Cost leadership at home and abroad in international banking Rients - - PowerPoint PPT Presentation

Literature and contribution Theory and empirics Results

Cost leadership at home and abroad in international banking

Rients Galema1, Caroline Liesegang2 and Michael Koetter3

1Utrecht University 2Deutsche Bundesbank 3Frankfurt School of Management

IMF/DNB - International Banking: Microfoundations and Macroeconomic Implications, 2014

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

New international trade theory

Very few firms in international goods trade

(Bernard and Jensen 1995; 1999)

Theory emphasizes firm-level productivity

Melitz (2003): only most productive fims internationalize Bernard et al. (2003): comparative cost advantages are key

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

New international trade theory

Very few firms in international goods trade

(Bernard and Jensen 1995; 1999)

Theory emphasizes firm-level productivity

Melitz (2003): only most productive fims internationalize Bernard et al. (2003): comparative cost advantages are key

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

New international trade theory

Very few firms in international goods trade

(Bernard and Jensen 1995; 1999)

Theory emphasizes firm-level productivity

Melitz (2003): only most productive fims internationalize Bernard et al. (2003): comparative cost advantages are key

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

New international trade theory

Very few firms in international goods trade

(Bernard and Jensen 1995; 1999)

Theory emphasizes firm-level productivity

Melitz (2003): only most productive fims internationalize Bernard et al. (2003): comparative cost advantages are key

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Can we apply goods trade models to (financial) services trade?

Buch et al. (JIE, 2011) show

Very few banks have foreign affilates Those that do are larger and more productive Almost all banks hold some cross-border assets

Breinlich and Criscuolo (2011)

find services exporters to be larger and more productive conclude that: These models [Melitz, 2003; Bernard et al., 2003] would seem, therefore, to provide a good starting point for explaining the basic characteristics of services exporters."

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Contributions to the literature

New empirical evidence on affiliate lending informed by trade theory Relate banks’ comparative costs to internationalization combining proprietary data with microdata Germany and foreign markets

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Service trade Contribution

Contributions to the literature

New empirical evidence on affiliate lending informed by trade theory Relate banks’ comparative costs to internationalization combining proprietary data with microdata Germany and foreign markets

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What do we do?

Take a model comparable to Bernard et al. (2003) and De Blas and Russ (2012) Models of comparative cost advantage Our basic assumptions

Banks have different levels of productivity.. ... and therefore different marginal costs operating abroad Cost difference used to overcome fixed costs of exporting

Next steps:

Define a profit function of banks’ affiliate lending. Banks are present abroad if profit (π) > 0 Derive an empirical specification (probit)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

When do banks conduct foreign affiliate lending?

When π > 0!

Consider profit bank i of lending to country j πij = RL

ij Lij − (RD i + ηj)(1/γ)Di − RD ij (1/ψ)Dij − Fj

Foreign lending financed by affiliate and parent deposits (1/γ)Di = αijLij (1/ψ)Dij = (1 − αij)Lij Substituting.. πij = RL

ij Lij − (RD i + ηj)αijLij − RD ij (1 − αij)Lij − Fj

(1)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

When do banks conduct foreign affiliate lending?

When π > 0!

Consider profit bank i of lending to country j πij = RL

ij Lij − (RD i + ηj)(1/γ)Di − RD ij (1/ψ)Dij − Fj

Foreign lending financed by affiliate and parent deposits (1/γ)Di = αijLij (1/ψ)Dij = (1 − αij)Lij Substituting.. πij = RL

ij Lij − (RD i + ηj)αijLij − RD ij (1 − αij)Lij − Fj

(1)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

When do banks conduct foreign affiliate lending?

When π > 0!

Consider profit bank i of lending to country j πij = RL

ij Lij − (RD i + ηj)(1/γ)Di − RD ij (1/ψ)Dij − Fj

Foreign lending financed by affiliate and parent deposits (1/γ)Di = αijLij (1/ψ)Dij = (1 − αij)Lij Substituting.. πij = RL

ij Lij − (RD i + ηj)αijLij − RD ij (1 − αij)Lij − Fj

(1)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What are the sources of profit?

