Competition in Hong Kong s banking industry

Transcription

1 Lngnan Journal of Bankng, Fnance and Economcs Volume /2013 Academc Year Issue Artcle 6 January 2013 Competton n Hong Kong s bankng ndustry La Yee CHU Yue CUI Nan YE Yueln YAN Follow ths and addtonal works at: Part of the Fnance Commons, and the Fnance and Fnancal Management Commons Recommended Ctaton Chu, L. Y., Cu, Y., Ye, N., & Yan, Y. (2013). Competton n Hong Kong s bankng ndustry. Lngnan Journal of Bankng, Fnance and Economcs, 4. Retreved from Ths Artcle s brought to you for free and open access by the Department of Economcs at Dgtal Lngnan Unversty. It has been accepted for ncluson n Lngnan Journal of Bankng, Fnance and Economcs by an authorzed edtor of Dgtal Lngnan Unversty.

2 CHU et al.: Competton n Hong Kong s bankng ndustry Competton n Hong Kong s Bankng Industry La Yee CHU, Yue CUI, Nan YE and Yueln YAN Abstract Ths paper tests the competton structure of Hong Kong s bankng ndustry usng the Panzar-Rosse approach and a panel dataset of the largest 20 banks n Hong Kong from 1998 to The estmaton results showed that the compettve pressures were equal across tme and across the dfferent szes of the banks n Hong Kong s bankng ndustry. Compettve pressures may heghten n the near future as banks wll seek to gan compettve edge and they can do so through mergers and acqustons. Regulatons are also slowly beng relaxed after the fnancal crss n order to boost the economy and more advances n technology are also antcpated. Key words: Competton Bankng Industry Panzar-Rosse Approach 77 Publshed by Dgtal Lngnan Unversty,

3 Lngnan Journal of Bankng, Fnance and Economcs, Vol. 4, Iss. 1 [2013], Art Introducton Hong Kong, as an nternatonal fnancal centre, has one of the most concentrated and compettve bankng ndustres n the world. Pror to the year 2000, the number of all authorzed nsttutons n Hong Kong s bankng ndustry was more than three hundred (the hghest number s 381 n 1995). But ths number keeps decreasng gradually n the recent years. At the end of 2012, the number of authorzed nsttutons n Hong Kong s bankng ndustry was just two hundred. The followng fgure descrbes the average number of all authorzed nsttutons n Hong Kong s bankng sector each year. Fgure 1 Average number of all authorzed nsttutons So what are the mplcatons of the deceasng number of banks for the remanng banks? Are they ncreasng ther proft because of larger market shares? In ths paper, we use the Panzar-Rosse approach and a panel dataset of the largest 20 banks n Hong Kong from 1998 to 2011 to test whether competton has a postve effect on the profts of banks. In our analyss we use only 20 banks because out of the 37 ranked banks n Hong Kong, more than 10 were formed after the year 2000 and the data of some banks were not readly avalable. Wong et al. (2006) analyse the evoluton of compettve condtons of Hong Kong s bankng ndustry for the perod 1991 to 2005 by usng the Panzar-Rosse assessment. Ths research clams that compettve pressure was hgher among larger banks and lower among smaller banks. Ths s already an mprovement on the study by Jang et al (2004) who use the aggregate data of the bankng sector to check the compettve condtons between

4 CHU et al.: Competton n Hong Kong s bankng ndustry In our study, we extend the perod from 2005 to 2011 to make a more accurate analyss of how competton affects proft. Of specal mportance s the year 2008 when there was the fnancal crss n the US. At the tme, Hong Kong s bankng ndustry came under a great deal of stran and makes 2008 partcularly mportant when analyzng whether competton has a postve or negatve effect on banks profts. The paper s organzed as follows: The lterature revew of how competton effect bankng sector s dscussed n part 2. Then we talk about Panzar-Rosse approach n part 3. Part 4 s the data descrpton and part 5 s dedcated to the emprcal model. The result and analyss are dscussed n part 6. The fnal part s the concluson. 2. Lterature Revew In 1987, Panzar and Rosse created a method to measure the compettve condtons n the bankng ndustry, whch s the H-statstc. Ths measure s based on the estmated effect of changng nput prces on revenue. The frst research applyng Panzar-Rosse approach was conducted by Shaffer n The author analyzed the bankng monopoly n New York and found that the value of H ranged from 0.32 to Jang et al (2004) appled Panzar-Rosse approach to Hong Kong s bankng sector and suggested that compettve pressures n the sector may have eased n the later years based on aggregate data of the bankng ndustry. Jm Wong et al followed up on the study by Jang et al n They separated the sample banks nto two groups and found that compettve pressure was hgher among larger banks and lower among smaller banks. Other researchers have also studed the effect of compettve condtons on profts of banks usng other approaches. John Boyd & De Ncolo (2006) stated that less compettve bankng systems are less fragle. Whle Frankln Allen & Douglas Gale (2003) argued competton leads to less fraglty. Uhde et al (2008) used data from 25 EU countres to prove that natonal bankng market competton has a negatve relatonshp wth the revenues of European banks. Berger et al (2009) also conducted a cross-country study and n ths paper they make great suggestons as to how control varables can be chosen. It s safe to say that across dfferent regons and tme, the relatonshp between competton and banks profts may dffer. We ntend to focus on the Hong Kong bankng ndustry and use the past 14 years data to draw conclusons. 79 Publshed by Dgtal Lngnan Unversty,

