J. Finan. Intermediation xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

J. Finan. Intermediation j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j fi

Market discipline and conflicts of interest between banks and pension funds Adolfo Barajas ⇑, Mario Catalán International Monetary Fund, United States

a r t i c l e

i n f o

Article history: Received 17 April 2012 Available online xxxx Keywords: Conflicts of interest Institutional investors Pension funds Banks Market discipline Argentina

a b s t r a c t We study the behavior of private pension funds as large depositors in a banking system. Using panel data analysis, we examine whether, and if so how, pension funds influence market discipline in Argentina in the period 1998–2001. We find that the disciplining role of pension funds depends on whether or not banks are connected to the pension fund industry through ownership of pension fund management companies. We find evidence that pension funds exert market discipline on unconnected banks but not on connected ones. On balance, pension funds undermine market discipline in the banking system as a result of conflicts of interest. We conclude that regulations aimed at averting these conflicts can enhance market discipline. Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction In the last three decades, a surge in the frequency and intensity of banking crises has destabilized economies worldwide, motivating research aimed at explaining their causes. A common view holds that government guarantees of bank liabilities weaken depositors’ response to changes in bankspecific fundamentals (‘‘market discipline’’) and result in excessive risk taking which ineffective regulation and supervision are unable to tame.1 An important unanswered question is whether large depositors enhance market discipline. ⇑ Corresponding author. E-mail address: [email protected] (A. Barajas). For a database documenting the high frequency and intensity of systemic banking crises in the period 1970–2007, see Laeven and Valencia (2012). On the connection between government guarantees and market discipline, see for example Demirguc-Kunt and Huizinga (2004) and Demirguc-Kunt and Detragiache (2002). 1

http://dx.doi.org/10.1016/j.jfi.2014.04.002 1042-9573/Ó 2014 Elsevier Inc. All rights reserved.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

2

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Private pension funds are large depositors in many Latin American and Central and Eastern European countries which implemented pension reforms from pay-as-you-go to fully funded systems over the last three decades. In these countries, pension funds therefore could exert influence on market discipline. This paper studies the behavior of private pension funds as large depositors in Argentina in the period 1998–2001. Argentina provides an excellent case study, as it introduced a funded pension scheme in 1994 and suffered a dramatic banking panic episode 7 years later, in 2001. We define market discipline as a situation in which depositors withdraw (increase) deposits in response to increases (declines) in banks’ risks as measured by a deterioration (improvement) in bank fundamentals.2 In theory, pension funds can influence market discipline in banking systems through different channels. On the positive side, in contrast to individuals, pension funds are sophisticated, large, and long-term depositors. Pension funds’ advantage in monitoring implies that they could enhance market discipline, penalizing banks for fundamental weaknesses or excessive risk-taking by withdrawing deposits. On the negative side, pension fund behavior could be affected by conflicts of interest in relation to banks. Pension funds could favor connected banks, undermining instead of enhancing market discipline. In Argentina, banks’ ownership of Pension Fund Management Companies raises the possibility that conflicts could have influenced the deposit allocation of pension funds. For these reasons, the case of Argentina is of particular interest. We conduct panel data analysis over the period 1998–20013 to address the questions whether, and if so how, pension funds influence market discipline in a banking system. The first question that we address is, ‘‘do pension funds exert market discipline on banks?’’ We find that the disciplining role of pension funds depends on whether or not banks are connected to the pension fund industry through ownership of pension fund management companies. We find evidence that pension funds exert limited market discipline on unconnected banks but not on connected ones. Regarding unconnected banks, we obtain the following two results. First, pension funds exert discipline with respect to two CAMEL-type fundamentals: capital adequacy and the non-performing loans ratio—an asset quality indicator. We find no evidence that pension funds exert discipline with respect to changes in bank profitability, exposure to the government, liquidity, or the bank fundamental indicator z-score—a distance-to-insolvency measure. Second, the discipline exerted by pension funds strengthens as the share of pension fund deposits in a bank rises; this suggests that a larger presence of pension funds in a bank’s deposit base improves their disciplining incentives. Regarding connected banks, however, we find that the disciplining behavior of pension funds is tainted by conflicts of interest. Pension funds undermine overall market discipline by shifting deposits toward connected banks with weakening fundamentals. In sharp contrast to unconnected banks, connected banks gain pension fund deposits as their capitalization and z-score measures decline. The second question that we address is, ‘‘do pension funds enhance market discipline?’’ In other words, ‘‘does the presence of pension funds affect the disciplining behavior of other depositors?’’ We find evidence that other (non-pension fund) depositors exert discipline with respect to some CAMEL-type fundamentals—capital adequacy and exposure to the government4—and the z-score when the share of pension fund deposits is small. As the share of pension fund deposits rises, however, the discipline exerted by non-pension fund depositors vanishes—possibly due to a crowding out effect whereby a larger presence of pension funds in a bank’s deposit base reduces the incentives for other depositors to exert market discipline. 2 A broader definition would also include situations where depositors demand higher interest rates in response to a deterioration in fundamentals—see Berger (1991) and Martínez Pería and Schmukler (2001). Our goal is to compare the disciplining behavior of pension funds and other depositors, but we have no data discriminating the interest rates earned by type of depositor. 3 This study period is limited by data availability; although bank-specific as well as macroeconomic information is available for a longer time period, 1998–2001 is the longest period for which the deposit allocation of pension funds across individual banks is available. 4 Exposure to the government sector is a highly relevant bank fundamental in the case of Argentina. The study period includes a banking crisis and a protracted sovereign debt crisis. Thus, we interpret high provision of government financing by a bank as a fundamental weakness—a high exposure to sovereign default risk.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

3

Focusing on pension funds’ behavior during the crisis period, we confirm that their disciplining role is undermined by conflicts of interest. When the crisis hit Argentina’s banking system in January of 2001,5 pension funds held 5.7% of the system’s time deposits. A striking fact about the crisis is that pension funds increased their deposits sharply, despite a generalized panic. Specifically, deposits held by pension funds increased by 19.0% from January to November of 2001 while time deposits in the system declined by 21.2% in the same period. By the time deposit convertibility was suspended on December 1, 2001, the share of pension fund deposits in the system had increased to 8.5%. In part, this happened because of tight regulations that limited pension funds’ investment options—including foreign investment restrictions preventing international diversification.6 These regulations, however, applied only to broad asset categories such as ‘bank deposits,’ so pension funds were free to exert discipline by reallocating deposits across banks. The behavior of aggregate pension fund deposits during the crisis—increasing while other depositors were withdrawing funds—suggests that pension funds might have played a stabilizing role in the banking system. From a market discipline perspective, however, we focus on disaggregated data and show that pension funds conducted a massive reallocation of deposits from unconnected to connected banks. Thus, motivated by conflicts of interest, pension funds undermined market discipline during the crisis. Consistent with our econometric results, we find that, during the crisis, pension funds allocated deposits to banks with poor fundamentals. At the beginning of this episode, pension fund deposits were concentrated in a small number of banks (15 out of 86 banks received 90% of pension fund deposits) that were ‘‘weak’’ in terms of capitalization, liquidity, and exposure to the government. Interestingly, due to poor fundamentals, banks that attracted pension fund deposits before the crisis suffered larger deposit losses than did other banks during the crisis. These results point to an important policy implication: regulations aimed at averting conflicts arising from bank ownership of pension fund management companies can enhance market discipline. The rest of the paper is organized as follows. Section 2 discusses related literature. Section 3 describes the data. Section 4 describes the behavior of pension funds during the banking crisis. Section 5 describes the panel data methodology. Section 6 presents the panel data results. Section 7 concludes. 2. Related literature In relation to previous research, this is the first paper that studies the behavior of a type of institutional investor (pension funds) as a depositor in a banking system. It connects to several strands of literature. First, it is related to the literature on market discipline in the banking sector. Park and Peristiani (1998) develop the basic empirical framework to test whether depositors respond to higher bank risk by withdrawing deposits or demanding higher interest rates. They propose a two-equation procedure: the first equation estimates the probability of bank failure based on observable fundamentals, and the second equation estimates how depositors respond to the estimated probability. Applying this framework to US thrifts in the late 1980s, the authors find evidence of market discipline through both the deposit growth and interest rate channels. Other studies use a single equation approach, whereby depositor behavior is directly determined by the set of bank fundamentals.7 Martínez Pería and Schmukler (2001) find evidence of market discipline in Argentina, Chile, and Mexico during the 1980s and 1990s. Barajas and Steiner (2000) examine 5 We date the beginning of the Argentine banking crisis at the end of January of 2001, when time deposits in the banking system reached a peak. Note that all pension fund deposits were in the form of time deposits. 6 The regulations, in place since the inception of the funded pension system in 1994, stated that for the main asset categories, the maximum limits on pension fund portfolio holdings were as follows: foreign assets (10%), bank deposits (30%), and government debt (50%). These limits imply that no less than 10% of pension fund portfolios had to be allocated to domestic stocks and corporate bonds. 7 Martínez Pería and Schmukler (2001) point out two advantages of the single equation approach: (1) it permits a market discipline test in cases where a lack of actual bank failures preclude an estimation of the probability of failure, and (2) it allows one to study specifically which fundamentals are affecting depositor behavior the most.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