The price and the cost markup.

Rewriting Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Price and cost margin Price markup: Intermediation margin (net interest margin) earned in the (imperfectly competitive) foreign market. Funding cost markup: Intermediation margin earned on fraction αij of lending financed with parent deposits.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What are the sources of profit?

The price and the cost markup.

Rewriting Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Price and cost margin Price markup: Intermediation margin (net interest margin) earned in the (imperfectly competitive) foreign market. Funding cost markup: Intermediation margin earned on fraction αij of lending financed with parent deposits.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What are the sources of profit?

The price and the cost markup.

Rewriting Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Price and cost margin Price markup: Intermediation margin (net interest margin) earned in the (imperfectly competitive) foreign market. Funding cost markup: Intermediation margin earned on fraction αij of lending financed with parent deposits.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

What are the sources of profit?

The price and the cost markup.

Rewriting Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Price and cost margin Price markup: Intermediation margin (net interest margin) earned in the (imperfectly competitive) foreign market. Funding cost markup: Intermediation margin earned on fraction αij of lending financed with parent deposits.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Why would parent banks have a funding cost advantage? Main hypothesis: parent banks have a funding cost advantage vis-à-vis foreign incumbants, e.g. due to: Scale economies in global wholesale funding. Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Accordingly, we define: C2j/Cij ≡ RD

ij − (RD i − ηj)

Substituting we arrive at: πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (C2j/Cij)

• Cost markup

−Fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Why would parent banks have a funding cost advantage? Main hypothesis: parent banks have a funding cost advantage vis-à-vis foreign incumbants, e.g. due to: Scale economies in global wholesale funding. Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Accordingly, we define: C2j/Cij ≡ RD

ij − (RD i − ηj)

Substituting we arrive at: πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (C2j/Cij)

• Cost markup

−Fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Why would parent banks have a funding cost advantage? Main hypothesis: parent banks have a funding cost advantage vis-à-vis foreign incumbants, e.g. due to: Scale economies in global wholesale funding. Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Accordingly, we define: C2j/Cij ≡ RD

ij − (RD i − ηj)

Substituting we arrive at: πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (C2j/Cij)

• Cost markup

−Fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Why would parent banks have a funding cost advantage? Main hypothesis: parent banks have a funding cost advantage vis-à-vis foreign incumbants, e.g. due to: Scale economies in global wholesale funding. Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Accordingly, we define: C2j/Cij ≡ RD

ij − (RD i − ηj)

Substituting we arrive at: πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (C2j/Cij)

• Cost markup

−Fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Why would parent banks have a funding cost advantage? Main hypothesis: parent banks have a funding cost advantage vis-à-vis foreign incumbants, e.g. due to: Scale economies in global wholesale funding. Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Accordingly, we define: C2j/Cij ≡ RD

ij − (RD i − ηj)

Substituting we arrive at: πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (C2j/Cij)

• Cost markup

−Fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Towards an empirical specification

Country lending income equals aggregate capital income MPKj × Kj ≡ R∗

j × Lj → Lij ≡ βijLD j ≡

• MPKj × Kj

R∗

j

• To derive a log-linear specification, we also define:

MP

ij ≡ µij(C2j/Cij)

NIMj ≡ µij/αij Banks are likely to be present when πij > 0: πij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj(1/R∗

j ) − Fj > 0

(2)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Towards an empirical specification

Country lending income equals aggregate capital income MPKj × Kj ≡ R∗

j × Lj → Lij ≡ βijLD j ≡

• MPKj × Kj

R∗

j

• To derive a log-linear specification, we also define:

MP

ij ≡ µij(C2j/Cij)

NIMj ≡ µij/αij Banks are likely to be present when πij > 0: πij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj(1/R∗

j ) − Fj > 0

(2)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Towards an empirical specification

Country lending income equals aggregate capital income MPKj × Kj ≡ R∗

j × Lj → Lij ≡ βijLD j ≡

• MPKj × Kj

R∗

j

• To derive a log-linear specification, we also define:

