DNB W O R K I N G P A P E R. DNB Working Paper. Internal Capital Markets and Lending by Multinational Bank Subsidiaries. No.

Transcription

1 DNB Working Paper No. 101/June 2006 Ralph de Haas and Iman van Lelyveld DNB W O R K I N G P A P E R Internal Capal Markets and Lending by Multinational Bank Subsidiaries

2 Internal Capal Markets and Lending by Multinational Bank Subsidiaries Ralph de Haas and Iman van Lelyveld * * Views expressed are those of the individual authors and do not necessarily reflect official posions of De Nederlandsche Bank. Working Paper No. 101/2006 June 2006 De Nederlandsche Bank NV P.O. Box AB AMSTERDAM The Netherlands

3 Internal Capal Markets and Lending by Multinational Bank Subsidiaries Ralph de Haas * and Iman van Lelyveld June 2006 Abstract We use panel data on the intra-group ownership structure and balance sheets of 45 of the largest banking groups from 1992 to 2004 to analyse what determines the cred growth of multinational bank subsidiaries. Both home- and host-country condions and characteristics of the subsidiaries themselves and their parent banks are taken into account. We find that the lending of multinational bank subsidiaries is influenced by substution effects, in which parent banks trade-off lending in several countries, as well as support effects, in which parent banks support weak subsidiaries. This provides strong evidence for the existence of internal capal markets through which multinational banks manage the cred growth of their subsidiaries. We also find that greenfield subsidiaries are more closely integrated into internal capal markets than subsidiaries that result from a take-over. JEL codes: Key words: F15, F23, F36, G21 multinational banks, cred supply, internal capal markets European Bank for Reconstruction and Development (EBRD), Office of the Chief Economist, One Exchange Square, London EC2A 2JN, Uned Kingdom. Tel.: Fax: dehaasr@ebrd.com. The contents of this working paper reflect the opinions of the individual authors and do not necessarily reflect the views of De Nederlandsche Bank or the EBRD. The authors would like to thank Jack Bekooij, Cees de Boer and Ron Wunderink for excellent research assistance and Arnoud Boot, Robert-Paul Berben and Maarten Gelderman for useful comments. De Nederlandsche Bank, Supervisory Policy Division, P.O. Box 98, 1000 AB Amsterdam, The Netherlands. Tel.: i.p.p.van.lelyveld@dnb.nl. Nijmegen School of Management, Radboud Universy, P.O. Box 9108, 6500 HK, Nijmegen, The Netherlands.

4 1. Introduction The wave of mergers and acquisions in the banking sector during the 1990s has increased the number of multinational banks, i.e. banks that own affiliates in more than one country. We use detailed data on the ownership structure and the cred supply of such regional and global banking groups to examine to what extent they manage the lending activies of their subsidiaries. In particular, we consider how the cred growth of multinational bank subsidiaries is influenced (i) by the macroeconomic suation in both the host country and the home country and (ii) by their own financial characteristics, the financial characteristics of their parent bank and the financial characteristics of other subsidiaries in the same banking group. 1 A better understanding of these issues is important now that in many regions most notably Central and Eastern Europe and Latin America large parts of the banking system consist of subsidiaries of multinational banks. In Western Europe consolidation has been primarily whin borders, but the recent take-overs of HVB Group and Banca Antonveneta by, respectively, Unicredo and ABN Amro have been heralded in the media as the start of further cross-border consolidation. Furthermore, in other parts of the world, including Russia and China, the process of multinational bank penetration may be expected to accelerate as well. More insight into the implications for the level and stabily of bank lending in these host countries is thus warranted. The remainder of this paper is structured as follows. In Section 2, we briefly discuss the existing lerature on the international cred supply of multinational banks. Section 3 then discusses our model specification and estimation methodology, followed by Section 4 which describes the data we use. Section 5 presents our empirical results and Section 6 concludes. 2. Multinational bank subsidiaries cred growth 2.1 Theoretical considerations on multinational bank lending The theoretical lerature that explicly deals wh the cred supply of the multinational banking firm is surprisingly limed. An important exception is the contribution by Morgan et al. (2004), who develop a two-region version of the Holmström and Tirole (1997) financial-intermediation model and apply this to state-level banking in the US. The authors model a multinational banking system that rebalances the international allocation of bank capal and cred in reaction to shocks to 1 We use the terms parent bank and bank holding alternately. Host country refers to the country where a multinational bank subsidiary operates and home country refers to the country where the head office is domiciled. Other countries are defined from the perspective of a particular subsidiary and include all other countries where the parent bank owns subsidiaries, excluding the host country of the particular subsidiary and excluding the home country. E.g. in the case of HSBC Mexico, the host country is Mexico, the home country is Great Brain, and other countries refers to all countries where HSBC owns subsidiaries except for Mexico and Great Brain. 1

5 eher bank capal self or to the capal of non-financial firms (in eher the home country or the host country). Through such rebalancing, multinational banks may act as a channel for the international transmission of financial and real-economic shocks. The two-country banking model shows that the presence of multinational banks tends to dampen bank-capal shocks in the host country because of cross-border capal support by the bank holding. At the same time, multinational banks tend to sharpen business cycles as they shift capal away from low-return countries to high return countries. We term these effects the support effect and the substution effect, respectively. The support effect implies that a multinational bank holding supports subsidiaries wh capal and liquidy in case of significant losses. The substution effect means that a multinational bank allocates capal over s subsidiaries depending on their expected risks and returns (portfolio view). In practise, multinational banks will use an internal capal market to allocate their scarce capal over various subsidiaries (Stein, 1997). The cred growth of subsidiaries that are connected through the internal capal market of the parent bank may then become correlated. 2 In the absence of capal market frictions, a multinational bank and s subsidiaries would not need to operate an internal capal market. Instead, subsidiaries would attract sufficient liabilies to finance each and every profable investment project themselves. 3 From a theoretical point of view, we can thus discern, on the one hand, a more or less frictionless world in which independent multinational bank subsidiaries choose their own optimal cred-growth strategy, independent of any financing they may receive from their parent bank. On the other hand, multinational banks may operate an internal capal market in which they allocate scarce capal to (some of) their subsidiaries, leading to support effects and substution effects. 4 We can now formulate five theoretical priors in order to test whether multinational bank subsidiaries in general are influenced by an internal capal market operated by their parent bank and whether some types of subsidiaries are controlled more strongly by their parent bank than other types. We expect subsidiaries that are the result of a take-over ( take-overs ) to be relatively independent, as they may not (yet) be fully integrated into the bank holding (cf. De Haas and Van Lelyveld, 2006). In contrast, we expect that greenfield subsidiaries ( greenfields ) which have 2 Houston et al. (1997) show for US bank holdings that the cred growth of a particular subsidiary is negatively correlated wh loan growth in other subsidiaries of the same holding. 3 In practice, banks do face limations wh regard to the amount of capal they can raise (Froot and Stein, 1998). Negative capal shocks may then force banks to reduce their assets in order to satisfy capal requirements (Van den Heuvel, 2002). 4 There is a parallel wh the Feldstein-Horioka puzzle in macroeconomics: the puzzle is that although capal markets are becoming more integrated, domestic investment is still linked to domestic savings. In our paper we are merely interested in establishing the degree of capal market integration whin the conglomerate. 2