4

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Colombian banks during 1985–1999, and find that depositors prefer banks with higher capital and loanloss provisions, those that are state-owned, and banks that offer a wider branch network. Schumacher (2000) studies the deposit runs in Argentina in 1995 and finds that banks with weaker fundamentals suffered larger deposit withdrawals. Levy-Yeyati et al. (2010) examine depositor behavior in Argentina in the run-up to the 2001 crisis and find that depositor discipline remains significant once macroeconomic factors are accounted for.8 Barajas et al. (2007) also study the 2001 Argentine crisis, and find that depositors favored banks with higher capital and loan quality, and lower exposure to the government. More recently, Ioannidou and Penas (2010) study the case of Bolivia and show that the introduction of a deposit insurance system increased bank risk taking due to a decrease in market discipline exerted by large depositors. Some studies focus on market discipline in a cross-country setting. Demirguc-Kunt and Huizinga (2004) estimate panel regressions for 30 countries during 1990–1997 and find that explicit deposit insurance schemes weaken market discipline. Berger and Turk-Ariss (2010) examine depositor behavior in the US and EU during the 11-year period leading up to the global financial crisis and find evidence of stronger market discipline in the US and for smaller banks; also, depositors are more sensitive to the equity–asset ratio than to measures of loan portfolio performance. In some of the above studies, the behavior of small and insured depositors is contrasted with that of large, well-informed and uninsured depositors. Our contribution to the literature is to address the question whether private pension funds, with characteristics in common with the latter group but with particular connections to banks, can serve to enhance or undermine the disciplining function. The second strand to which this paper contributes is the incipient literature on the effects of financial crises on retirement systems. In light of the global crisis, Mitchell (2010) discusses risks and vulnerabilities associated with existing pension systems and calls for the development of a new framework for retirement security. Whitehouse (2009) documents the heavy losses inflicted by the crisis on private pension funds, and the policy responses adopted in OECD countries; and Munnell et al. (2008) discuss the effects of the crisis on public pension systems in the US. These studies focus on the current crisis; however, our knowledge about the impact of previous crises on funded pension systems remains anecdotal. In this paper, we show that a risk that has been ignored in the literature— originating in conflicts of interest between banks and pension funds—can undermine market discipline during a crisis. This paper also contributes to a third strand of literature, on pension fund governance, by documenting behavior of pension funds that is inconsistent with their fiduciary duty to act in the best interest of beneficiaries.9 Related literature studies conflicts of interest in financial institutions, including universal banks and mutual funds.10 We identify a new conflict—created because pension funds act as depositors in banks that own Pension Fund Management Companies—and show evidence supporting its existence. 3. Data description We exploit a unique dataset showing pension fund deposit allocation across banks in Argentina for the period January 1998–December 2001. Our empirical analysis combines two types of panel data: bank-specific balance sheet information and pension fund deposit holdings. All data are available at a monthly frequency. We obtained the banking data from the Central Bank of Argentina and the data on pension fund deposits and ownership from the pension funds regulator, Superintendencia de Administradoras de Fondos de Jubilaciones y Pensiones (SAFJP). 8 Calomiris and Powell (2001) provide a detailed account of regulatory developments in Argentina aimed at establishing discipline during the 1990s. For an account of macroeconomic developments in Argentina before and during the crisis, see Mussa (2002). 9 Central references in the literature on pension fund governance are Clark (2000, 2004). Catalán (2004) offers an early discussion of conflicts of interest between pension funds and banks in Latin America. 10 Conflicts of interest within universal banks include those associated with bank lending/underwriting; sell side analysis/ underwriting; analysis/brokerage; and asset management/underwriting; Mehran and Stulz (2007) offer an excellent literature review. Conflicts within the mutual fund industry are reviewed by Mahoney (2004). Ljungqvist et al. (2007) show that institutional investors moderate conflicts in sell-side research.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

5

3.1. Bank-specific data We follow previous literature and use the volume of time deposits along with variables that provide proxies of risk, including the z-score and CAMEL-type fundamental variables.11 The z-score (Z-SCORE) measures distance-to-insolvency: by how many standard deviations must the bank’s profitability fall in a given 6-month period in order to deplete the bank’s capital.12 The CAMEL-type fundamental variables include measures of capital adequacy, asset quality, profitability, and liquidity. Capital adequacy is measured by the capital to assets ratio (CAPR). Asset quality is measured by credit risk related variables: the ratio of non-performing loans to total loans (NPLL) and the ratio of net exposure to the government to total assets (NGOVB). Profitability is measured by the ratio of before-tax profits to total assets (PROFIT), and liquidity by the ratio of liquid assets to total assets (LIQ).13 We also include bank size, measured by each bank’s market share in terms of assets (SIZE). Note that all of these variables are standard and widely considered in the literature as possible determinants of deposit growth. In particular, banks of larger size, that have higher z-scores, higher capitalization, more liquidity, and higher-quality assets, are considered stronger, or equivalently, less risky. Also, in the dataset we identify the ownership of each bank, whether public, private, or foreign, and construct the dummy variables PRIVATE and FOREIGN. Finally, using data on ownership of Pension Fund Management Companies, we construct the dummy variable CONNECT that takes the value 1 if a bank is ‘‘connected’’ to the pension fund industry—owns a Pension Fund Management Company—and 0 otherwise.14 3.2. Pension fund deposit data We obtained the deposits held by pension funds as a whole in each individual bank on a monthly basis.15 Combining pension fund deposit data with bank-specific data, we constructed a series of the share of pension fund deposits in total time deposits of each bank (PFDEP). There were 18 pension funds at the beginning of our sample period in January of 1998. A process of consolidation reduced the number of pension funds to 13 by the end of 1999. The 13 pension funds in operation at the beginning of the banking crisis in January of 2001 held time deposits with fixed and variable interest rates and were almost entirely denominated in domestic currency.16 The duration (average maturity) of pension fund deposits was 6 months, and 12% of the deposits matured in December of 2001 or later. Table 1 reports summary statistics across banks and over the study period for the variables used in the analysis. 4. Pension funds’ behavior as depositors during the crisis This section describes pension fund deposit allocation and its relation to bank fundamentals at the beginning of the crisis, and the cross-bank deposit reallocations during the crisis. 11 Specifically, the selection of the CAMEL-type variables is based on previous studies of depositor discipline in Argentina, in particular Martínez Pería and Schmukler (2001) and Barajas et al. (2007). As suggested by one of the referees, we also include the z-score as a fundamental determinant of depositor discipline. 12 Consistent with the average maturity of pension fund deposits—taken as the period needed for pension pffiffiffi funds to exert discipline—the z-score is calculated for a 6-month horizon: Z  SCORE ¼ ðCAPR þ 6  PROFITÞ=ðrPROFIT  6Þ: The standard deviation of PROFIT (rPROFIT) is calculated on a rolling basis using monthly observations for a full year. 13 The numerator of NGOVB is calculated as bank holdings of government bills and bonds, plus loans to the non-financial public sector, net of government deposits denominated in both domestic and foreign currency. The numerator of PROFIT is calculated as monthly before-tax profits. The numerator of LIQ includes gold and cash assets and reserves (required and voluntary) denominated in domestic and foreign currency; it excludes government bills and bonds as well as private sector assets. 14 During the study period, banks owned 80% of the Pension Fund Management Companies; the remaining ownership was divided among insurance companies, labor unions, and non-financial companies. 15 It is important to note a limitation of our dataset: we do not have deposit holdings of each pension fund in each individual bank. Our dataset only has the aggregate deposit holdings of all pension funds as a group in each individual bank. For this reason, our analysis of conflicts of interest must be interpreted in relation to the connection of banks to the pension fund industry. 16 By January of 2001, pension fund deposits were distributed as follows: 86% were fixed-rate deposits denominated in pesos; 4.5% were fixed-rate deposits in foreign currency (2.6% in US dollars and 1.9% in euros); 6.5% were floating-rate deposits in pesos; and 3.5% were peso-denominated deposits embedding early withdrawal options.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

6

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Table 1 Summary statistics. Sources: Central Bank of Argentina; SAFJP; and authors’ calculations. Mean Deposit growth DDEP (l = 6)

SD

Min

Max

25%

50%

75%

0.53

9.93

1.00

383.32

0.10

0.01

0.10

48.7 0.16 0.07 0.0005 0.14 0.06

49.0 0.14 0.05 0.0073 0.16 0.10

2.6 0.01 0.00 0.1705 0.00 0.26

180.4 0.98 0.34 0.0356 0.92 0.70

14.2 0.08 0.03 0.0008 0.02 0.00

30.5 0.11 0.06 0.0003 0.09 0.04

65.2 0.18 0.10 0.0013 0.19 0.09

Size SIZE

0.011

0.022

0.000

0.136

0.001

0.003

0.009

Pension fund deposits PFDEP PFDEP (>0)

0.06 0.17

0.17 0.25

0.00 0.00

1.00 1.00

0.00 0.02

0.00 0.06

0.02 0.16

DUM CONNECT FOREIGN PRIVATE

0.13 0.40 0.89

Bank-specific fundamentals Z-SCOREa CAPR LIQ PROFIT NPLL NGOVB

We applied 90% winsorization to the z-score variable to reduce the influence of outliers. a This table reports the summary statistics for the z-score variable defined in reference to a 6 month horizon (footnote 12).

4.1. Deposit allocation of pension funds at the beginning of the crisis Table 2 shows basic facts that characterize the cross-bank deposit allocation of pension funds at the beginning of the crisis, in January of 2001. The following are noteworthy: (1) Pension funds held deposits in 29 out of 86 banks. They allocated 90% of their deposits to 15 banks and were ‘‘large depositors’’ in 9 banks—holding more than 10% of the banks’ deposits; (2) Deposits from pension funds accounted for 5.7% of the system’s deposits and 6.7% of the deposits of the 29 banks in which they invested. Within the group of 15 banks that attracted 90% of pension fund deposits, these accounted for 8.5% of total deposits. Within the group of 9 banks in which pension funds were large depositors, pension fund deposits represented 32.7% of total deposits;17 (3) Pension funds allocated their deposits to banks that were fundamentally strong in terms of the z-score and non-performing loans but ‘‘weak’’ in terms of capitalization, liquidity, profitability, and exposure to the government.