MP

ij ≡ µij(C2j/Cij)

NIMj ≡ µij/αij Banks are likely to be present when πij > 0: πij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj(1/R∗

j ) − Fj > 0

(2)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Towards an empirical specification

Country lending income equals aggregate capital income MPKj × Kj ≡ R∗

j × Lj → Lij ≡ βijLD j ≡

• MPKj × Kj

R∗

j

• To derive a log-linear specification, we also define:

MP

ij ≡ µij(C2j/Cij)

NIMj ≡ µij/αij Banks are likely to be present when πij > 0: πij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj(1/R∗

j ) − Fj > 0

(2)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Towards an empirical specification

Country lending income equals aggregate capital income MPKj × Kj ≡ R∗

j × Lj → Lij ≡ βijLD j ≡

• MPKj × Kj

R∗

j

• To derive a log-linear specification, we also define:

MP

ij ≡ µij(C2j/Cij)

NIMj ≡ µij/αij Banks are likely to be present when πij > 0: πij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj(1/R∗

j ) − Fj > 0

(2)

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Deriving the probability of having foreign lending

Define Z > 1 if bank i is in country j Zij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj FjR∗

j

> 1 Probit is derived by taking the natural logarithm: zijt = γ0 + γ0nimjt + γ2c2jt − γ3cit − γ4αjt + γ5βijt + γ6r ∗

jt + γ7mpkjt + γ8kjt

− γ9fj − γ10φt + εijt

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Deriving the probability of having foreign lending

Define Z > 1 if bank i is in country j Zij = (1 + NIMj)(C2j/Cij)αijβijMPKjKj FjR∗

j

> 1 Probit is derived by taking the natural logarithm: zijt = γ0 + γ0nimjt + γ2c2jt − γ3cit − γ4αjt + γ5βijt + γ6r ∗

jt + γ7mpkjt + γ8kjt

− γ9fj − γ10φt + εijt

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Which signs do we expect?

Probit: ρijt = Pr(Tijt = 1|Observables) ρijt = Φ(γ0 + γ1nimjt + γ2c2jt − γ3cijt − γ4αijt + γ5βijt + γ6rjt − γ7mpkjt + γ8kjt − γ9fj − γ10φt) Probability of presence... ...positively related to the intermediation margin nimjt ...positively related to MC foreign cost leaders c2jt ...negatively related to MC German bank cijt ...positively related to comparative MCs: I(c2jt > cit) ...negatively related to foreign loan rates rjt (quantity effect) ...negatively related to fixed operating costs fj

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

How do we estimate marginal costs and prices?

Key starting premise German affiliates’ cost advantage in country j is due to the parent’s cost advantage. Estimation steps Sample of world banks from Bankscope Estimate cost and profit frontiers for all world banks Calculate marginal costs Use marginal costs German banks as cit c2jt: 5th percentile of marginal costs in each country rjt: 5th percentile of average revenues in each country

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Bank data

Bankscope data 2003-2010 Commercial, savings, cooperative and investment banks Drop entries with negative factor prices, outputs, costs, equity and total assets Winsorize factor prices and outputs at 1% Deflate all monetary volumes using CPI Cost and profit function estimates for 108,704 observations

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Foreign status data

External Position Report (Auslandsstatus data) from the Bundesbank End-of-year data loans and advances of German banks to foreign affiliates Exclude interbank lending; only enterprises, households and governments Data from 2003-2010 After merge with Bankscope data: 1590 German banks in 52 countries

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Foreign status data

External Position Report (Auslandsstatus data) from the Bundesbank End-of-year data loans and advances of German banks to foreign affiliates Exclude interbank lending; only enterprises, households and governments Data from 2003-2010 After merge with Bankscope data: 1590 German banks in 52 countries

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Foreign status data

External Position Report (Auslandsstatus data) from the Bundesbank End-of-year data loans and advances of German banks to foreign affiliates Exclude interbank lending; only enterprises, households and governments Data from 2003-2010 After merge with Bankscope data: 1590 German banks in 52 countries