6 been erected from scratch by the parent bank self tend to be closely integrated into the bank holding and s capal allocation mechanism. Our first theoretical prior is that in case of dependent multinational bank subsidiaries, hostcountry lending will be negatively related to the business cycle in the home country and in the other countries where the parent bank operates. More generally, host-country lending will be negatively related to macroeconomic developments that make lending in the home and other countries more attractive compared to cred expansion in the particular host country (substution effect). In contrast, lending by independent subsidiaries will not be related to economic developments in the home country and in the other countries of operation. Second, host-country lending by independent multinational bank subsidiaries will be negatively related to host-country financial or banking crises that wipe out part of the local banking capal. This may be the case because their own capal is reduced, because there is a reduced inflow (or even an outflow) of deposs or because the risks of lending have increased. In contrast, we expect dependent subsidiaries to be relatively insensive to episodes of host-country banking instabily because they can rely on parental capal and liquidy back-up (support effect). Parent banks may even stimulate subsidiaries to take advantage of banking system problems in the hostcountry by expanding market share. Third, host-country lending by independent multinational bank subsidiaries will be insensive to banking or financial crisis in the home country, where the parent bank is chartered, or in other countries of operation. In contrast, the lending of dependent subsidiaries will be sensive to such home-country or other-country banking problems, as the parent bank may reallocate capal from the host-country to the affected country (support effect). Fourth, lending by independent multinational bank subsidiaries will be influenced by the financial characteristics of the subsidiary self, such as capal posion, but not by the financial characteristics of the parent bank or of other subsidiaries. In contrast, lending by dependent subsidiaries that form part of an internal capal market will not (only) be sensive to the financial posion of the subsidiary but also to the financial posion of the parent bank or other subsidiaries. In particular, the financial characteristics of the parent bank will determine the strength and persistence of any support and substution effects. Fifth, the host-country lending of both dependent and independent multinational bank subsidiaries can be expected to be posively related to the host-country business cycle. More generally, the cred supply will be posively related to any host-country macroeconomic development that improves the risk/return characteristics of local investment opportunies. However, whereas independent subsidiaries will mainly increase lending on their own account, dependent subsidiaries will also increase lending because of their parent bank s influence 3

7 (substution effect). As the substution effect implies that the cred supply of a dependent subsidiary explicly reflects the trade-off between lending in the host country versus lending in the home country, we expect that the sensivy of dependent subsidiaries (such as greenfields) to hostcountry macroeconomic developments will be higher than the sensivy of independent subsidiaries (such as take-overs). 2.2 Empirical findings on multinational bank lending A number of empirical studies on multinational bank lending have been conducted in recent years. An important insight that can be derived from these studies is that parent banks tend to control their dependent subsidiaries. In line wh support effects, lending by multinational banks tends to dampen host-country financial shocks 5 and to transm home-country financial shocks. 6 Multinational bank lending also tends to be posively correlated wh the host-country business cycle 7 and to be correlated, eher posively or negatively, wh the home-country business cycle. 8 This last finding points to substution effects. A number of recent empirical results thus point in the direction of support and substution effects whin multinational banks. Unfortunately, most of these studies suffer from one or two important limations. First, many studies lim themselves to multinational bank links between specific home regions (Uned States, Japan or Western Europe) and specific host regions (Latin- America or Central and Eastern Europe). Second, a lot of studies are based on aggregate banklending data, so that no bank-level information on intrabank linkages can be exploed. We contribute to this empirical lerature by taking a global perspective in which we consider a much broader sample of host, home and other countries of operation. We also use a new and detailed data set wh bank-level information on multinational banks ownership structures and their international lending activy. This allows us to analyse the international linkages between the various subsidiaries of a multinational banking group in more detail and to come to more robust conclusions as regards the importance of internal capal markets whin such groups. 9 5 Dages et al. (2000), Peek and Rosengren (2000b), Goldberg (2001), Crystal et al. (2002), Martinez Peria et al. (2002) and De Haas and Van Lelyveld (2004, 2006). 6 Peek and Rosengren (1997, 2000a) and Van Rijckeghem and Weder (2000, 2001). 7 Dahl and Shrieves (1999), Buch (2000), Barajas and Steiner (2002), Jeanneau and Micu (2002), Morgan and Strahan (2004) and De Haas and Van Lelyveld (2006). 8 Calvo et al. (1993), Hernandez and Rudolph (1995), Dahl and Shrieves (1999), Goldberg (2001), Jeanneau and Micu (2002), Martinez Peria et al. (2002) and De Haas and Van Lelyveld (2006). Depending on the sample period and the sample of countries, some studies find a posive correlation between multinational bank lending and the home-country business cycle, while other studies find a negative relationship. 9 This aspect is also examined by Van Rijckeghem and Weder (2000, 2001), who find that multinational banks transm currency crises as they adjust cred lines to third countries in reaction to significant losses in a 4

8 3. Model specification and estimation methodology 3.1 Model specification We use a basic theoretical framework that treats multinational bank subsidiaries as independent banks that are neher capal constrained nor influenced by parental support or substution effects. The model, which follows Molyneux et al. (1998), is based on a portfolio-choice model in which risks and returns associated wh the host country and wh the characteristics of the subsidiary self fully determine the cred expansion of the subsidiary. We assume that the multinational bank subsidiary i is risk averse and sets a cred-growth target at the beginning of the year. The parent bank lets each subsidiary independently maximise s expected utily, which is a concave function of end-of-period profs? i,t+1. The logarhm of expected utily for a subsidiary is then: 1 ( Π ) Var ( ) U (1) = Et i, t+ 1 t Πi, t+ 1 2b Here, b is a time-varying risk-appete parameter that is determined by a vector Z of subsidiary specific and host-country specific variables that reflect the subsidiary s willingness to lend: b = Var. t Π i, t +1 b(z ). The variance of profs is denoted by ( ) End-of-period profs consist of the realised net return r i,t+1 on loans L minus realised costs C i,t+1 : Π i, t + 1 = i, t + 1L Ci, t + 1 r (2) The net return on loans equals the gross return on loans minus all costs that are proportional to the amount of loans. C i, t+1 then denotes all other costs and risks that are independent of the lending volume. At the beginning of the year, the multinational bank subsidiary is able to use current information to predict the returns that will materialise at the end of the year, although a certain part will remain unpredictable: ( X ) u 1 r (3) i, t + 1 = f + t + crisis country. However, the authors use aggregate BIS data by nationaly of lender and are thus unable to directly measure intrabank linkages at the bank level. Our data set overcomes this shortcoming. 5