4.2. Behavior during the crisis Turning to the banking panic, Fig. 1 shows a striking fact: pension funds increased their deposits sharply, moving against the tide during this turbulent period. Specifically, deposits held by pension funds increased by 19.0% from January to November of 2001 while time deposits in the system declined by 21.2% in the same period. One reason why this happened is that tight regulations—including those restricting international diversification—limited pension funds’ investment options, although pension funds were free to reallocate deposits across banks. Fig. 2 shows relative frequency distributions of deposit changes across banks from January to November of 2001 for all banks in the system and for the group of ‘‘banks with large pension fund deposits’’—the 9 banks in which pension funds held more than 10% of total deposits in January of 17

These numbers correspond to the weighted average calculations shown in Table 2.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

7

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Table 2 Pension fund deposits and bank fundamentals at the beginning of the banking crisis, January of 2001 (in percent, except for numbers corresponding to Z-SCORE). Sources: Central Bank of Argentina; IMF, International Financial Statistics; SAFJP; and authors’ calculations. Banks with large pension fund deposits (9 banks) Pension fund deposits PFDEP Simple 52.6 average Deposit- 32.7 weighted average

Banks receiving 90% of pension fund deposits

All banks with pension fund deposits

Banks with no pension fund deposits

Banks not connected to the pension fund industry

Banks connected to the pension fund industry

All banks

(15 banks)

(29 banks)

(57 banks)

(75 banks)

(11 banks)

(86 banks)

32.1

19.0

–

7.4

6.4

7.3

8.5

6.7

–

7.3

3.9

5.7

43.0 20.1 7.0 0.03 24.9 4.3

46.7 17.9 6.6 0.01 19.7 6.0

62.3 8.8 7.0 0.02 15.4 6.9

48.7 16.2 6.7 0.02 18.9 5.3

0.4

0.6

5.3

1.3

Bank-specific fundamentals (simple averages) Z-SCORE 50.5 66.2 58.4 CAPR 15.1 9.6 10.9 LIQ 4.7 4.5 6.1 PROFIT 0.07 0.03 0.01 NPLL 10.5 10.1 10.7 NGOVB 10.4 9.4 7.9 Size SIZE

1.4

4.0

2.7

56.0

3.8

54.0

3.6 3.4 3.2

50.0

3.0

48.0

2.8

Banking System (left)

Nov-2001

Set-2001

Jul-2001

May-2001

Mar-2001

Jan-2001

Nov-2000

Set-2000

2.2 Jul-2000

2.4

42.0 May-2000

2.6

44.0 Mar-2000

46.0

Jan-2000

Billion $

52.0

Pension Funds (right)

Fig. 1. Time deposits of pension funds in the Argentine Banking System (January 2000–November 2001).

2001. Note that the distribution corresponding to this sub-group of banks is skewed to the right compared to that corresponding to all banks: pension funds were large depositors in banks that suffered larger than average deposit losses during the crisis. Table 3 confirms this observation: deposit changes in three groups of banks are respectively 21.2% (all banks), 25.7% (banks receiving 90% of pension funds’ deposits), and 24.7% (banks with large pension fund deposits). It also shows the breakdown of deposit changes across bank groups by depositor type. Regarding the behavior of pension funds as depositors, we highlight the following facts:

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

8

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Fig. 2. Histogram of change in deposits during the crisis all banks and banks with large pension fund deposits (January– November 2001).

Table 3 Bank deposit changes during the banking crisis by depositor type, January–November 2001a (in percent). Sources: Central Bank of Argentina ; SAFJP; and authors’ calculations.

All depositors Pension funds Other depositors a

Banks with large pension fund deposits (9 banks)

Banks receiving 90% of pension fund deposits (15 banks)

All banks with pension fund deposits (29 banks)

Banks with no pension fund deposits (57 banks)

Banks not connected to the pension fund industry (75 banks)

Banks connected to the pension fund industry (11 banks)

All banks

24.7 24.8

25.7 19.7

24.2 17.7

10.1 ...

27.0 0.5

17.3 43.6

21.2 19.0

24.6

26.2

27.3

10.1

29.1

19.8

23.4

(86 banks)

Numbers are calculated as deposit-weighted averages across banks.

(4) Pension funds increased their aggregate deposits sharply; (5) Pension funds conducted a large reallocation of deposits across banks, withdrawing 17.7% of their deposits from the 29 banks in which they held deposits at the beginning of the crisis, and reallocating them along with additional deposits toward a new group of banks. This reallocation generally occurred from unconnected to connected banks: pension funds withdrew 0.5% of their deposits from unconnected banks and increased their deposits in connected banks by 43.6%; (6) Pension funds’ behavior as depositors helps explain the fate of different bank groups. Banks that received pension fund deposits before the crisis lost on average 24.2% of deposits. In contrast, banks that had no pension fund deposits before the crisis lost 10.1% of deposits. Table 4 shows the deposit reallocation conducted by pension funds during the crisis. Banks are listed in decreasing order by the value of the (net) pension fund deposit changes observed in the period January–November 2001. The following facts emerge: (7) 15 banks received net inflows of pension fund deposits: 12 were foreign owned, 2 were publicly owned, and one was domestically and privately owned;

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Bank name

Pension funds

Banks Ownership type

Connection to pension fund industry

23.6 37.6

F F

Y

2.3

6.5

F

Y

6.8

4.3

5.0

F

Y

7,278

4.4

6.7

23.0

F

Y

36,558

5,389

17.3

0.6

14.3

F

25,353 3,084 22,801 29,552 4,757 12,448

27,175 4,678 21,917 26,864  6,791

1,823 1,595 884 2,689 4,757 5,657

7.2 51.7 3.9 9.1 100.0 45.4

0.7 1.4 1.0 0.2 1.3 7.3

31.0 7.8 59.9 50.7 19.3 19.5

F F F F F F

75,412 115,086

36,228 72,837

39,185 42,249

52.0 36.7

1.7 3.0

26.6 20.9

F F

161,859

103,684

58,175

35.9

0.7

39.2

F

205,751 79,248 187,597

147,354 – 90,586

58,397 79,248 97,011

28.4 100.0 51.7

8.6 1.5 0.6

22.3 11.0 54.2

F F F

141,407

9,987

131,420

92.9

0.5

85.8

F

Size, asset market share January 2001

Deposit change January– November 2001

(in percent)

(in percent of system)

(in percent)

127,256 66,677

52.5 59.6

6.2 1.5

139,799

48,127

52.5

396,166

422,938

26,772

166,533

173,811

31,169

Share of bank deposits January 2001

Deposits January 2001

Deposits November 2001

Deposit change January– November 2001

(in percent)

(thousands of pesos)

(thousands of pesos)

(thousands of pesos)

242,490 111,814

369,747 178,491

91,671

Banks with pension fund deposits in January 2001 Foreign banks CITIBANK N.A. 7.6 DEUTSCHE BANK 90.5 ARGENTINA SCOTIABANK QUILMES 6.6 S.A. HSBC BANK ARGENTINA 18.7 S.A. BANKBOSTON, NATIONAL 4.9 ASSOCIATION BANCO SOCIETE 5.9 GENERALE S.A. LLOYDS TSB BANK PLC. 6.2 DEL SUQUIA S.A. 0.3 ABN AMRO BANK N.V. 5.5 B.I.CREDITANSTALT S.A. 19.0 BANSUD S.A. 0.5 BBVA BANCO FRANCES 0.3 S.A. BISEL S.A. 6.2 NAZIONALE DEL LAVORO 5.6 S.A. EUROPEO PARA AMER. 99.7 LATINA RIO DE LA PLATA S.A. 4.2 CAJA DE AHORRO S.A. 6.4 MORGAN GUARANTY 98.7 TRUST CO. OF NY ING BANK N.V. 80.0

Y Y

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Y

(continued on next page) 9

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 4 Pension fund deposit flows across banks during the crisis, January–November 2001.a Sources: Central Bank of Argentina; SAFJP; and authors’ calculations.

10

Bank name

Domestic public banks DE LA CIUDAD DE BS. AS. DE LA PCIA. DE BS. AS. Domestic private banks CREDICOOP COOP. LTDO. COMAFI S.A. PATAGONIA S.A. GENERAL DE NEGOCIOS S.A. MACRO MISIONES S.A. SAENZ S.A. HIPOTECARIO S.A. DE GALICIA Y BUENOS AIRES Sub-total

Pension funds

Banks Deposits January 2001

Deposits November 2001

Deposit change January– November 2001

Size, asset market share January 2001

Deposit change January– November 2001

(in percent)

(thousands of pesos)

(thousands of pesos)

(thousands of pesos)

(in percent)

(in percent of system)

(in percent)

2.8 2.3

50,862 124,857

75,269 86,694

24,407 38,163

48.0 30.6

2.6 9.4

10.9 30.2

D-Pub D-Pub

0.2 1.3 1.1 10.0

2374 1,120 1,590 44,211

13,231 – – 41,563

10,857 1,120 1,590 2,649

457.4 100.0 100.0 6.0

1.5 0.2 0.3 1.1

12.6 50.0 36.9 52.3

D-Priv D-Priv D-Priv D-Priv

0.8 15.1 41.8 7.8

2,948 8,793 166,297 540,903

– 2,254 62,370 358,557

2,948 6,538 103,927 182,346

100.0 74.4 62.5 33.7

0.5 0.1 3.3 9.3

25.6 48.0 56.9 37.5

D-Priv D-Priv D-Priv D-Priv

3,048,155

2,509,381

538,774

17.7

78.6

-

+

F

0.0 0.8 0.1

7.5 + 82.5

F F F

11.1

4.2

D-Pub

Banks that attracted new pension fund deposits in January–November 2001 SUDAMERIS ARGENTINA – – 250,097 S.A. ITAU BUEN AYRE S.A. – – 24,069 BANK OF AMERICA NA – – 280,257 LINIERS SUDAMERICANO – – 1024 S.A. DE LA NACION – – 563,021 ARGENTINA Sub-total Total a

Ownership type

Share of bank deposits January 2001

–

1,118,468

1,118,468

+

12.0

3,048,155

3,627,849

579,694

19.0

90.7

Connection to pension fund industry

Y

Y

Y

Ownership of banks is indicated by F (foreign), D-pub (domestic public), and D-priv (domestic private). Y indicates that the bank has ownership stakes in a pension fund management company, as of April 2001.