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Foreign status data

External Position Report (Auslandsstatus data) from the Bundesbank End-of-year data loans and advances of German banks to foreign affiliates Exclude interbank lending; only enterprises, households and governments Data from 2003-2010 After merge with Bankscope data: 1590 German banks in 52 countries

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Foreign status data

External Position Report (Auslandsstatus data) from the Bundesbank End-of-year data loans and advances of German banks to foreign affiliates Exclude interbank lending; only enterprises, households and governments Data from 2003-2010 After merge with Bankscope data: 1590 German banks in 52 countries

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Macroeconomic data and other controls

Macro data mpkjt & kjt: PWT 8.0 data, Feenstra et al. (2013) German FDI: MiDI database Bundesbank Distance: CPII Capital regulation & Activity restrictions: Barth & Levine (2001) Other bank controls Return on equity Loan impairment charges Cost-to-income ratio Z-score

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Commercial banks most important

Commercials Savings Coops Total # banks 104 468 1018 1590 # banks abroad 24 13 6 43 # countries ≥ 1 affiliate 52 51 49 52 In million euros...

• Av. lending per bank

57.80 16.90 3.65 39.10

• Av. lending per country

17,60 16.50 17.70 17.20

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Theory Probit Marginal costs and prices Data

Summary of ijt sample

Not Abroad Abroad N = 394,780 N =3,610 mean std mean std Bank marginal costs (cit) 0.038 0.008 0.029 0.015 Marginal costs abroad (c2jt) 0.025 0.021 0.024 0.019 Cost leadership (I(c2jt > cit)) 0.179 0.384 0.366 0.482 Price of capital (rjt) 4.433 3.341 4.244 3.181 Return on equity (roe) 17.290 7.566 14.154 14.959 Z-score (zs) 33.323 19.136 16.014 11.245 Credit risk (cr) 0.893 0.745 0.635 0.950 Cost-to-income ratio (cit) 40.189 8.996 24.784 14.910 Loans and advances abroad 0.000 0.000 987.918 7800.031

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Comparative cost advantage matters (1)

Marginal effects, ’03-’10

Controls: None Individual Joint cit −0.0258∗∗∗ −0.0139∗∗∗ −0.0094∗∗∗ −0.0037∗∗∗ c2jt −0.0034∗∗∗ −0.0023∗ −0.0023∗ −0.0012 I(c2jt > cit) 0.0047∗∗∗ 0.0027∗∗∗ 0.0028∗∗∗ 0.0016∗∗∗ nimjt 0.0044 0.0030 0.0030 0.0038 rjt 0.0022 0.0017 0.0016 0.0009 k −0.0010 −0.0002 0.0001 0.0003 mpk 0.0017 0.0015 0.0015 0.0011 αjt 0.0012∗∗ 0.0008 0.0007 0.0005 βjt 0.0029 0.0026 0.0029 0.0024

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Comparative cost advantage matters (2) I

Marginal effects, ’03-’10

Controls: None Individual Joint roe −0.0168∗∗∗ −0.0275∗∗∗ z-score −0.0052∗∗∗ credit risk −0.0006∗∗ cost-income ratio −0.0198∗∗∗ −0.0213∗∗∗ fj Yes Yes Yes Yes φt Yes Yes Yes Yes bank group dummies No Yes Yes Yes N 398,390 398,390 398,390 397,547 R2 0.3254 0.3267 0.4289 0.4643

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

How to interpret the size of the effects?

Aren’t they very small?

No, because number of banks abroad is very small. The 5th percentile of fitted values is almost 0% The 95th percentile of fitted values is typically around 2%

Some quantitative effects:

cit ↓ 1% → Probability abroad ↑ 0.37%. c2jt > cit → Probability abroad ↑ 0.16%.