9 where X i,t is a vector of variables that can be used, at the beginning of the year, to predict returns and where u t+1 represents the unpredictable component, which has a zero mean and a constant variance s 2. W i,t+1 is a vector of variables related to costs and risks which determine those costs that are not related to the lending amount: ( W ) C C (4) i, t + 1 = i, t + 1 This leads to the following condional expectation and condional variance of profs: t ( Π ) = E ( r ) L E ( C ) E (5) i, t + 1 t i, t + 1 t i, t + 1 Var t ( ) = σ L + σ Π (6) ic where the variance of costs 2 σ ic is assumed to be constant. We can now substute (5) and (6) into (1) and differentiate wh respect to L in order to get the first-order condion for prof maximisation: U L = f ( X ) σ b 2 L ( Z ) = 0 (7) * Solving (7) yields the cred target ( L ) that subsidiary i sets as a function of variables related to s risk aversion (Z ) and variables related to s expected returns (X ): ( Z X ) L * * = L, (8) We assume that the multinational bank subsidiary adapts the amount of cred in accordance wh a standard partial adjustment mechanism: * ( L L ) L L L = λ (9) 1 1 or 6

10 L ( 1 λ) L 1 * = L + λ (10) where? is the speed of adjustment parameter (which we assume is equal for all subsidiaries whin the same banking holding). This adjustment mechanism reflects that the subsidiary cannot instantaneously alter s loan portfolio on the basis of new risk/return information, but will only be able to adjust gradually. 10 The combination of the bank prof maximisation model and the partial adjustment model leads to a reduced form equation in which the cred growth of a subsidiary exclusively depends on variables that determine s risk aversion, variables that influence expected return and a lagged cred variable. 11 The risk and return variables Z and X, respectively can be eher subsidiary or host-country specific. 3.2 Estimation approach and methodology We estimate three types of regressions in order to test whether multinational bank subsidiaries independently determine their cred supply or whether their lending is (also) influenced by parental substution and support effects. In all three estimations, the dependent variable is the cred growth of subsidiary i and the independent variables include a standard set of host-country condions and a standard set of financial characteristics relating to subsidiary i self. This basic empirical set-up thus corresponds wh the theoretical model set out in the previous sub-section, but we are interested in explaining cred growth rather than the level of bank lending. In order to test for the presence of substution and support effects, we add a number of addional determinants in each of the three regressions (cf. the priors in Section 2.1). The first regression (11) is the basic model where we treat the subsidiary as if is operating on a stand alone basis. Cred growth is then solely determined by subsidiary-specific variables and host-country macroeconomic variables. In the second regression (12) we add as addional explanatory variables a number of parent bank characteristics. To the extent that support effects are important, we expect to find an influence of the financial characteristics of the parent bank On the supply side, existing loan commments lim the room for downward adjustments as the bank can only decide not to roll-over expiring loans. Upward adjustments are limed if banks are capal constrained. 11 Molyneux et al. (1998) find that lagged loan growth is the one and only significant determinant of current bank loan growth for a sample of foreign banks in the Uned States ( ). 12 As we use consolidated information on parent banks, endogeney problems may occur in case of large subsidiaries. However, as a percentage of total parent bank assets, individual subsidiaries are relatively small, especially in our sample of the world s largest banks which in many cases also have significant non-financial operations and equy participations. Reverse causaly is thus not likely to be an issue. 7

11 In the third regression (13) we do not measure the influence of the financial condion of the parent bank, but explicly test for substution effects. We include variables related to the (weighted) average risk/return characteristics of the other subsidiaries of the same multinational bank holding. In doing so, we explicly estimate the trade-off that parent banks face when deciding what parts of their international cred portfolio to expand or to reduce. For similar reasons, we also include the average (weighted) GDP growth in other countries where the multinational bank operates: L = α l m 1 + γ1 L 1 + βkhostk, + βksubk, + ε (11) k = 1 k = l+ 1 L = n p q 2 + γ 2 L 1 + βkhostk, + βksubk, + βkparentk, + ε (12) k = m+ 1 k = n+ 1 k = p+ 1 α L = α r s u v 3 + γ 3 L 1 + βkhostk, + βksubk, + βkothcok, + βk k = q+ 1 k = r+ 1 k = s+ 1 k = u+ 1 HOME k, + w k = v+ 1 β kothsu k, + ε (13) where?l (?L -1 ) is the percentage cred growth of subsidiary i in year t (t-1 if one period lagged) a 1, a 2 and a 3 are intercept terms;? 1,? 2 and? 3 are coefficients and ß k are coefficient vectors HOST is a matrix of host-country macroeconomic variables SUB is a matrix of characteristics related to bank subsidiary i PARENT is a matrix of characteristics related to the parent bank holding of subsidiary i HOME is a matrix of home-country macroeconomic variables OTHCO is a matrix of (weighted) macroeconomic variables related to the other countries where the bank holding operates (excluding the home country) OTHSU is a matrix of (weighted) characteristics related to other subsidiaries of the parent bank 2 ε is the idiosyncratic error, ε IID ( σ ) ~ 0, ε i=1,..., N where N is the number of bank subsidiaries in the sample t=1,..., T i where T i is the number of years in the sample for bank subsidiary i HOST (HOME) includes a number of host-country (home-country) macroeconomic variables that are likely to influence the (relative) attractiveness in terms of risks and returns of expanding cred in a 8