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 4 (continued)

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

11

(8) Banks that received the largest inflows of pension fund deposits were connected to the pension fund industry—as owners of pension fund management companies; (9) The new group of banks that attracted pension fund deposits during the crisis (bottom panel, Table 4) consisted of only 5 banks: 4 were foreign-owned and one was the largest public bank; (10) There was a large reallocation of pension fund deposits within the group of foreign banks, generally from unconnected to connected banks. These facts provide preliminary support for the hypothesis that conflicts of interest hamper the disciplining role of pension funds. Note that although foreign banks are often considered more safe and sound than domestic banks in emerging markets, Fact 10 suggests that conflicts of interest dominated pension funds’ behavior as depositors during the crisis, trumping even flight-to-safety concerns. We now turn to the econometric analysis, which provides further evidence of the conflicts of interest hypothesis. 5. Econometric methodology We conduct two main econometric exercises to determine the nature of pension fund behavior as depositors. First, in Section 5.1, we present the panel data models that we use to evaluate whether pension funds and other depositors exert market discipline on banks, and in particular, the effect of bank connection to the pension fund industry. Second, in Section 5.2, we also assess to what extent the share of pension fund deposits in individual banks affects market discipline. Finally, in Section 5.3, we use the results from Section 5.2 to construct partial influence functions that describe how market discipline changes as the pension fund deposit share increases. 5.1. Models to evaluate the presence of market discipline and the effect of bank connection to the pension fund industry We separate the total time deposits of individual banks DEP into two components: those held by pension funds (DEPPF) and those held by other depositors (DEPOTHER), where

DEP ¼ DEPPF þ DEP OTHER We define a two-level model with depositors nested within banks, which is given by: j

j

DDEP ji;tl ¼ lji þ kt þ b0  F i;tl þ b0d  ðd  F i;tl Þ þ /0  ðF i;tl  CONNECT i Þ þ /0d  ðd  F i;tl  CONNECT i Þ þ eji;t ; ð1Þ  0 if j ¼ OTHER j . The indexes i, t, j, and l where i ¼ 1; . . . ; I; t ¼ 1; . . . ; T; j ¼ OTHER; PF; and d ¼ 1 if j ¼ PF indicate, respectively, the banking institution, the time period, the type of depositor, and the measurement lag. The panel is unbalanced, so T, the number of observations per bank, varies across institutions. The dependent variable DDEPji;tl measures the growth rate of time deposits held by depositor type j in bank i between the times t and t  l. Determinants of bank deposit growth are lagged by l months—measured at the beginning of the corresponding growth period. In all the regressions presented below, the reported lag is given by l = 6, owing to the fact that the average maturity of pension fund deposits is about 6 months during the study period. The bank and depositor specific fixed effects lji capture time-invariant unobserved characteristics of bank i and depositor type j. The time-specific effects kt control for macroeconomic and banking sector developments that have common effects across banks. Bank-specific fundamental variables as well as SIZE are time-variant and are collected in the 0 vector F i;tl ¼ ½Z  SCOREi;tl ; FUND0i;tl ; SIZEi;tl  , where FUND denotes the vector of CAMEL-type 0 fundamentals FUNDi,tl = [CAPRi,tl, LIQi,tl, PROFITi,tl, NPLLi,tl, NGOVBi,tl] . Estimated coefficients on interaction terms of the bank-specific fundamentals with the dummy variable dj will indicate

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

12

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

whether the disciplining behavior of pension funds is different from that of other depositors. Standard t-tests performed on those coefficients will determine whether such differences are statistically significant. Note that the treatment of the variable SIZE in Eq. (1) is similar to that of fundamentals.18 Finally, Eq. (1) includes interactions between bank fundamentals and the variable CONNECT, and between these and the dummy variable dj. Estimated coefficients on these interaction terms will indicate whether the disciplining behavior of both types of depositors is different toward connected and unconnected banks. To obtain robust conclusions and assess the significance of certain bank-specific characteristics, we also consider a variation of the model. Bank-specific characteristics are collected in the vector 0 DUMi = [CONNECTi, FOREIGNi, PRIVATEi] and the intercepts lji are replaced with the terms j j 0 0 a0 þ a1  d þ a2  DUMi þ a3  d  DUMi in Eq. (1). The DUM variables and their interactions with the dummy variable dj are time-invariant and thus act as intercept shifts in ordinary least square (OLS) regressions.19

5.2. Models to evaluate the effect of the share of bank deposits held by pension funds on market discipline A larger presence of pension funds in a bank’s deposit base (higher value of PFDEP) could improve their disciplining incentives; it could also weaken the incentives of other depositors to exert market discipline. We assess whether higher shares of pension fund deposits affect market discipline—measured by how strongly deposits respond to changes in bank fundamentals—through interaction effects. We define the following model: j

j

DDEP ji;tl ¼ lji þ kt þ b0  F i;tl þ c0  ðF i;tl  PFDEPi;tl Þ þ b0d  ðd  F i;tl Þ þ c0d  ðd  F i;tl  PFDEPi;tl Þ j

j

þ /0d  ðd  F i;tl  CONNECT i Þ þ s0d  ðd  F i;tl  CONNECT i  PFDEPi;tl Þ þ g  PFDEP i;tl j

j

þ gd  ðd  PFDEP i;tl Þ þ qd  ðd  CONNECT i  PFDEP i;tl Þ þ eji;t ;

ð2Þ

Eq. (2) includes bank and depositor-specific fixed effects and time-specific effects.The terms 0

c  (Fi,tl  PFDEPi,tl) collect the interactions of the bank fundamentals and SIZE with the share of pension fund deposits. This model also incorporates the effects of bank connection to the pension fund industry—through the dummy variable CONNECT—as well as its interactions with the bank fundamentals, SIZE, and the pension fund deposit share. Note that Eq. (1) only includes interactions of bank fundamentals with dummy variables; in contrast, Eq. (2) also includes interactions of fundamentals with a continuous variable (PFDEP). This implies that while in Eq. (1) all forms of depositor discipline can be assessed by testing whether single coefficients or sums of coefficients are different from zero, in Eq. (2) the task is more complex. In the latter case, market discipline could vary with the pension fund deposit share, making possible results such as ‘‘non-pension fund depositors exert discipline with respect to a fundamental F when the pension fund deposit share is small, but not when it is large.’’ To evaluate depositor discipline in the presence of interactions of bank fundamentals with the continuous variable PFDEP, we must now define partial influence functions. 18 On theoretical grounds, the relation between size and market discipline is ambiguous and complex. First, a ‘‘too big to fail’’ problem could weaken the market discipline exerted on big banks relative to small banks. In the case of Argentina, however, the capacity of the government to bail out banks was severely constrained by the widespread dollarization of deposits and the currency board monetary regime. Second, large depositors could account for a larger share of deposits in big banks than in small banks—wealthy families, corporations, and institutional investors could prefer to do business with big banks that offer a broader set of financial and support services. In such case, large banks would be subject to more strict discipline than small banks. 19 Bank-specific characteristics are time invariant for all the institutions in our sample. However, some bank mergers and acquisitions, which occurred during the study period, changed bank-specific characteristics. Data corresponding to an acquired bank was discontinued in the sample. Data corresponding to the individual merged banks is discontinued and a new consolidated institution is added to the sample.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

13

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

5.3. Market discipline measures based on partial influence functions The model specified above by Eq. (2) allows for variation in depositor responses to changes in bank fundamentals depending on the share of pension fund deposits and depositor type. A partial influence function summarizes these interactions by measuring the effect on deposit growth of a marginal change in a fundamental. Consider first the case of an unconnected bank (CONNECTi = 0) and let F be any bank fundamental or SIZE: F e {Z  SCORE, CAPR, LIQ, PROFIT, NPLL, NGOVB, SIZE}. Denote bF and bdF the coefficients (elements of the vectors b and bd) corresponding to the variables Fi,tl and dj  Fi,tl in Eq. (2). Similarly, denote cF and cdF the coefficients corresponding to the variables Fi,tl  PFDEPi,tl and dj  Fi,tl  PFDEPi,tl. The partial influence function corresponding to the fundamental F and depositor type j (PIjF ) is given by the partial derivative of the function DDEPji;tl with respect to Fi,tl:

PIjF ¼

@ DDEPji;tl @F i;tl

 ¼

bF þ cF  PFDEP i;tl

if

j ¼ OTHER

ðbF þ bdF Þ þ ðcF þ cdF Þ  PFDEP i;tl

if

j ¼ PF

:

Note that the PIjF functions are linear and differ across depositor types. The intercepts of the PI functions indicate the discipline exerted by both types of depositors—with respect to the fundamental F—on a bank when the share of pension fund deposits is insignificant (as PFDEPi,tl ? 0). The sum of the intercepts and slopes (bF + cF and bF + bdF + cF + cdF) indicate the discipline exerted by depositors on a bank whose share of pension fund deposits approaches unity (as PFDEPi,tl ? 1). We use the parameter estimates from Eq. (2) to construct empirical PI functions. To account for the stochastic nature of parameter estimates, we construct confidence bounds using the Fieller method described in Appendix A. Our empirical definition of depositor discipline is as follows: a depositor type j exerts discipline with respect to the fundamental F for a given pension fund deposit ratio PFDEPi,tl if all the points in the confidence interval are of the ‘‘correct’’ sign and strictly different from 0. More precisely, we take the ‘‘correct’’ sign to mean that deposits grow more rapidly in banks with higher Z-SCORE, CAPR, LIQ, and PROFIT; or lower NPLL and NGOVB. And we conclude that there exists market discipline when we reject the ‘‘no discipline’’ hypothesis at the 95% confidence level. Note that according to this definition, conclusions about depositor discipline can vary with the pension fund deposit share. Starting with the partial influence functions corresponding to the two depositors’ groups, we can construct an aggregate partial influence function. The growth rate of total time deposits in bank i between the times t and t  l (DDEPi,tl) can be expressed as follows: DDEPi;tl ¼ DDEPPF i;tl  PFDEP i;tl þ DDEPOTHER  ð1  PFDEP i;tl Þ: The aggregate partial influence function is quadratic and given i;tl by:

@ DDEPi;tl ¼ bF þ ðbdF þ cF Þ  PFDEPi;tl þ cdF  ðPFDEP i;tl Þ2 : @F i;tl Finally, the model specified by Eq. (2) in Section 5.2 allows variation in pension funds’ disciplining behavior toward connected and unconnected banks. To evaluate the effect of bank connection to the pension fund industry on market discipline, we define and estimate different PI functions of pension fund depositors for both types of banks—by setting the value of CONNECTi equal to 0 or 1, respectively—as follows:

PIPF F ¼

@ DDEPji;tl @F i;tl

 ¼

ðbF þ bdF Þ þ ðcF þ cdF Þ  PFDEP i;tl

if

CONNECT i ¼ 0

ðbF þ bdF þ /dF Þ þ ðcF þ cdF þ sdF Þ  PFDEP i;tl

if

CONNECT i ¼ 1

:

Appendix A shows how to construct confidence intervals for these specific PI functions. 6. Econometric results In Section 6.1 we discuss regression results corresponding to Eq. (1) and evaluate the disciplining behavior of pension funds and other depositors on connected and unconnected banks. In Section 6.2, Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

14

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

we discuss regression results corresponding to Eq. (2) and evaluate whether changes in the pension fund deposit share affect market discipline.

