hart (1986).
Focus on the importance of ownership rights as a source of power when contracts are incomplete. Core result: Residual rights of control should be assigned to the party whose investment contributes most to the value of the final output.
Vertical organization of the firm (Nunn and Tre er, 2008; Alfaro and Charlton, 2009; Alfaro et al., 2016)
Organization of value chains within and across national borders (McLaren, 2000; Antras and Helpman, 2004; Antras and Chor, 2013; Alfaro et al., 2019)
Literature
The most recent advances focus mainly on backward integration:
Models of global sourcing (Grossman and Helpman, 2003; Antras and Helpman, 2004).
Organizational structure of global production networks (Antras, 2005).
Sequential multi-stage production networks (Antras and Chor, 2013; Alfaro et al., 2019)
Contribution
We build on model of forward and backward integration (Acemoglu et al., 2010).
Extending the model to include fixed integration costs.
Empirically assessing hypotheses regarding both integration directions.
Providing evidence for the property rights framework approach on forward integration.
We run our empirical analysis on a panel covering the country-sector pairs of the entire world.
Model setup
Backward integration (B) Forward integration (F) Independence (I)
Model setup
o S ) = xS (peo
P + 1) (1)
xS ∈ (0, 1) is a customized input.
indicates to what extent the final output relies on the provision of the customized input; (1− ) is importance of standard input.
eo P ≥ 0 and eo
S ≥ 0 are the (endogenous) investment levels of P and S , respectively.
p and s are the respective marginal products of investment.
Timing
1 P offers organizational form o ∈ {F , I ,B} and transfers, T o
P + T o S = 0.
2 S decides whether to accept the offer.
3 S and P simultaneously decide on their investment levels eo
P ≥ 0, eo S ≥ 0.
4 After investments are realized, S and P bargain over shared revenues.
Outside options
In order to determine the outcome of the Nash bargaining, we have to define the respective outside options V o
i in case of disagreement for player i under organizational form o, where {λS , λP} are retained investment fractions and θ is the outside-marketability share of a customized input in case of disagreement:
Producer Supplier
F S )
No Integration Y (xS = 0, e I P , 0) θ(se I
S + 1) Backward Y (xS = 1, eB
P , (1− λS )eB S ) 0
Profits
We can then compute the gross revenue accruing to each party under each organizational form,
yo i (eo
i + 1
P , e o S )− V o
S − V o P ) (2)
which, in turn, we can use to calculate profits,
πo i = yo
i + T o i (3)
where F o i denotes fixed costs of integration and Ci (e
o i ) is a simple
quadratic form for the costs of investments. Optimal investment
Total surplus
Equilibrium organizational form maximizes total surplus So = πo S + πo
P at optimal investment levels and noting that T o
P + T o S = 0:
SF = 1
2 s2 +
2
) s2 − F (6)
Total surplus
Which organizational form will be chosen depends on the joint surplus. Let:
F = SF − S I (7)
B = SB − S I (8)
γ = p
s (9)
Graphical representation
Forward integration Independence Backward integration
Predictions I
∂γ ≥ 0
A relatively higher investment intensity of the producer vis-a-vis the supplier induces backwards integration (γ > γB∗)
A relatively higher investment intensity of the supplier vis-a-vis the producer induces forward integration (γ < γF∗)
Corollary: The shareholder has relatively higher investment intensity than the subsidiary.
Predictions II
∂ < 0.
Marketability of the customized input (θ). Larger θ ⇒ Higher outside option for S ⇒ Larger [γF∗, γB∗] ⇒ any form of integration becomes less likely.
Dependence on the input for final production (). Larger ⇒ Lower outside option for P ⇒ Smaller [γF∗, γB∗] ⇒ any form of integration becomes more likely.
Predictions III
∂F > 0.
Fixed costs of integration (F ). Larger F ⇒ Integration more costly ⇒ Larger [γF∗, γB∗] ⇒ any form of integration becomes less likely.
Predictions IV
∂F∂θ < 0.
The cross-derivative of fixed integration costs and input marketability on {γB∗, γF∗} indicates that an increase in the marketability of inputs reduces the detrimental effect of fixed integration costs on backward integration but increases it on forward integration.
ORBIS data on ownership
annual firm data for 2007-2013.
J = 199 countries (i ,j).
S = 38 ISIC Revision-4 one-digit (two-digit for manufacturing) sectors (r ,s).
1992 · 382 = 57, 183, 844 country-sector-pair cells of potential ownership links per year.
For the period 2007-2013 this yields a panel data-set of 400,286,908 observations.
ORBIS data on ownership
CF rs ij as the number of firms in country i and sector r that are owned by
firms from sector s in country j (number of connected firms).
Figure: Number of Connected Firms
World Input-Output Table
Figure: World Input-Output Table - Antras and Chor (2017)
Input-output coefficients
ij,t
): normalized inputs of sector-r
output (regardless of its geographic origin) as used by country j in its production of sector-s output in year t.
Output coefficient (brs i ,t =
∑J j=1 Z rs
ij,t
): which sectors (regardless of
the country) are the main users for country i ’s sector-r output at year t.
(ars j , b
rs i ) are the average coefficient over the years in the sample
Backward and Forward variable
0 otherwise.
Indicates whether sector r of the affiliates is among the top-5 supplying sectors of shareholders in j and s.
Forwardrs j =
0 otherwise.
Indicates whether sector r of the affiliates is among the top-5 using sectors of shareholders in j and s.
Investment intensity (γ)
We proxy γ with R&D intensity (expenditures on R&D over total sales of a firm).
We calculate γs for every shareholder sector and γs for every affiliate sector.
Finally we define the following indicator on whether the shareholder has relatively higher investment intensity than the subsidiary
γrs =
{ 1 if γs ≥ γr , 0 otherwise.