12 particular host country (home country). These include GDP growth (+), the unemployment rate (-), the average lending rate (+), inflation (-) and the government budget balance as percentage of GDP (+) (expected sign in parentheses). 13 Banks are likely to expand lending if GDP growth increases, unemployment is low, lending rates are higher, inflation is lower and government finances are sound. We also include crisis dummies, HOSTCRIS (HOMECRIS), which take on the value of 1 in case the host (home) country experienced a banking crisis in a particular year. 14 We expect bank cred to decline during banking system stress, unless the subsidiary is supported by s parent bank. Similarly, OTHCO includes variables measuring the (weighted) macroeconomic development in the other countries of operation of the parent bank. These variables thus reflect the (relative) attractiveness of expanding cred in other countries and proxy for the opportuny costs of expanding cred in a particular host country. For each year, we weigh the values for each country where a significant subsidiary is present wh the size of the particular subsidiary relative to the sum of all subsidiaries of the same parent in our sample. For instance, in the earlier example of HSBC Mexico, we construct an OTHCO GDP-growth variable which measures the weighted average growth rate of all countries where HSBC operates except Mexico and Great Brain. If the HSBC subsidiary in Canada would be twice as large as the one in Ireland, Canadian GDP growth would count twice as much as Irish GDP growth when constructing the OTHCO GDP-growth measure. 15 Finally, the OTHCRIS dummy is 0 in all years except for those years in which there was a banking crisis in one or more other countries of the same banking group. 16 As for bank-specific risk and return variables, SUB includes the following characteristics of each subsidiary (expected signs in parentheses). First, we include solvency (total equy to total assets (±)) and liquidy (total liquid assets to total assets (±)) as measures of the bank s risk aversion and capal/liquidy constrainedness of the bank. On the one hand, high capal and liquidy ratios may signal that a bank is relatively risk-averse and we thus expect a negative effect on loan growth. On the other hand, they may signal that liabily constraints are less important, so that banks have ample room to let their lending expand. The a priori sign of these variables is thus 13 To the extent that host-country inflation leads to a significant increase in the nominal value of loan portfolios there would be a posive effect of inflation on (nominal) cred growth. However, as we redenominated our data to US dollars inflationary effects should disappear to the extent that PPP holds. 14 Information for constructing this crisis dummy is taken from Caprio and Klingebiel (2003) and Carstens et al. (2004). For the year 2004, we constructed the dummy ourselves on the basis of several publications in print and on the internet. 15 In case of more than one subsidiary in a host country, we calculate the OTHCO macroeconomic variables on the basis of the other host countries only. For instance, when constructing the OTHCO GDP-growth variable for one of several Mexican subsidiaries of the same parent bank, we do not include Mexican GDP growth in the weighted OTHCO GDP-variable, since this, by construction, would lead to multicollineary between this variable and the Mexican GDP-growth variable (included as a separate determinant). 16 This dummy is constructed by averaging for each year the 0 and 1 scores for the other countries were the parent bank has subsidiaries (weighted wh the size of the respective subsidiaries). The OTHCRIS dummy is then assigned value 1 if the weighted value equals or exceeds 0.25 and a value of 0 otherwise. 9

13 indeterminate. Second, we include loan loss provisions to net interest revenue (-) as a proxy for the general financial condion of the bank as well as s willingness to take on risk. 17 An increase reflects that higher cred risk is only partially compensated for by higher interest margins and we thus expect a negative effect on loan growth. Third, we include profabily (return on equy (+)) and efficiency (net interest margin (-)) as (net) return indicators. Finally, PARENT includes the same bank-specific variables for the parent bank of the particular subsidiary, whereas OTHSUB consists of (weighted) variables for the other subsidiaries of the bank holding (including those in the home country). 18 To take into account that not all (types of) subsidiaries need to be equally sensive to parent bank policies, we make an a priori distinction between relatively dependent and relatively independent subsidiaries by constructing an ownership dummy that is 1 for all greenfields and that is 0 for all take-overs. We expect that, on average, greenfields are more strongly integrated in the multinational bank holding than acquired banks, at least for a number of years. In a separate set of estimations of equations (10) to (12), we use this dummy to construct interaction terms wh the variables that measure the characteristics of the parent bank and the other subsidiaries in the same bank holding. In this way we can explicly test whether any substution and support effects are stronger for greenfields than for acquired banks. For all regressions, we use two estimation methodologies: fixed effects and a dynamic GMM panel-data estimator (Hansen, 1982; Arellano and Bond, 1991). The choice for fixed effects estimations is based on Hausman tests, which consistently point out that fixed effects are to be preferred to random effects as the determinants of cred growth are correlated wh the bankspecific effects. However, as shown by Nickell (1981), in a dynamic context, the lagged dependent variable in particular may depend on the panel-level effects, potentially leading to an inconsistent estimator when the time dimension is limed. To solve this potential inconsistency problem, we also estimated all empirical models using a GMM framework. Two addional advantages of using GMM are that does not require distributional assumptions, such as normaly, and that can allow for heteroskedasticy of unknown form. We use one-step estimators of the robust variety, in order to correct for heteroskedasticy (based on the Sargent test statistics). 17 Peek and Rosengren (2000b) find that non-performing loans of parent banks may have an even more significant negative impact on host-country lending than parent banks capalisation. 18 Note that in this case, contrary to the macroeconomic OTHCO variables, we also include information on other subsidiaries in the same host-country as well as information on any separate subsidiaries in the home country (cf. footnote 15). 10

14 4. Data and descriptive statistics Our sample of multinational banks is based on the Top 1000 list of the world s largest banks (asset rank) as published by The Banker in July Of the 150 largest banks, we included all banks that operated more than one significant foreign subsidiary. For each of the 45 remaining bank holdings, we identified on the basis of information taken from BankScope, from banks webses and based on correspondence wh banks all domestic and foreign subsidiaries of which (i) the assets accounted for 0.5 per cent or more of the assets of the parent bank in 2004 and that (ii) were at least for 50% owned by the parent bank. 19 We thus focus on relatively large subsidiaries in which the parent bank has a controlling stake. 20 If parent banks were the result of a merger or acquision in year t we only included them from year t+1 onwards. Since we also only include banks for which we have at least three consecutive years of data, this meant that we had to exclude all Chinese and most Japanese banks. For each subsidiary, we traced back in which year t became part of the bank holding. In case of new establishments greenfields we included data for this subsidiary right from year t onwards. Subsidiaries that resulted from a take-over, are included from t A list of all bank holdings included in the sample can be found in Annex 1. As a double check on the qualy of our bank database, we sent a letter to each parent bank in which we asked to confirm that the subsidiaries we had identified were indeed those considered as material by the parent bank self. We also asked for the dates at which non-greenfield subsidiaries were acquired. In a typical year, each parent bank in our sample is associated wh on average 4.3 subsidiaries. The geographical distribution of parent banks (subsidiaries) used in the final analysis is as follows: 83% (73%) in Europe; 13% (15%) in North-America; 0% (2%) in South-America; 0% (0%) in Africa; 1% (8%) in Asia; and 3% (2%) in Australia. A graphical representation of the location of the banks in our dataset for 2004 is shown in Figure 1. Here we have plotted the parent (subsidiary) banks wh a triangle (circle). It is clear that most of the banks in our sample are in Europe and that European parent banks are also the most internationally oriented when comes to subsidiary location. 19 For each significant subsidiary ( level 1 ) we also checked whether owned significant sub-subsidiaries ( level 2 ) that were larger than 0.5% of the ultimate bank holding ( level 0 ). If this was not the case, we included consolidated data for the level 1 subsidiaries. If this was the case, we included unconsolidated data for the level 1 subsidiary and separately included consolidated data for the sub-subsidiary. 20 We only included commercial banks, savings banks, co-operative banks, real estate/mortgage banks and medium and long-term cred banks. We excluded investment banks, securies houses, government-owned banks or non-banking cred instutions. 21 We do not include any subsidiaries based in Luxembourg or Swzerland as in these countries, multinational bank subsidiaries are typically private banks wh activies that are mainly driven by the supply of deposs by (foreign) residents, rather than driven by eher local macroeconomic developments or by parent banks steering and capal support. 11