6.1. Market discipline and the effects of bank connection to the pension fund industry Table 5a presents results for regressions corresponding to Eq. (1), and Table 5b summarizes the results of tests on the significance of different sums of coefficients. In regressions 1–3 and 6–7, deposit growth is expressed as a function of a single summary fundamental, the z-score; in regressions 4 and 5, individual CAMEL-type fundamentals are also included. Regressions 1–5 show estimations with bank and depositor specific fixed effects over the entire sample of depositors, while regressions 6 and 7 are run on pension funds and other depositors separately. The simplest test of depositor discipline—whether depositors as a whole respond to changes in the z-score—provides weak evidence of market discipline: regression 1 yields a positive coefficient for ZSCORE, but only at a 10% significance level. However, once we differentiate the disciplining behavior of pension funds from that of other depositors in regression 2—by including an interaction term between the variable Z-SCORE and the pension fund dummy dj —we find that other depositors exert market discipline, but pension funds do not, as Table 5b indicates that the sum of the two coefficients is negative and significant. Specifically, the semi-annual growth rate of non-pension fund deposits increases by about 0.4 percentage points in response to a 10-point increase in the z-score. For a bank that increases its z-score from 14.2 (the 25th percentile value in the indicator’s distribution across banks) to 65.2 (the 75th percentile value), the semi-annual growth rate of non-pension fund deposits would increase by about 2 percentage points (4 percentage points annually). Regression 3 allows for different disciplining behavior of both types of depositors across connected and unconnected banks. The results show that other (non-pension fund) depositors exert market discipline on both connected and unconnected banks. In sharp contrast, pension funds do not exert discipline: the response to the z-score is negative for both connected and unconnected banks. Regressions 6 and 7 show pooled estimations of Eq. (1) separately for pension funds and other depositors, and include the three (unreported) bank-specific characteristics, CONNECT, PRIVATE, and FOREIGN. The results confirm the different disciplining behavior of pension funds and other depositors, and show the undermining effect of bank connection to the pension fund industry on market discipline. Pension funds do not respond to changes in the z-score in unconnected banks, but respond negatively in connected banks (regression 6), whereas other depositors respond positively and significantly to these changes in unconnected banks only (regression 7).20 These results are consistent with a reallocation of pension fund deposits away from unconnected banks with relatively strong fundamentals toward connected banks, with little consideration of the latter group’s fundamentals. In this manner, the presence of connection between banks and pension funds appears to have undermined market discipline in the banking system.21 Regressions 4 and 5 expand the set of determinants of deposit growth to include CAMEL-type fundamentals and SIZE. Since the z-score is a summary indicator that embeds the effects of capitalization and profitability, regression 4 incorporates the z-score together with only the remaining CAMEL-type fundamentals (LIQ, NPLL, and NGOVB). Regression 5 excludes the z-score but incorporates capitalization and profitability (CAPR and PROFIT). 20 Note that in regression 6, the non-interacted CONNECT dummy (not reported) has a positive and significant coefficient. Thus, regardless of all other characteristics such as fundamentals, banks connected to the pension fund industry experience higher deposit growth. This reinforces the relevance of links between certain banks and pension funds. 21 Throughout our analysis we adopt the standard view of the market discipline literature that the behavior of banks deposits is ‘‘supply-driven’’. That is, depositors observe bank-specific characteristics and macroeconomic conditions and decide where to place their funds. Thus, we interpret our results as reflecting a preference by pension funds for connected banks. An alternative, ‘‘demand-driven’’ interpretation might attribute the reallocation of deposits to an effort by banks to attract pension fund deposits, particularly as their own fundamentals were deteriorating during the crisis episode. However, it is not clear why banks would favor these deposits. Certainly, although these were more stable at the aggregate level, it should have been apparent to banks that pension funds were quite willing to shift funds rapidly across banks. Moreover, whether connected banks in particular were trying to attract pension fund deposits or pension funds were favoring connected banks, the end result would be the same: a weakening of market discipline brought about by conflicts of interest.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Dependent variable: real growth rate of deposits (DDEP), 6 months (l = 6)

Fixed effects

Pooled

All depositors Explanatory variables Z-SCORE

Z-SCORE (l)

FUND

CAPR (l)

Pension funds

(1)

(2)

(3)

(4)

0.00025 (1.78)⁄

0.00039 (2.94)⁄⁄⁄

0.00034 (2.36)⁄⁄

0.00047 (3.37)⁄⁄⁄

LIQ (l)

0.265 (1.59)

PROFIT (l) NPLL (l) NGOVB (l) j

d  Z-SCORE j

d  FUND

dj  Z-SCORE (l)

0.00086 (3.72)⁄⁄⁄

0.00082 (3.21)⁄⁄⁄

0.077 (1.63) 0.002 (0.02) 0.00108 (3.22)⁄⁄⁄

j

d  CAPR (l) dj  LIQ (l)

1.150 (2.99)⁄⁄⁄

dj  PROFIT (l) dj  NPLL (l) dj  NGOVB (l) Z-SCORE  CONNECT

Z-SCORE (l)  CONNECT

FUND  CONNECT

CAPR (l)  CONNECT LIQ (l)  CONNECT

0.00039 (0.90)

0.004 (0.02) 0.003 (0.01) 0.00015 (0.33)

0.517 (1.65)⁄

PROFIT (l)  CONNECT NPLL (l)  CONNECT NGOVB (l)  CONNECT

0.296 (2.20)⁄⁄ 0.298 (1.17)

(5)

Other depositors

(6)

(7)

0.00029 (1.13)

0.00045 (5.26)⁄⁄⁄

0.00156 (2.47)⁄⁄

0.00031 (2.02)⁄⁄

0.542 (4.35)⁄⁄⁄ 0.244 (2.04)⁄⁄ 0.316 (0.57) 0.029 (1.01) 0.163 (1.91)⁄

0.143 (0.71) 1.293 (3.96)⁄⁄⁄ 3.367 (0.89) 0.163 (0.90) 0.071 (0.44)

1.314 (2.26)⁄⁄ 0.314 (1.07) 0.200 (0.05) 0.271 (2.73)⁄⁄⁄ 0.082 (0.51)

15

(continued on next page)

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 5a Deposit response to changes in the Z-SCORE and CAMEL-type fundamentals: pension funds versus other depositors and connected versus unconnected banks (Eq. (1)).a

16

Dependent variable: real growth rate of deposits (DDEP), 6 months (l = 6)

Fixed effects

Pooled

All depositors Explanatory variables

(1)

dj  Z-SCORE  CONNECT

dj  Z-SCORE (l)  CONNECT

dj  FUND  CONNECT

dj  CAPR (l)  CONNECT

(2)

(3)

(4)

0.00007 (0.24)

0.00140 (2.33)⁄⁄

dj  LIQ (l)  CONNECT

1.989 (3.72)⁄⁄⁄

j

d  PROFIT (l)  CONNECT dj  NPLL (l)  CONNECT dj  NGOVB (l)  CONNECT SIZE j

SIZE (l)

d  SIZE

dj  SIZE (l)

SIZE  CONNECT

SIZE (l)  CONNECT

dj  SIZE  CONNECT

dj  SIZE (l)  CONNECT

dj

dj

Number of observations F-test for DUM and fixed effects (p-value) R2

2292 4.36 (0.00) 0.259

2292 4.28 (0.00) 0.265

2292 4.11 (0.00) 0.266

0.491 (1.33) 0.577 (1.51) 1.481 (0.55) 1.830 (2.86)⁄⁄⁄ 0.741 (0.21) 3.305 (3.68)⁄⁄⁄ 0.096 (2.57)⁄⁄⁄ 2275 4.00 (0.00) 0.300

Pension funds

Other depositors

(5)

(6)

(7)

0.040 (0.08) 1.136 (2.39)⁄⁄ 18.736 (1.27) 0.075 (0.24) 0.285 (1.03) 0.127 (0.08) 0.488 (1.08) 2.832 (1.38) 1.366 (2.10) 0.076 (2.13)⁄⁄ 3200 4.48 (0.00) 0.269

447 11.11 (0.00) 0.163

1839 27.09 (0.00) 0.197

a This table reports OLS and fixed effects regressions with robust standard errors of real growth of time deposits held by pension funds and other depositors on bank-specific factors. All regressions include time effects (not reported). Regressions (1)-(5) include fixed effects for banks and depositor types (not reported); regressions (6) and (7) include the variables DUM as regressors, but the estimated coefficients are not reported. In all regressions a constant is estimated but not reported. The table is based on monthly data; deposit growth rates are calculated over 6 month periods and regressors are measured at the beginning of the periods. t-statistics are indicated in parentheses, with 10% (⁄), 5% (⁄⁄), and 1% (⁄⁄⁄) significance levels. The table also reports F-tests on the significance of DUM variables in pooled regressions and bank dummies in fixed effects regressions (which are compared against ‘‘restricted’’ regressions that include the variables DUM, d, and dxDUM), with p-values in parentheses.