According to the corollary of Prediction 1 the coefficient on γrs should be positive for the forward and backward integration.
Investment intensity (γ). Assessing Prediction 1
Table: R&D Investment Intensity
Number of Firm-to-Firm Connections (CF rs ij ,t) (1) (2) (3)
Rel. high shareholder R&D intensity (γrs t ) 0.689∗∗∗ 0.745∗∗∗ 0.541∗∗∗
(0.065) (0.069) (0.067) Backwardrs
(0.075) (0.079) Forwardrs
j 0.327∗∗∗ 0.339∗∗∗
(0.051) (0.057) Forwardrs
(0.056) (0.068)
Country-pair FE X X X Shareholder-country-industry-year FE X X X Subsidiary-country-industry-year FE X X X Domestic-year FE X X X
Obs. 28,484,832 28,484,832 28,484,832 R2 0.92838 0.92813 0.93018
Standard errors are clustered at country-industry pairs level and reported in parentheses.
* p < 0.1, ** p < 0.05, *** p < 0.01
Competition (θ)
We proxy θ as the ratio of the total number of firms in downstream sector over the total number of firms in upstream.
θrs ij ,t =
for forward integration.
According to Prediction 2 the coefficient on θrs ij ,t should be negative for the
forward and backward integration.
Competition (θ). Assessing Prediction 2
Table: Competition Effects
Market thickness of shareholder industry rel. to subsidiary industry (θB rs ij ,t) −0.000023∗∗∗ −0.000011∗∗∗
(0.000) (0.000) Backwardrs
j 0.915836∗∗∗ 0.765719∗∗∗
(0.000) (0.000)
Market thickness of subsidiary industry rel. to shareholder industry (θF rs ij ,t) 0.000012∗∗∗ 0.000015∗∗∗
(0.000) (0.000) Forwardrs
j 0.802499∗∗∗ 0.624699∗∗∗
Obs. 28,600,089 28,600,089 28,600,089 R2 0.92432 0.92314 0.92713
Column (3) also includes Output coef. × θB and Input coef. × θF as controls.
* p < 0.1, ** p < 0.05, *** p < 0.01
Input-consumption effects ()
We proxy as the share of total input consumption over production
sB j ,t =
R∑ r=1
ars j ,t
rF i ,t =
S∑ s=1
asr i ,t
According to Prediction 2 the coefficient on s j ,t should be positive for the
Input-consumption effects (). Assessing Prediction 2
Table: Total Input-consumption Effects
Backwardrs j −0.031 −1.109∗∗∗
Backwardrs j × Rel. importance of inputs for shareholder (B s
j ,t) 1.985∗∗∗ 1.636∗∗∗
Forwardrs j × Rel. importance of inputs for subsidiary (F r
i ,t) 1.933∗∗∗ 0.941∗∗∗
Obs. 28,576,343 28,560,530 28,536,807 R2 0.92498 0.92368 0.92921
Column (3) also includes Output coef. × B and Input coef. × F as controls.
* p < 0.1, ** p < 0.05, *** p < 0.01
Fixed integration costs (F )
We proxy a reduction of Fixed Integration Costs (F−1) as Bilateral Investment Treaty (BIT) coming into force.
F−1 ij ,t = BITij ,t =
{ 1 if a BIT is in force between i and j at year t, 0 otherwise.
According to Prediction 3 the coefficient on F−1 ij ,t should be positive for the
Fixed integration costs (F ). Assessing Prediction 3
Table: Fixed-cost Effects
BIT (F−1 ij ,t ) −0.036 −0.005 −0.053
(0.030) (0.031) (0.034) Backwardrs
Obs. 28,600,089 28,600,089 28,600,089 R2 0.92433 0.92314 0.92711
* p < 0.1, ** p < 0.05, *** p < 0.01
Cross Effects of Key Parameters. Assessing Prediction 4
Table: Competition and Fixed-cost Effects
BIT (F−1 ij ,t ) −0.050309 −0.022309 −0.074162∗∗
(0.031) (0.032) (0.034) Rel. high shareholder R&D intensity (γrs
t ) 0.688817∗∗∗ 0.743656∗∗∗ 0.539057∗∗∗
(0.064) (0.069) (0.067)
Market thickness of shareholder industry rel. to subsidiary industry (θB rs ij ,t) −0.000013∗∗ −0.000006
(0.000) (0.000) Backwardrs
j 0.324891∗∗∗ 0.351065∗∗∗
(0.061) (0.064) Backwardrs
(0.075) (0.079)
(0.000) (0.000)
Market thickness of subsidiary industry rel. to shareholder industry (θF rs ij ,t) 0.000013∗∗∗ 0.000012∗∗∗
(0.000) (0.000) Forwardrs
j 0.320228∗∗∗ 0.334782∗∗∗
(0.053) (0.058) Forwardrs
(0.056) (0.068)
(0.000) (0.000)
Obs. 28,484,832 28,484,832 28,484,832 R2 0.92844 0.92818 0.93028
* p < 0.1, ** p < 0.05, *** p < 0.01
Robustness
Different definitions of Forward/Backward creating TopH Inputrs j
for Backward and TopH Outputrs j for Forward with H measuring
whether a sector is among the H most-important ones with H ∈ {1, ..., 10} Different measure of investment intensity (γ) using physical investment intensity (expenditures on physical investment over total sales).
Subsamples of the data keeping in the sample only those countries that are reported in the WIOT.
Thank You!
Acknowledgement: This project has received funding from the European Union’s
Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie
agreement No 721916.
Investment level
Using (3) we can compute the optimal level of investment under each organizational form taking the other party’s investment as given:
eF∗ S = s, eF∗
back
Motivation and literature

backward vs. forward integration of firms in gvcs

Documents