15 Figure 1 Location of parent banks (triangles) and subsidiaries (circles) (2004) Colour code parent banks: Over time both the number and size of the banks included in our dataset has grown, as shown in Figures 2 and 3. Figure 2 illustrates that internationally active banks have become more international, measured in terms of the number of foreign subsidiaries, especially though takeovers. In terms of total assets, we see in Figure 3 that growth has been even faster. The relative proportions of the assets of take-overs, greenfields and the parent bank are stable over time wh on average shares of 4%, 20% and 76% of total assets in our sample, respectively. Figure 2 Number of banks ( ) Figure 3 Total assets ( ) No. banks Trillion $ Parent Takeover Greenfield Parent Takeover Greenfield Source: BankScope Source: BankScope We obtained financial data on all parent banks and subsidiaries from Bureau van Dijk s BankScope database. Our sample period is , but the panel is unbalanced as we do not have data for all years for each bank. We corrected for the fact that not all banks report in the same currency by redenominating all balance-sheet and income-statement variables in millions of US dollars. 12

16 As explained in Section 3, our dependent variable is the percentage growth of gross loans. In order to construct this gross cred growth variable, we add loan loss reserves to net loans. In doing so, we correct for changes in (net) loans that are not due to changes in banks output of new loans, but that are related to changes in banks loan loss provisioning. We checked for outliers and removed all observations wh implausible values for one or more variables. In order to control for mergers and acquisions, we also removed all observations where the absolute growth of gross loans exceeded 75%. All in all, this data cleansing reduced our number of observations by 16%. Table 1 summarises our bank-level data for both parent banks and subsidiaries. Subsidiaries have on average a higher solvency, expressed as total equy to total assets, than parent banks. At the same time, the subsidiaries in our sample expand their lending somewhat faster than their parent banks, which may contribute to their on average lower profabily (return on equy) and the higher ratio between loan loss provisions and net interest revenue. The latter finding may indicate a greater willingness to take on risk at the subsidiary level. In a typical year, the assets of the average subsidiary account for about 11% of the assets of the average parent bank. Table 1 Summary statistics parent banks and subsidiaries Solvency Liquidy Efficiency Profabily Weakness Loan growth Total assets (in %) (in %) (in %) (in %) (in %) (in %) (mil. $) Parent banks Subsidiaries Finally, our macroeconomic data was taken from the IMF International Financial Statistics (IFS), the Transion Reports of the European Bank for Reconstruction and Development (EBRD), the IMF World Economic Outlook and Datastream. 5. Empirical results 5.1 Basic empirical results We first estimated a basic model in which multinational bank subsidiaries cred growth only depends on lagged cred growth, characteristics of the subsidiary self and some host-country characteristics (Table 2, columns 1 and 2). Experimenting wh various macroeconomic variables lead us to only include GDP growth, the unemployment rate and the crisis dummies, as these turned out to be relatively robust macroeconomic determinants of cred growth Contrary to what one might expect, economic growth and the unemployment rate are not very strongly correlated over time and across countries in our sample. 13

17 Table 2 Basic estimations Dependent: Cred growth (%) Fixed GMM Fixed GMM Fixed GMM Effects Effects Effects (1) (2) (3) (4) (5) (6) Lagged cred growth (0.07)* (0.01)** (0.36) (0.09)* (0.09)* (0.16) Weakness (0.95) (0.73) (0.81) (0.63) (0.97) (0.84) Profabily (0.00)*** (0.08)* (0.00)*** (0.05)** (0.20) (0.92) Liquidy (0.00)*** (0.00)*** (0.00)*** (0.01)*** (0.01)** (0.06)* Solvency (0.00)*** (0.01)** (0.00)*** (0.03)** (0.01)*** (0.09)* GDP gr. host country (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)*** Unemploy. host coun (0.02)** (0.38) (0.01)** (0.15) (0.00)*** (0.06)* Crisis dum. host coun (0.63) (1.00) (0.99) (0.91) (0.72) (0.54) Weakness parent bank (0.41) (0.14) Profab. parent bank (0.19) (0.29) Liquidy parent bank (0.00)*** (0.05)** Efficiency parent bank (0.00)*** (0.00)*** GDP gr. home country (0.00)*** (0.01)*** Weakness other subs (0.37) (0.15) Profabily other subs (0.08)* (0.00)*** Liquidy other subs (0.35) (0.14) Solvency other subs (0.93) (0.14) Crisis other coun (0.21) (0.16) Constant (0.00)*** (0.11) (0.00)*** (0.33) (0.00)*** (0.12) Observations Number of banks R-squared Hausman & AB test AR AB test AR Wald ~ p values in parentheses 14

18 * significant at 10%; ** significant at 5%; *** significant at 1% Arellano-Bond (1991) dynamic panel-data estimator (one-step, robust) & P value of Hausman specification test P value of Arellano-Bond test that average autocovariance in residuals of order 1 is 0 P value of Arellano-Bond test that average autocovariance in residuals of order 2 is 0 ~ P value of Wald test that parameters associated wh parent bank/other subsidiaries variables are jointly 0 We find that profable subsidiaries expand their cred faster and that bank subsidiaries that are eher relatively solvent or relatively liquid show lower rates of cred expansion. Our conjecture is that relatively solvent banks show a higher risk aversion, whereas relatively liquid banks grow more slowly because they invest mainly in liquid assets. 23 Subsidiary banks also grow faster when economic growth is higher and unemployment is lower in the host country. Finally, we find that subsidiaries of multinational banks do not reduce their cred supply when the host country is h by a systemic banking crisis. This last finding may point to parent banks giving financial support to their subsidiaries when the latter are confronted wh adverse financial condions. However, this result may also simply reflect that subsidiaries themselves are on average strong enough to whstand host-country crisis periods. Columns 3 and 4 show similar estimations, but now we also add characteristics of the parent bank and the home country. Our earlier results do not change: subsidiaries of multinational banks grow faster when host-country economic growth is high and when they are relatively profable. In contrast, subsidiaries that aim for high levels of solvency and/or liquidy generally expand their cred less swiftly. Interestingly, this expanded empirical model also shows that the balance-sheet structure of the parent bank matters for subsidiary growth. In particular, subsidiaries of more efficient parent banks tend to grow faster. 24 On the contrary, subsidiaries of parent banks that are highly liquid tend to grow slower, which is in line wh our result for the liquidy of the subsidiaries themselves. These last two findings are in line wh support effects. Finally, we estimate a set of regressions in which we include (weighted) variables related to the other subsidiaries of the same multinational bank holding (columns 5 and 6). 25 Again, we replicate our finding that subsidiaries cred supply is posively related to the host-country business cycle and negatively to their own liquidy and solvency. However, profabily is no 23 Reverse causaly could have been an issue here: banks that grow faster will become less liquid and less solvent if then cannot replenish their liquid funds and capal fast enough. However, note that GMM is an instrumental variables technique (where the instruments are constuted by lagged values of the variables themselves) and as such controls for any reverse causaly that may arise. 24 Because of the high (0.8) correlation between parent-bank efficiency and parent-bank solvency we only include parent-bank efficiency. 25 Because of the high (0.6) correlation between other subsidiaries efficiency and other subsidiaries solvency we only include other subsidiaries solvency. Also, we include home country GDP growth but not the weighted GDP growth in the other countries of operation as both economic growth rates tend to be highly correlated as well. 15