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 5a (continued)

Bank fundamental Regression in Table 5a Z-SCORE

CAPR

LIQ

PROFIT

NPLL

NGOVB

Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks

(1) (2)

(3)

(4)

() F = 3.28 (0.07) () F = 3.31 (0.07) () F = 2.92 (0.09) (+) F = 2.32 (0.13) (+) F = 3.20 (0.07) (+) F = 0.61 (0.43)

Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks

(5)

(6)

(7)

() F = 3.43 (0.06) () F = 9.69 (0.00) () F = 1.09 (0.30) (+) F = 2.87 (0.09) (+) F = 0.44 (0.51) () F = 1.84 (0.17) () F = 12.30 (0.00) () F = 20.25 (0.00) () F = 47.53 (0.00) () F = 40.90 (0.00) (+) F = 0.82 (0.37) (+) F = 0.06 (0.80)

Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks

(+) F = 0.95 (0.33) (+) F = 2.48 (0.12) (+) F = 0.00 (0.98)

Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks

(+) F = 0.14 (0.71) (+) F = 2.83 (0.09) () F = 2.98 (0.08)

(+) F = 0.55 (0.46) (+) F = 0.68 (0.41) () F = 9.82 (0.00)

Pension funds on all banks Pension funds on unconnected banks Pension funds on connected banks Other depositors on connected banks

() F = 0.00 (0.99) (+) F = 2.41 (0.12) () F = 1.59 (0.21)

(+) F = 2.09 (0.15) (+) F = 0.04 (0.84) () F = 3.16 (0.08)

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

a This table reports additional F-tests for sums of coefficients corresponding to the pooled and fixed effects regressions presented in Table 5, with the signs of coefficient sums and pvalues in parentheses. Specifically, in regression (2), the hypothesis tested is the following: Pension funds on all banks H0: Coeff. Z-SCORE (l) + Coeff. dj  Z-SCORE (l) = 0. In regressions (3)–(5) and for any fundamental variable Y, the hypothesis tested is the following: Pension funds on unconnected banks H0: Coeff. Y (l) + Coeff. dj  Y (l) = 0; Pension funds on connected banks H0: Coeff. Y (l) + Coeff. dj  Y (l) + Coeff. dj  Y (l)  CONNECT = 0; Other depositors on connected banks H0: Coeff. Y (l) + Coeff. Y (l)  CONNECT = 0. In regression (6): Pension funds on connected banks H0: Coeff. Z-SCORE (l) + Coeff. Z-SCORE (l)  CONNECT = 0. In regression (7): Other depositors on connected banks H0: Coeff. Z-SCORE (l) + Coeff. Z-SCORE (l)  CONNECT = 0.

17

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 5b F-tests on significance of disciplining behavior of pension funds and other depositors on connected and unconnected banks (based on results of Table 5, Eq. (1)).a

18

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

These regressions confirm previous findings: evidence of market discipline for other depositors only, and primarily with respect to unconnected banks. Regarding the other fundamentals included, the capitalization ratio provides the strongest evidence of market discipline. There is a positive and significant response of pension funds and other depositors to increases in bank capitalization— except for connected banks, where the response is not significant. To illustrate the magnitude of the effects, we find that a 1 percentage point increase in the capital to assets ratio is associated with an increase in the semi-annual growth rate of non-pension fund deposits of about 0.5 percentage points. For a bank that increases its capital to assets ratio from 0.08 to 0.18 (from the 25th to the 75th percentile value in the indicator’s distribution across banks), the semi-annual growth rate of non-pension fund deposits would increase by about 5 percentage points (10 percentage points annually). On the other hand, pension funds respond perversely to liquidity across all banks,22 and to non-performing loans of connected banks in particular (regression 4), although the latter result is somewhat weaker. There is evidence that non-pension fund depositors favor banks with low exposure to the government. In contrast, there is no evidence that pension funds exert discipline with respect to this exposure on either connected or unconnected banks. 6.2. Results on the effect of the pension fund deposit share Table 6 displays estimations corresponding to Eq. (2), with regression 1 including the z-score along with a subset of CAMEL-type fundamentals, and regression 2 including all individual CAMEL-type fundamentals but excluding the z-score. Figs. 3 and 4 plot the respective PI functions showing the effect of each of the bank fundamentals on deposit growth as the pension fund deposit share increases. The main results can be described as follows. Based on the PI functions presented in Fig. 3 (upper panels), we find some evidence that pension funds and other depositors exert market discipline on banks considered as a single group. First, non-pension fund depositors exert discipline only when the share of pension fund deposits (PFDEP) is small. Such discipline is exerted with respect to the z-score (Z-SCORE), the capitalization ratio (CAPR), and the net exposure to the government ratio (NGOVB), and vanishes completely as the share of pension fund deposits rises above a threshold (somewhere in the range 0.2–0.4).23 Thus, the presence of large pension fund deposits in a bank possibly crowds out the market discipline exerted by non-pension fund depositors.24 Second, pension funds also exert discipline only when the aggregate share of pension fund deposits in a bank is small. This discipline is exerted only with respect to the capitalization ratio (CAPR) and vanishes when the share of pension fund deposits rises above a threshold of about 50%. We also find that pension fund deposit growth favors less liquid banks, seemingly at odds with disciplining behavior. However, this behavior might be justifiable. During non-crisis times, pension funds could view liquidity as a sign that banks are not generating sufficient profitability, and prefer to move their deposits to banks that are pursuing more aggressive credit expansion strategies.25 During crisis

22

For a detailed discussion of the response of depositors to changes in the liquidity ratio, see Section 6.2. The magnitude of the response of non-pension fund deposits to changes in the net exposure to the government ratio (NGOVB) when the share of pension fund deposits is small—less than 20%—can be described as follows. A 1 percentage point increase in the net exposure to the government ratio reduces the semi-annual deposit growth rate by about 0.16 percentage points. For a bank that increases its net exposure to the government ratio from 0 to 0.09 (from the 25th to the 75th percentile value in the indicator’s distribution across banks), the semi-annual growth rate of non-pension fund deposits would decrease by about 1.4 percentage points, and the annual growth rate would decrease by about 2.7 percentage points. Note that as the pension fund deposit share increases, this estimated effect loses statistical significance. 24 Strictly speaking, according to our analysis of the disciplining behavior of non-pension fund depositors, we reject the ‘‘no discipline’’ hypothesis when the share of pension fund deposits is low, but we cannot reject it when the share of pension fund deposits is sufficiently high. 25 This interpretation is consistent with previous results in the empirical banking literature in emerging market economies; see Barajas and Steiner (2000). 23

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Dependent variable: real growth rate of deposits (DDEP), 6 months (l = 6) Explanatory variables Z-SCORE

(1) Z-SCORE (l)

FUND

CAPR (l) LIQ (l)

0.089 (0.58)

PROFIT (l) NPLL (l) NGOVB (l) dj  Z-SCORE dj  FUND

dj  Z-SCORE (l)

0.054 (1.16) 0.072 (0.72)

FUND  PFDEP

CAPR (l)  PFDEP (l) LIQ (l)  PFDEP (l)

NPLL (l)  PFDEP (l) NGOVB (l)  PFDEP (l)

dj  CAPR (l)  PFDEP (l)

0.570 (2.23)⁄⁄ 0.970 (3.36)⁄⁄⁄

SIZE (l)

0.766 (0.41)

2.265 (1.86)⁄

SIZE  PFDEP

dj  SIZE (l)

0.871 (0.68) 0.097 (0.61)

0.328 (0.39) 0.229 (1.65)⁄

dj  SIZE  PFDEP

0.202 (0.93)

0.112 (0.67)

0.987 (2.01)⁄⁄

j

d  NGOVB (l)

PFDEP (l)

dj  PFDEP (l)

(2)

0.00136 (0.97)

2.012 (1.06)

PROFIT (l)  PFDEP (l)

dj  FUND  PFDEP

dj  CAPR (l)

dj  NPLL (l)

dj  PFDEP

Z-SCORE (l)  PFDEP (l)

0.222 (0.64) 0.951 (2.11)⁄⁄ 11.716 (1.27) 0.513 (2.12)⁄⁄ 0.523 (2.04)⁄⁄

0.00092 (2.29)⁄⁄

dj  PROFIT (l)

PFDEP

(1)

Z-SCORE  PFDEP

dj  Z-SCORE (l)  PFDEP (l)

d  LIQ (l)

dj  SIZE

0.484 (3.88)⁄⁄⁄ 0.189 (1.70)⁄ 0.404 (0.60) 0.032 (1.15) 0.205 (2.55)⁄⁄

Explanatory variables

dj  Z-SCORE  PFDEP

j

SIZE

(2)

0.00043 (3.03)⁄⁄⁄

1.489 (1.59) 0.423 (1.05)

0.173 (0.27) 0.961 (0.67) 0.122 (0.05) 0.078 (0.10) 0.115 (0.28)

0.002 (1.31)

4.254 (3.06)⁄⁄⁄ 2.187 (3.83)⁄⁄⁄

0.697 (0.66) 3.165 (1.36) 15.171 (1.30) 4.785 (3.63)⁄⁄⁄ 0.972 (2.00)⁄⁄

SIZE (l)  PFDEP (l)

14.107 (1.91)⁄

9.773 (1.65)⁄

dj  SIZE (l)  PFDEP (l)

14.027 (1.02)

5.783 (0.64)

j

d  LIQ (l)  PFDEP (l)

0.663 (0.23)

dj  PROFIT (l)  PFDEP (l) dj  NPLL (l)  PFDEP (l) j

d  NGOVB (l)  PFDEP (l)

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

(continued on next page) 19

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 6 Deposit response to changes in the Z-SCORE, CAMEL-type fundamentals, and the pension fund deposit share: pension funds versus other depositors and connected versus unconnected banks (Eq. (2))a.

20

Dependent variable: real growth rate of deposits (DDEP), 6 months (l = 6) Explanatory variables

(1)

j

j

d  ZSCORE  CONNECT

d  Z-SCORE (l)  CONNECT

(2)

Explanatory variables j

0.00191

d  ZSCORE  PFDEP  CONNECT

(1) j

d  Z-SCORE (l)  PFDEP (l)  CONNECT

d  FUND  CONNECT

j

d  CAPR (l)  CONNECT

dj  LIQ (l)  CONNECT

0.441

2.336 (2.44)⁄⁄

dj  PROFIT (l)  CONNECT dj  NPLL (l)  CONNECT

dj  NGOVB (l)  CONNECT j

d  SIZE  CONNECT

dj  PFDEP  CONNECT

Number of observations F-test for fixed effects (p-value) R2

j

d  SIZE (l))  CONNECT

dj  PFDEP (l)  CONNECT

0.038 (2.77)⁄

(1.50) j

(2)

j

d  FUND  PFDEP  CONNECT

(0.37) 2.096 (2.14)⁄⁄ 6.957

0.232

(0.22) 0.945

(0.37) 0.218

(2.74)⁄⁄⁄ 0.568

(0.28)

(0.96)

1.521

2.372

(0.73)

(1.95)⁄

5.713

5.699

(3.41)⁄⁄⁄

(3.19)⁄⁄⁄

2275

3200

3.737 (0.00) 0.330

4.20 (0.00) 0.293

j

d  CAPR (l)  PFDEP (l)  CONNECT dj  LIQ (l)  PFDEP (l)  CONNECT

46.159

9.618 (0.80)

dj  PROFIT (l)  PFDEP (l)  CONNECT dj  NPLL (l)  PFDEP (l)  CONNECT dj  NGOVB (l)  PFDEP (l)  CONNECT j

j

d  SIZE  PFDEP  CONNECT

d  SIZE (l)  PFDEP (l)  CONNECT

dj

dj

(2.94)⁄⁄⁄ 15.697 (1.66)⁄ 82.763

0.962

(0.28) 14.713

(0.11) 0.196

(2.59)⁄⁄⁄ 12.558

(0.03)

(1.76)⁄

118.550

88.871

(4.55)⁄⁄⁄

(4.65)⁄⁄⁄

0.004

0.305

(0.07)

(1.63)

a This table reports a fixed-effects regression with robust standard errors of real growth of time deposits held by pension funds and other depositors on bank-specific factors. The regression includes time effects and fixed effects for depositor types and banks (not reported). The table is based on monthly data; deposit growth rates are calculated over 6 month periods and regressors are measured at the beginning of the corresponding periods. t-statistics are indicated in parentheses, with 10% (⁄), 5% (⁄⁄), and 1% (⁄⁄⁄) significance levels. The table also reports the result of an F-test on joint significance of bank dummies (compared against ‘‘restricted’’ regressions that include the variables DUM, d, and dxDUM), with p-values in parentheses.