These effects are economically significant compared to:

German FDI ↑ 1% Probability abroad ↑ 0.16%. An uncondictional likelihood of foreign presence of 0.9%.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Fixed costs have their predicted signs (1)

Marginal effects ’03-’10

Fixed costs Branches Subsidiaries cit −0.0037∗∗∗ 0.0002 −0.0026∗∗∗ c2jt 0.0011 −0.0001 −0.0014 I(c2jt > cit) 0.0008 0.0005 0.0009∗ nimjt −0.0002 0.0028 0.0003 rjt −0.0013 −0.0001 0.0012 k 0.0007 −0.0006 0.0000 mpk −0.0014 −0.0002 0.0016 αjt 0.0003 0.0004 0.0002 βjt −0.0035∗∗∗ 0.0031 0.0013

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Fixed costs have their predicted signs (2)

Marginal effects ’03-’10

Fixed costs Branches Subsidiaries German FDI 0.0016∗∗∗ Distance 0.0001 Capital restrictions −0.0004 Activity restrictions −0.0030∗∗ Bank controls Yes Yes Yes fj No Yes Yes φj Yes Yes Yes bank group dummies Yes Yes Yes N 370,607 397,547 397,547 R2 0.5421 0.4373 0.5746

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Additional results

Country groups:

Cost leadership only significant for less-developed and non-Euro and non-OECD countries. Cost leadership signficantly negative for financial centers.

Volume of affiliate lending:

Amount lent depends mostly on country characteristics, like the net interest margin. It does not depend on bank characteristics like cost advantage, marginal cost or risk considerations.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

Conclusion

Our starting point:

Theory: foreign presence driven by a price and a funding cost markup. Hypothesis: funding cost markup driven by relative parent cost advantages.

How did we test this empirically?

Combine proprietary data on foreign lending with data on foreign competitors. Test whether cost advantages drive foreign presence.

What did we find?

Likelihood of German banks operating foreign affiliates depends on domestic and relative cost advantages. It’s not just costs that matter: riskier and larger banks more likely to operate affiliates abroad. Relative cost advantages especially relevant for less developed countries.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (1)

Funding cost advantage vis-à-vis foreign incumbants due to:

1

Scale economies in global wholesale funding.

2

Cost advantages from managing liquidity on a global scale (Cetorelli and Golberg, JF and JIE 2012). Alternatives to C2j/Cij ≡ RD

ij − (RD i − ηj)

Use RD

i =interest expenses / interest bearing liabilites to

proxy for

1 .

Use number of countries in which bank i has affiliaties to proxy for

2 .

Use monetary policy stance to reflect foreign funding rates RD

ij .

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (2)

Again consider Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Using data on the interbank capital market... ...we can more directly measure (1/γ)Di = αijLij as Fij. Running a system of equations to explain whether... ...Lij > 0 can be explained by the Price markup. ...Fij > 0 can be explained by the Funding cost markup.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (2)

Again consider Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Using data on the interbank capital market... ...we can more directly measure (1/γ)Di = αijLij as Fij. Running a system of equations to explain whether... ...Lij > 0 can be explained by the Price markup. ...Fij > 0 can be explained by the Funding cost markup.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (2)

Again consider Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Using data on the interbank capital market... ...we can more directly measure (1/γ)Di = αijLij as Fij. Running a system of equations to explain whether... ...Lij > 0 can be explained by the Price markup. ...Fij > 0 can be explained by the Funding cost markup.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (2)

Again consider Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Using data on the interbank capital market... ...we can more directly measure (1/γ)Di = αijLij as Fij. Running a system of equations to explain whether... ...Lij > 0 can be explained by the Price markup. ...Fij > 0 can be explained by the Funding cost markup.

Galema, Liesegang, Koetter Cost leadership

Literature and contribution Theory and empirics Results Main results Probit Main results Probit without country-fixed effects Additional results

What if it is not relative cost advantages?

Some recent plans (2)

Again consider Equation (1) πij = Lij (RL

ij − RD ij )

• Price markup

+αijLij (RD

ij − (RD i + ηj))

• Funding cost markup

−Fj Using data on the interbank capital market... ...we can more directly measure (1/γ)Di = αijLij as Fij. Running a system of equations to explain whether... ...Lij > 0 can be explained by the Price markup. ...Fij > 0 can be explained by the Funding cost markup.

Galema, Liesegang, Koetter Cost leadership