19 longer a significant determinant of cred growth. 26 Instead, we find that when other subsidiaries in the same group are relatively profable, this posively influences the cred supply of a subsidiary. Note that in case substution effects would dominate, we would expect a negative relationship between other subsidiaries profabily and cred growth. Apparently, support effects are more important here: wh (other) profable subsidiaries, parent banks are able to give more capal support to a given subsidiary. This support effect also seems to dominate the influence of subsidiaries own profabily. We also find that GDP growth in the home country exerts a negative influence on subsidiaries cred growth. Apparently, parent banks trade expanding cred growth in the home against growth in the host country (substution effect). A drawback of our approach is that our sample does not contain a control group of domestic banks. We can thus not determine whether the significant negative sign for the influence of home country GDP only applies to the foreign subsidiaries. Other, domestic banks might be equally sensive to changes in the GDP of home countries. In particular, if banks expand abroad into neighbouring countries, is likely that GDP movements in the home country will also be relevant for the domestic banks in the host country. To overcome this issue, we would need to extend our dataset to also include all other banks in the respective host countries. This is, however, given the data availabily, too ambious. Still, domestic banks are unlikely to be influenced by developments in the home countries of their foreign competors if domestic banks and foreign banks supply cred do different market segments. In particular, local banks are likely to focus more on (smaller) local customers, whereas multinational bank subsidiaries often focus on large, multinational companies. 27 In addion, domestic banks will be relatively insensive to the economic circumstances in the home countries of their foreign competors if the geographical distance between the host and the home country is large, so that macroeconomic and trade linkages between both countries are relatively small. 5.2 Greenfield versus take-over multinational bank subsidiaries As explained in Section 2.1, we expect that subsidiaries that are the result of a take-over ( takeovers ) are relatively independent as they may not (yet) be fully integrated into the bank holding. In contrast, we expect that greenfield subsidiaries ( greenfields ) which have been erected from scratch by the parent bank self tend to be more closely integrated into the bank holding and s capal allocation mechanism. To test whether this is borne our by the data, we re-estimated the 26 Also note that the number of banks in these two sets of estimations is smaller because we removed all subsidiary banks that operate in the same country as their parent bank. For these subsidiaries, the home country and the host country coincide so that we had to exclude them from our sample in order to prevent any artificial multicollineary between the home country variables and the host country variables. 27 See, for instance, Sabi (1988) and Berger et al. (2001) for empirical evidence. 16

20 above empirical models, this time including interaction effects of the main determinants wh an ownership dummy. 28 The results can be found in Table 3. Table 3 Estimations wh interaction terms Fixed GMM Fixed GMM Fixed GMM Effects Effects Effects (1) (2) (3) (4) (5) (6) Lagged cred growth (0.07)* (0.03)** (0.33) (0.12) (0.09)* (0.17) Weakness (0.07)* (0.08)* (0.07)* (0.13) (0.14) (0.12) Weakness/greenfield interact. (0.06)* (0.07)* (0.06)* (0.12) (0.07)* (0.10)* Profabily (0.01)*** (0.07)* (0.00)*** (0.05)** (0.20) (0.74) Liquidy (0.00)*** (0.00)*** (0.00)*** (0.01)*** (0.01)** (0.05)** Solvency (0.00)*** (0.01)*** (0.00)*** (0.01)*** (0.01)*** (0.05)** GDP gr. host country GDP/greenfield interaction (0.55) (0.65) (0.46) (0.48) (0.96) (0.81) (0.00)*** (0.01)*** (0.00)*** (0.00)*** (0.01)*** (0.00)*** Crisis dum.host coun (0.57) (0.96) (0.96) (0.79) (0.76) (0.51) Unemploy. host coun (0.85) (0.37) (0.70) (0.99) (0.20) (0.40) Unemployment/ greenfield interaction (0.38) (0.17) (0.43) (0.47) (0.99) (0.78) Weakness parent bank (0.45) (0.13) Profabily parent (0.35) (0.42) Liquidy parent bank (0.00)*** (0.04)** Efficiency parent bank (0.00)*** (0.00)*** GDP gr. home country (0.00)*** (0.00)*** Weakness other subs (0.32) (0.14) Profabily other subs (0.05)** (0.00)*** 28 We interacted all determinants wh an ownership dummy which is 1 for greenfields and 0 for takeovers. We wanted to keep the empirical models as parsimonious as possible and eliminated the most insignificant interaction terms, starting wh the least significant ones. This was necessary since including too many interaction effects would lead to an overly large instrument matrix in the GMM estimations. 17