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table 6 (continued)

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

21

Fig. 3. Partial influence functions by type of depositor. The PI functions corresponding to Other Depositors, Pension Funds on Unconnected Banks, and Pension Funds on Connected Banks are obtained from regressions (1) and (2) in Table 6 (Eq. (2)). The PI functions corresponding to Pension Funds are obtained from the regressions presented in Appendix B.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

22

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Fig. 3 (continued)

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

23

Fig. 3 (continued)

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

24

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Fig. 3 (continued)

times, although conventional wisdom would interpret asset liquidity as a fundamental strength allowing banks to withstand larger deposit withdrawals if faced with bank runs, the opposite is true when deposit withdrawals are motivated by fear of a currency crisis—as was the case during 2001 in Argentina. In this case, banks holding more liquid assets denominated in domestic currency may suffer withdrawals from large depositors—such as pension funds—who follow a rule of ‘‘selling liquid assets first.’’ Thus, our result on the relation between bank liquidity and pension funds’ deposit growth is consistent with rational behavior by pension funds in both crisis and non-crisis times. Finally, regarding the empirical validity of results corresponding to Eq. (2), note that F-tests performed on the regressions reject the pooled model and support the inclusion of bank and depositor specific fixed effects. To summarize, we consider banks as a single group and find little evidence that pension funds contribute to depositor discipline. Pension funds exert some discipline when their weight as depositors is small, and thus, whatever discipline exists is driven mainly by non-pension fund depositors. But we find no evidence that pension funds exert discipline when they are large, and hence, the main drivers of aggregate deposit behavior. The aggregate PI functions shown in Fig. 4 further support these conclusions, which are critically influenced by the effect of some banks’ connection to the pension fund industry, reflecting conflicts of interest. We also find that when pension funds exert discipline, they do so only on unconnected banks. The PI functions shown in Fig. 3 (lower panels) reveal a sharp contrast between pension funds’ disciplining behavior toward connected and unconnected banks. First, connected banks gain pension fund deposits when their z-scores decrease, and this effect gets stronger as the share of pension fund deposits in the banks increases. For unconnected banks, the zscore has a non-significant effect on pension fund deposits. Second, unconnected banks gain pension fund deposits when their capitalization ratios increase. The opposite holds true for connected banks: they gain pension fund deposits when their capitalization ratios decrease.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

25

Fig. 4. Aggregate partial influence functions. (The aggregate PI functions are obtained from the regressions presented in Appendix B. They indicate percentage changes in semi-annual deposit growth in response to a 1 percentage point increase in CAMEL-type fundamentals or the SIZE variable. For the z-score, the PI function shows the response of deposit growth to a 10 percentage point increase in Z-SCORE.)

Third, unconnected banks lose pension fund deposits when their asset liquidity ratios increase. That is not the case for connected banks, where higher liquidity has no significant effect on pension fund deposits (except when the share of pension fund deposits is small). Thus, we refine earlier results and conclude that pension funds raided only unconnected banks to mop up their liquidity. Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

26

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Fourth, unconnected banks lose pension fund deposits when their non-performing loans ratios increase,26 but connected banks do not—consistent with connected banks attracting pension fund deposits even when their assets become riskier. Fifth, while bank size does not affect pension fund deposit growth in unconnected banks, smaller connected banks appear to be preferred by pension funds. All these results point to a detrimental impact of conflicts of interest—stemming from bank-pension fund connections—on depositor discipline. 7. Conclusion We study the behavior of private pension funds as large depositors in Argentina in the period 1998–2001. Using panel data analysis, we address the questions whether, and if so how, pension funds influence market discipline. We reach two main conclusions. First, pension funds do exert limited market discipline; this discipline is exerted on unconnected banks, but it weakens as the share of pension fund deposits in a bank increases. Second, conflicts of interest undermine the disciplining role of pension funds. Banks’ ownership of Pension Fund Management Companies has detrimental effects on pension funds’ disciplining behavior. Banks connected to the pension fund industry tend to gain pension fund deposits as their fundamentals deteriorate. Our results also suggest that an increasing share of pension fund deposits in the deposit base could crowd out the discipline exerted by non-pension fund depositors. Finally, this paper has a clear policy implication: regulations should address the conflicts of interest between pension funds and banks to enhance depositor discipline. One possibility is to preclude pension funds from investing their deposits in banks that control them. However, reciprocity arrangements among banks could render this intervention ineffective; therefore it might be necessary to forbid banks’ ownership of companies involved in pension fund management to avoid the conflicts of interest studied in this paper. Acknowledgments We thank Badi Baltagi, Luis Brandao Marques and two anonymous referees for excellent comments and suggestions on previous drafts, and seminar participants at the IMF Western Hemisphere, Monetary and Capital Markets, and Institute for Capacity Development departments, as well as at the Research Department of the Central Bank of Brazil. The views expressed are those of the authors and do not necessarily represent those of the IMF or IMF policy. Appendix A. Using the Fieller method to construct confidence intervals for PI functions The estimated values of PFDEPi,tl that set the PI functions equal to 0 are given by:

PIjF

¼

@ DDEP ji;tl @F i;tl

8 > > <

b PFDEP i;tl ¼  bF if cbF ¼ 0 () > b c > : PFDEP i;tl ¼  ð bF þ bdF Þ if ð cbF þ c cdF Þ

j ¼ OTHER ; j ¼ PF

b and b where b c are estimates of b and c, respectively. Following Hirschberg and Lye (2010), 100  (1  a)% confidence intervals (CI) for the empirical PI functions are defined by:

26 We can illustrate the sensitivity of pension fund deposits to changes in the ratio of non-performing loans in unconnected banks as follows. Using the PI function, and taking the sensitivity to be 0.5 (corresponding to a share of pension fund deposits in the deposit base of about 0.2), we find that a 1 percentage point increase in the ratio of non-performing loans reduces the semi-annual growth rate of pension fund deposits by 0.5 percentage points. For an unconnected bank that increases its ratio of non-performing loans from 0.02 to 0.19 (from the 25th to the 75th percentile value in the indicator’s distribution across banks), the semi-annual growth rate of pension fund deposits would decrease by about 8 percentage points, and the annual growth rate would decrease by about 16 percentage points.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

27

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   dv ðb ; c Þ  PFDEP i;tl þ PFDEP 2  r b 2b þ 2  Co b 2c if j ¼ OTHER; CI ¼ c cF  PFDEP i;tl  ta=2  r bF þ c F F i;tl F F h   i  cF þ cc  PFDEP bF þ d bdF þ c CI ¼ c i;tl dF qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 d b ðb þb Þ þ 2  Cov ðbF þ bdF ; cF þ cdF Þ  PFDEP i;tl þ PFDEP 2i;tl  r b 2ðc þc Þ ta=2  r F

dF

F

dF

if j ¼ PF :

d b 2b ; r b 2c ; and Co In the previous expressions, r v ðbF ; cF Þ are estimates of the variances and the covariF F d b F and b b 2ðb þb Þ ; r b 2ðc þc Þ ; and Co ance corresponding to b c F . Similarly, r v ðbF þ bdF ; cF þ cdF Þ are estimates F dF F dF bF þ b b dF Þ and ð b of the variances and the covariance corresponding to ð b c F þ bc dF Þ:

dv ðb ; b Þ; r dv ðc ; c Þ; b 2ðc þc Þ ¼ r b 2c þ r b 2c þ 2  Co rb 2ðbF þbdF Þ ¼ rb 2bF þ rb 2bdF þ 2  Co F dF F dF F dF F dF dv ðb þ b ; c þ c Þ ¼ Co dv ðb ; c Þ þ Co dv ðb ; c Þ þ Co dv ðb ; c Þ þ Co dv ðb ; c Þ: Co F dF F F dF dF F dF F dF F dF Numerical calculations are based on robust estimation of the variance–covariance matrix of regression coefficients. Aggregate PI function: in this case the aggregate PI function is non-linear and the Fieller method cannot be applied directly. However, we can obtain confidence intervals associated with linear approximations to the PI function at each point PFDEPi,tl. We approximate the PI function bF + (bdF + cF)  PFDEP + cdF  (PFDEP)2 around the point PFDEP: The slope of the linearized function is given by: bdF þ cF þ 2  cdF  PFDEP; and its intercept is given by: bF  cdF  PFDEP2 . Hence, the linear approximation to the function can be written as follows:

 @ DDEPi;tl  ffi bF  cdF  PFDEP 2 þ bdF þ cF þ 2  cdF  PFDEP  PFDEP i;tl : @F i;tl Applying the Fieller method, a 100  (1  a)% confidence interval (CI) for the (linearized) PI function at the point PFDEP is given by: CI ¼

h

   i b c dF  PFDEP2 þ bb dF þ bc F þ 2  bc dF  PFDEP  PFDEP  ta=2 bF  b rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   dv b  c  PFDEP 2 ; b þ c þ 2  c  PFDEP  PFDEP þ PFDEP 2  r b2 b2 ;  r þ 2  Co F dF dF F dF 2 ðbF cdF PFDEP Þ