21 Liquidy other subs (0.41) (0.17) Solvency other subs (0.82) (0.13) Crisis other countries (0.18) (0.23) Constant (0.00)*** (0.10)* (0.00)*** (0.31) (0.00)*** (0.09)* Observations Number of banks R-squared Hausman & AB AR AB AR Wald ~ p values in parentheses * significant at 10%; ** significant at 5%; *** significant at 1% Arellano-Bond (1991) dynamic panel-data estimator (one-step, robust) & P value of Hausman specification test P value of Arellano-Bond test that average autocovariance in residuals of order 1 is 0 P value of Arellano-Bond test that average autocovariance in residuals of order 2 is 0 ~ P value of Wald test that parameters associated wh parent bank/other subsidiaries variables are jointly 0 Ownership is a dummy which takes on the values of 0 (1) for takeovers (greenfields) As before, we find that lending by multinational bank subsidiaries is sensive to their own profabily (+), liquidy (-) and solvency (-). In case other subsidiaries in the same group are relatively profable, this also posively influences the cred supply of a subsidiary. We also confirm our earlier finding that subsidiaries of more efficient parent banks tend to grow faster, whereas subsidiaries of parent banks that are highly liquid tend to grow slower. On the basis of the interaction terms, we also find weak evidence (10% confidence level) that shows that while the lending of take-over subsidiaries is sensive to the weakness of their balance sheet, this does not seem to hold for greenfield subsidiaries. This is in line wh a stronger role of the parent bank in the case of greenfields compared to take-overs. Second, we find that hostcountry GDP growth only seems to matter for greenfield foreign banks. The cred supply of takeovers is much less procyclical, perhaps because these banks have not yet been well-integrated into the parent bank and substution effects are thus less important. Home-country GDP-growth has a negative influence on the cred growth of both types of multinational bank subsidiaries. 6. Summary and conclusions We estimate dynamic cred growth regressions to explore how the cred growth of multinational bank subsidiaries is influenced by home- and host-country macroeconomic condions as well as by the characteristics of the subsidiary self and s parent bank. We find that multinational bank subsidiaries are not fully stand-alone operations, but are influenced by developments in the rest of 18

22 the multinational banking group and in other countries of operation. Our findings provide strong evidence for the hypothesis that multinational banks manage the cred growth of their subsidiaries through the operation of an internal capal market. In particular, we find that subsidiaries that aim for high levels of solvency and/or liquidy generally expand their cred less swiftly, whereas profable subsidiaries are able to expand cred faster. However, subsidiaries are not only affected by their own balance sheets, but also by the developments in other parts of the multinational banking group to which they belong. In line wh substution effects, we find that subsidiaries expand their lending faster if economic growth in their home country decreases. Apparently, parent banks trade off expanding cred growth in the home versus the host country. Likewise, we find that higher host-country growth posively influences cred growth, but here the substution effect is limed to greenfield subsidiaries. We also find evidence for support effects. In particular, subsidiaries of more efficient parent banks tend to grow faster, while subsidiaries of highly liquid parent banks tend to grow more slowly. In case other subsidiaries in the same group are relatively profable, this posively influences the cred supply of a subsidiary: wh (other) profable subsidiaries parent banks are able to give more capal support. We also find that subsidiaries of multinational banks do not tend to reduce their cred supply when the host country is h by a banking crisis. Finally, we find that whereas the cred growth of (strongly integrated) greenfield subsidiaries is not sensive to their own balance-sheet strength, this does not hold for the (more independent) take-overs. Apparently, the latter group can rely less on parental support and is thus forced to slow down their lending if their balance sheet gets weaker. References Arellano, M. and S. Bond (1991), Some tests of the specification for panel data: Monte Carlo evidence and an application to employment equations, The Review of Economic Studies, 58, Barajas, A. and R. Steiner (2002), Cred stagnation in Latin America, IMF Working Paper No. 53, International Monetary Fund, Washington D.C. Berger, A.N., L.F. Klapper and G.F. Udell (2001), The abily of banks to lend to informationally opaque small businesses, Journal of Banking & Finance, 25, Buch, C.M. (2000), Why do banks go abroad? Evidence from German data, Financial Markets, Instutions & Instruments, 9(1), Calvo, G.A. and F. Coricelli (1993), Output collapse in Eastern Europe. The role of cred, IMF Staff Reports, 40,

23 Caprio, G. and D. Klingebiel (2002), Episodes of systemic and borderline banking crises, World Bank, Washington D.C. Carstens, A.G., D.C. Hardy and C. Pazarbasioglu (2004), Avoiding banking crises in Latin America, Finance & Development, September, Crystal, J.S., B.G. Dages and L.S. Goldberg (2002), Has foreign bank entry led to sounder banks in Latin America?, Current Issues in Economics and Finance, 8(1), 1-6. Dahl, D. and R.E. Shrieves (1999), The extension of international cred by US banks: A disaggregated analysis, , Journal of International Money and Finance, 18, Dages, B.G., L.S. Goldberg and D. Kinney (2000), Foreign and domestic bank participation in emerging markets: Lessons from Mexico and Argentina, Federal Reserve Bank of New York Economic Policy Review, September, De Haas, R.T.A. and I.P.P. van Lelyveld (2004), Foreign bank penetration and private sector cred in Central and Eastern Europe, Journal of Emerging Market Finance, 3(2), De Haas, R.T.A. and I.P.P. van Lelyveld (2006), Foreign banks and cred stabily in Central and Eastern Europe. A panel data analysis, Journal of Banking & Finance, forthcoming. Froot, K.A. and J.C. Stein (1998), Risk management, capal budgeting, and capal structure policy for financial instutions: An integrated approach, Journal of Financial Economics, 47, Goldberg, L.S. (2001), When is U.S. bank lending to emerging markets volatile?, NBER Working Paper No. 8209, National Bureau of Economic Research, Cambridge, MA. Hansen, L.P. (1982), Large sample properties of generalized method of moments estimators, Econometrica, 50, Hernandez, L. and H. Rudolph (1995), Sustainabily of private capal flows to developing countries is a generalised reversal likely?, World Bank Policy Research Working Paper No. 1518, World Bank, Washington D.C. Holmström, B. and J. Tirole (1997), Financial intermediation, loanable funds, and the real sector, Quarterly Journal of Economics, CXII, Houston, J.F., C. James and D. Marcus (1997), Capal market frictions and the role of internal capal markets in banking, Journal of Financial Economics, 46, Jeanneau, S. and M. Micu (2002), Determinants of international bank lending to emerging market countries, BIS Working Papers No. 112, Bank for International Settlements, Basel. Martinez Peria, S., A. Powell and I. Vladkova Hollar (2002), Banking on foreigners: The behavior of international bank lending to Latin America, , World Bank Working Paper No. 2893, World Bank, Washington D.C. Molyneux, P., E. Remolona and R. Seth (1998), Modelling foreign bank performance and lending behaviour, Financial Markets, Instutions & Instruments, 7(4),