ðbdF þcF þ2cdF PFDEPÞ

dv ðb ; c Þ; b 2b þ r b 2c  PFDEP 4  2  PFDEP 2  Co ¼r F dF F dF   2 2 2 2 2 d b b b b r ðb þc þ2c PFDEPÞ ¼ r bdF þ r cF þ r cdF  4  PFDEP þ 2  Cov ðbdF ; cF Þ dF F dF h i dv ðb ; c Þ þ Co dv ðc ; c Þ ; þ4  PFDEP  Co dF dF F dF   dv b  c  PFDEP 2 ; b þ c þ 2  c  PFDEP ¼ Co dv ðb ; b Þ and Co F dF F dF dF F dF

b2 where r ðb

F cdF PFDEP

2Þ

dv ðb ; c Þ þ PFDEP  Co dv ðb ; c Þ þ Co F F F dF h i dv ðc ; b Þ þ Co dv ðc ; c Þ  PFDEP 3  r b 2c :  PFDEP 2  Co dF dF dF F dF A.1.1. PI Functions for analysis of banks’ connection to the pension fund industry Confidence intervals for the PI functions of pension fund depositors on connected banks (CONNECTi = 1) are given by:

CI ¼

h

  i  c d c c c bF þ bc dF þ /dF þ cF þ cdF þ sdF  PFDEP i;tl  t a=2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dv ðb þ b þ / ; c þ c þ sdF Þ  PFDEP i;tl þ PFDEP 2  r b 2ðc þc þs Þ b 2ðb þb þ/ Þ þ 2  Co  r F dF dF F dF i;tl F

dF

dF

F

dF

dF

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

28

Dependent variable: real growth rate of deposits (DDEP), 6 months (l = 6) Explanatory variables Z-SCORE

(1) Z-SCORE (l)

FUND

CAPR (l) LIQ (l)

0.152 (1.00)

PROFIT (l) NPLL (l) NGOVB (l) dj  Z-SCORE dj  FUND

dj  Z-SCORE (l)

0.052 (1.14) 0.060 (0.60)

CAPR (l)  PFDEP (l) LIQ (l)  PFDEP (l)

2.283 (1.24)

PROFIT (l)  PFDEP (l) NPLL (l)  PFDEP (l) NGOVB (l)  PFDEP (l)

1.671 (1.79)⁄ 0.369 (0.93)

0.284 (0.45) 1.031 (0.71) 0.316 (0.12) 0.154 (0.20) 0.041 (0.10)

dj  CAPR (l)  PFDEP (l)

0.514 (2.39)⁄⁄ 0.941 (3.92)⁄⁄⁄

SIZE (l)

1.620 (0.87)

1.269 (1.07)

SIZE  PFDEP

SIZE (l)  PFDEP (l)

18.507 (2.61)⁄⁄⁄

8.364

dj  SIZE (l)

0.283 (0.37)

0.242 (0.41)

dj  SIZE  PFDEP

dj  SIZE (l)  PFDEP (l)

0.204 (0.02)

7.082 (1.00)

PFDEP (l)

0.143 (0.94)

0.203 (1.48)

dj  PFDEP (l)

0.364 (2.16)⁄⁄

0.303 (1.93)⁄

dj

0.017 (0.36)

0.003 (0.07)

dj  NPLL (l) dj  NGOVB (l)

dj

FUND  PFDEP

(2)

0.00150 (1.06)

dj  FUND  PFDEP

1.743 (4.54)⁄⁄⁄

d  PROFIT (l)

dj  PFDEP

Z-SCORE (l)  PFDEP (l)

0.323 (1.01) 1.345 (3.80)⁄⁄⁄ 14.413 (1.80)⁄ 0.204 (0.87) 0.589 (2.99)⁄⁄⁄

dj  CAPR (l) d  LIQ (l)

PFDEP

(1)

Z-SCORE  PFDEP

dj  Z-SCORE (l)  PFDEP (l)

0.00016 (0.53)

j

dj  SIZE

0.476 (3.84)⁄⁄⁄ 0.236 (2.12)⁄⁄ 0.362 (0.54) 0.035 (1.26) 0.197 (2.45)⁄⁄

Explanatory variables

dj  Z-SCORE  PFDEP

j

SIZE

(2)

0.00031 (2.22)⁄⁄

j

d  LIQ (l)  PFDEP (l)

0.00383 (2.08)⁄⁄

0.990 (0.37)

j

d  PROFIT (l)  PFDEP (l) dj  NPLL (l)  PFDEP (l) dj  NGOVB (l)  PFDEP (l)

2.848 (2.35)⁄⁄ 2.350 (4.20)⁄⁄⁄

0.453 (0.41) 3.507 (1.64) 19.374 (1.87) 1.797 (1.23) 1.134

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Table B1 Deposit response to changes in the Z-SCORE, CAMEL-type fundamentals, and the pension fund deposit share: pension funds versus other depositors (Eq. (2)).

2275 4.06 (0.00) 0.307

3200 4.60 (0.00) 0.271

Note: This table reports fixed effects regressions with robust standard errors of real growth of time deposits held by pension funds and other depositors on bank-specific factors and macroeconomic risk indicators. The regressions include time effects and a constant that are estimated but not reported. The table is based on monthly data; deposit growth rates are calculated over 6 month periods and regressors are measured at the beginning of the periods. t-statistics are indicated in parentheses, with 10% (⁄), 5% (⁄⁄), and 1% (⁄⁄⁄) significance levels. The table also reports F-tests on the significance of bank dummies (which are compared against ‘‘restricted’’ regressions that include the variables DUM, d, and dxDUM), with p-values in parentheses.

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx 29

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002

Number of observations F-test for fixed effects (p-value) R2

30

A. Barajas, M. Catalán / J. Finan. Intermediation xxx (2014) xxx–xxx

Confidence intervals for the PI functions of pension fund depositors on unconnected banks are obtained as before. Appendix B. Additional regressions used to construct partial influence functions The partial influence functions of ‘‘Pension Funds’’ on all banks taken as a group were constructed based on the following regressions (see Table B1). References Barajas, A., Basco, E., Juan-Ramón, H., Quarracino, C., 2007. Banks during the argentine crisis: were they all hurt equally? Did they all behave equally? IMF Staff Pap. 54 (4), 621–662. Barajas, A., Steiner, R., 2000. Depositor Behavior and Market Discipline in Colombia. International Monetary Fund Working Paper No. 214. Berger, A., 1991. Market discipline in banking. In: Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, pp. 419–437. Berger, A., Turk-Ariss, R., 2010. Do Depositors Discipline Banks? An International Perspective. Moore School of Business, University of South Carolina, Columbia, South Carolina, Unpublished. Calomiris, C., Powell, A., 2001. Can emerging market bank regulators establish credible discipline? The Case of Argentina, 1992– 99. In: Mishkin, F. (Ed.), Prudential Supervision: What Works and What Doesn’t. University of Chicago Press, Chicago and London, pp. 147–191. Catalán, M., 2004. Pension funds and corporate governance in developing countries: what do we know and what do we need to know? J. Pension Econ. Finance 3 (2), 197–232. Clark, G., 2000. Pension Fund Capitalism. Oxford University Press, Oxford. Clark, G., 2004. Pension fund governance: expertise and organizational form. J. Pension Econ. Finance 3 (2), 233–253. Demirguc-Kunt, A., Detragiache, E., 2002. Does deposit insurance increase banking system stability? An empirical investigation. J. Monetary Econ. 49, 1373–1406. Demirguc-Kunt, A., Huizinga, H., 2004. Market discipline and deposit insurance. J. Monetary Econ. 51 (2), 375–399. Ioannidou, V., Penas, M.F., 2010. Deposit insurance and bank risk taking: evidence from internal loan ratings. J. Financ. Intermed. 19 (1), 95–115. Laeven, L., Valencia, F., 2012. Systemic Banking Crises Database: An Update. IMF Working Paper 12/163. Levy-Yeyati, E., Martínez Pería, M.S., Schmukler, S., 2010. Depositor behavior under macroeconomic risk: evidence from bank runs in emerging market economies. J. Money, Credit Bank. 42 (4), 585–614. Ljungqvist, A., Marston, F., Starks, L., Wei, K., Yan, H., 2007. Conflicts of interest in sell-side research and the moderating role of institutional investors. J. Financ. Econ. 85, 420–456. Martínez Pería, M.S., Schmukler, S., 2001. Do depositors punish banks for bad behavior? Market discipline, deposit insurance, and banking crises. J. Finance 56 (June), 1029–1051. Mahoney, P., 2004. Manager-investor conflicts in mutual funds. J. Econ. Perspect. 18 (2), 161–182, Spring. Mehran, H., Stulz, R., 2007. The economics of conflicts of interest in financial institutions. J. Financ. Econ. 85, 267–296. Mitchell, O., 2010. Implications of the Financial Crisis for Long Run Retirement Security. Pension Research Council Working Paper No. 2. Munnell, A., Aubry, J., Mudon, D., 2008. The Financial Crisis and State/Local Defined Benefit Plans. Center for Retirement Research at Boston College, No. 8–19, November. Mussa, M., 2002. Argentina and the Fund: From Triumph to Tragedy, Policy Analyses in International Economics, vol. 67. Institute for International Economics, Washington. Park, S., Peristiani, S., 1998. Market discipline by thrift depositors. J. Money, Credit, Bank. 30 (3), 347–364. Schumacher, L., 2000. Bank runs and currency run in a system without a safety net: Argentina and the ‘Tequila’ shock. J. Monetary Econ. 46 (August), 257–277. Whitehouse, E., 2009. Pensions during the crisis: impact on retirement income systems and policy responses. The Geneva Pap. Risk Insurance 34, 536–547. Hirschberg, J., Lye, J., 2010. A reinterpretation of interactions in regressions. Appl. Econ. Lett. 17 (5), 427–430.

Please cite this article in press as: Barajas, A., Catalán, M.. J. Finan. Intermediation (2014), http://dx.doi.org/ 10.1016/j.jfi.2014.04.002