24 Morgan, D., B. Rime and P.E. Strahan (2004), Bank integration and state business volatily, Quarterly Journal of Economics, 119(4), Morgan, D. and P.E. Strahan (2004), Foreign bank entry and business volatily: Evidence from U.S. states and other countries, In: Ahumada, L.A. and J.R. Fuentes (eds.), Banking Market Structure and Monetary Policy, , Central Bank of Chile, Santiago. Nickell, S. (1981), Biases in dynamic models wh fixed effects, Econometrica, 49(6), Peek, J. and E.S. Rosengren (1997), The international transmission of financial shocks: The case of Japan, American Economic Review, 87, Peek, J. and E.S. Rosengren (2000a), Implications of the globalization of the banking sector: The Latin American experience, New England Economic Review, September/October, Peek, J. and E.S. Rosengren (2000b), Collateral damage: Effects of the Japanese bank crisis on the Uned States, American Economic Review, 90(1), Sabi, M. (1988), An application of the theory of foreign direct investment to multinational banking in LDCs, Journal of International Business Studies, fall, Stein, J.C. (1997), Internal capal markets and the competion for corporate resources, Journal of Finance, 52, Van den Heuvel, S. (2002), Does bank capal matter for monetary transmission?, FRBNY Economic Review, May, Van Rijckeghem, C. and B. Weder (2000), Spillovers through banking centers: A panel data analysis, IMF Working Paper No. 88, International Monetary Fund, Washington D.C. Van Rijckeghem, C. and B. Weder (2001), Sources of contagion: Finance or trade?, Journal of International Economics, 54,

25 Annex 1 List of multinational banks in our sample Name bank holding Home country Number of Host countries subsidiaries 1 ABN Amro Holding The Netherlands 5 AU, BR, FR, NL, US 2 Allied Irish Banks Ireland 2 PL, GB 3 Banca Intesa Italy 11 HR, FR, HU, IE, IT, PE, SK 4 Banco Bilbao Vizcaya Argentaria Spain 9 AR, CL, CO, MX, PE, PT, ES, US, VE 5 Banco Comercial Português Portugal 2 MZ, PL 6 Banco Popular Español Spain 8 FR, ES 7 Banco Santander Central Hispano Spain 11 BR, CL, DE, MX, ES, US, VE 8 Bank of America USA 2 HK, US 9 Bank of Ireland Ireland 1 GB 10 Bank of Montreal Canada 2 IE, US 11 Barclays Bank Great Brain 1 ES 12 Bayerische Hypo- und Vereinsbank Germany 9 CZ, HU, PL, RU, AT 13 BNP Paribas France 4 FR, US 14 Caja de Ahor. y Pens. de Barcelona Spain 2 FR, ES 15 Cicorp USA 8 BR, CA, MY, MX, PL, US 16 Commerzbank Germany 4 DE, NL, PL 17 Danske Bank Denmark 1 NO 18 Deutsche Bank Germany 6 DE, IT, ES, US 19 Dexia Belgium 7 BE, FR, DE, IT, NL, ES 20 Dresdner Bank Germany 1 DE 21 DZ Bank Germany 4 DE, IE 22 Erste Bank Austria 8 CZ, HU, AT, SK 23 FöreningsSparbanken (Swedbank) Sweden 5 EE, LV, LT, SE 24 Fortis Bank Belgium 3 BE, FR, NL 25 Gruppo Monte dei Paschi di Siena Italy 3 FR, IT 26 HBOS Great Brain 4 IE, GB 27 HSBC Holdings Great Brain 12 BR, CA, FR, DE, HK, IND, MY, SA, GB, US 28 ING Bank The Netherlands 5 BE, CA, FR, NL, PL 29 KBC Bank Belgium 7 BE, CZ, DE, HU, IE, PL 30 MBNA Corp Uned States 3 CA, GB, US 31 Msubishi Tokyo Financial Group Japan 4 JP, US 32 National Australia Bank Australia 2 NZ, GB 33 National Bank of Greece Greece 4 BG, CA, CY, US 22

26 34 Nordea Bank Sweden 6 DK, FI, NO, SE 35 Rabobank Group The Netherlands 1 IE 36 Raiffeisen Zentralbank Oesterreich Austria 8 BG, HR, CZ, PL, RU, SI, UA, SK 37 Royal Bank of Canada Canada 2 GB, US 38 Royal Bank of Scotland Group Great Brain 5 GB, US 39 Scotiabank Canada 4 SV, MX, GB, US 40 SEB Sweden 5 EE, DE, LV, LT, SE 41 Standard Chartered Great Brain 4 KE, MY, TH, GB 42 Société Générale France 7 AU, CA, CZ, FR, DE 43 UBS Swzerland 1 GB 44 UniCredo Italiano Italy 5 BG, IE, IT, PL 45 WGZ Bank Germany 2 DE, IE 23

27 Previous DNB Working Papers in 2006 No. 81 Arthur van Soest, Arie Kapteyn and Julie Zissimopoulos, Using Stated Preferences Data to Analyze Preferences for Full and Partial Retirement No. 82 Dirk Broeders, Valuation of Condional Pension Liabilies and Guarantees under Sponsor Vulnerabily No. 83 Dirk Brounen, Peter Neuteboom and Arjen van Dijkhuizen, House Prices and Affordabily A First and Second Look Across Countries No. 84 Edwin Lambregts and Daniël Ottens, The Roots of Banking Crises in Emerging Market Economies: a panel data approach No. 85 Petra Geraats, Sylvester Eijffinger and Carin van der Cruijsen, Does Central Bank Transparency Reduce Interest Rates? No. 86 Jacob Bikker and Peter Vlaar, Condional indexation in defined benef pension plans No. 87 Allard Bruinshoofd and Clemens Kool, Non-linear target adjustment in corporate liquidy management: an endogenous thresholds approach No. 88 Ralph de Haas, Monoring Costs and Multinational-Bank Lending No. 89 Vasso Ioannidou and Jan de Dreu, The Impact of Explic Depos Insurance on Market Discipline No. 90 Robert Paul Berben, Kerstin Bernoth and Mauro Mastrogiacomo, Households Response to Wealth Changes: Do Gains or Losses make a Difference? No. 91 Anne Sibert, Central Banking by Commtee No. 92 Alan Blinder, Monetary Policy by Commtee: Why and How? No. 93 Céline Christensen, Peter van Els and Maarten van Rooij, Dutch households perceptions of economic growth and inflation No. 94 Ellen Meade, Dissents and Disagreement on the Fed s FOMC: Understanding Regional Affiliations and Lims to Transparency No. 95 Jacob Bikker, Laura Spierdijk, Roy Hoevenaars and Pieter Jelle van der Sluis, Forecasting Market Impact Costs and Identifying Expensive Trades No. 96 Cees Ullersma, Jan Marc Berk and Bryan Chapple, Money Rules No. 97 Jan Willem van den End, Indicator and boundaries of financial stabily No. 98 Zsolt Darvas, Gábor Rappai and Zoltán Schepp, Uncovering Yield Pary: A New Insight into the UIP Puzzle through the Stationary of Long Matury Forward Rates No. 99 Elisabeth Ledrut, A tale of the water-supplying plumber: intraday liquidy provision in payment systems No. 100 Ard den Reijer, The Dutch business cycle: which indicators should we monor?

28