5 Lngnan Journal of Bankng, Fnance and Economcs, Vol. 4, Iss. 1 [2013], Art Theoretcal Model The assumpton of the model s that banks wll adopt dfferent prcng strateges accordng to dfferent market structures of nput costs. Through the analyss of a bank's proft and nput costs, one can determne whch market condtons a bank operates n. The Panzar - Rosse method was derved from a general fnancal market model, whch model determnes how sngle bank maxmses proft gven equlbrum output and equlbrum number of banks. When margnal cost equals margnal revenue, bank maxmze profts: ' ' R ( x, n, z ) C ( x, w, t ) 0 ' R represent bank s margn revenue; ' C represents bank s margn cost; x represents bank s output; n s the number of banks; w s a vector of unt prce; z s the bank s proft functon of exogenous varables and t s the bank s proft functon of exogenous varables. Next, we solve for equlbrum at the market level: * * R ( x, n, z) C * * ( x *, w, t) 0 * varables represent the equlbrum value. Panzar and Rosse use the H ndex to measure the market power, namely, market structure and competton. H ndex s measured by ncome elastcty of the nput prces: H R w m * k * k 1 wk R The H-statstc s calculated by summng the estmated elastcty of revenue to factor prces, wth a value of one ndcatng perfect competton, a value of zero (or less) ndcatng monopoly, and ntermedate values ndcatng the degree of monopolstc competton. Table 1 Meanng of H value Compettve structure Values of H Monopoly H 0 Monopolstc Competton 0<H<1 Perfect Competton H=1 4. Data Descrpton In the Panzar-Rosse framework approach, the H-statstc s the sum of total coeffcents of three major nputs whch affect the bank's total ncome

6 CHU et al.: Competton n Hong Kong s bankng ndustry Table 2 Varables, descrpton and data source Varables Descrpton and calculaton Data source Captal cost Unt prce of Captal - Other Bankscope database Operatng expenses/fxed Asset Labour cost Unt prce of Labour - Personnel Bankscope database expenses/total Asset Funds cost Unt prce of Funds Interest Bankscope database Expense/Depost (from customers) Rsk of Asset (Equty / Total Assets+ROA) / Bankscope database Standard Devaton of ROA Asset sze Level of Assets for the banks Bankscope database GDP level Level of GDP IMF Inflaton rate (nflaton) Inflaton rate IMF When t comes to the selecton of the three major varables, we choose PF (unt prce of funds), PL (unt prce of labor) and PK (unt prce of captal). Frst, PF, also called the cost of fundng rate, refers to the rato of nterest expense to total fundng. If we assume that the man source of fundng for banks s customer deposts, then the correspondng cost s nterest payment. Therefore, PK should equal total nterest expense dvded by total depost from customers. Second, PL, whch stands for cost of labor rate, s computed as the rato of staff expense to total asset (Bkker and Groeneveld 1998), (Gelos and Roldos 2002). Note that other measures of unt prce of labor as the rato of staff expense to the number of employees are also frequently used. Thrd, PK represents the cost of fxed asset. Generally speakng, the total expenses of a bank can be dvded nto two parts; one s total nterest expenses and the other s total non-nterest expenses. Total non-nterest expenses less personnel expenses s the value of general cost of fxed assets of a bank. Therefore, we can use equaton other operatng expenses dvded by fxed assets to calculate the cost of fxed asset. In order to study such problem more reasonably, we add another mportant factor, Z-score n our model. Z-score equals the sum of μ (the return on average assets before taxes, ROAA) and k (the equty captal as a percent of total assets) dvded by σ (the standard devaton of 81 Publshed by Dgtal Lngnan Unversty,

7 Lngnan Journal of Bankng, Fnance and Economcs, Vol. 4, Iss. 1 [2013], Art. 6 ROAA). Therefore, the Z-score s a combnaton factor that contans banks proftablty (μ), captal rato (k) and return volatlty (σ). Generally speakng, these 3 ndcators can comprehensvely reflect the operatng stuaton of a bank. To be precse, proftablty (μ) s an ndcator of the proftablty of a bank's assets. It s also used to evaluate a bank s performance. Captal rato (k) s a key fnancal rato measurng the fnancal stablty and captal adequacy of a bank. The hgher the rato, the more stable the bank. Compared to the low captal rato banks, banks wth hgh captal rato are better able to protect themselves aganst operatng losses. Return volatlty (σ), on the other hand, shows the extent of fluctuatons n ROAA. It s clear that the Z-score wll ncrease wth μ (the banks proftablty) and k (captal rato) and decrease wth ncreasng σ (return volatlty). From an economc vewpont, the Z-score measures the probablty of a bank gong nsolvent when the value of assets becomes lower than the value of debt. Hence, a hgher (lower) Z-score mples a lower (hgher) probablty of nsolvency rsk (Uhde and Hemeshoff, 2008). Macroeconomc control varables are also mportant n our model. We nclude the GDP growth rate and the annual nflaton rate (provded by the Internatonal Monetary Fund) to capture macroeconomc developments that are lkely to affect the qualty of banks assets. Frst, GDP growth rate reflects the economc performance of the country and closely related to the bankng ndustry. Hence, we expect a postve symbol of the coeffcent that the banks performances mprove n perods of economc prosperty. In addton, borrowers solvency should be hgher under ncreasng economc performance whch n sum rases banks asset qualty. Second, the nflaton rate s also mportant. Inflaton s an ncrease n general prce level and s typcally expressed as an annual percentage rate of change. Hgher nflaton can decrease the real rate of return on assets and then dscourage savng but encourage borrowng. 5. Emprcal Model For the purpose of analyss, the panel-data analyss technque s employed as t has the advantage of contanng the nformaton necessary to deal wth both the ntertemporal dynamcs and the ndvdualty of the enttes beng nvestgated (Avral and Mha 2011). There are bascally three types of panel-data models, namely, a pooled Ordnary Least Square (OLS) regresson, panel model wth random effects and panel model wth fxed effects. Usng the varables descrbed above, the equaton for the pooled OLS regresson can be specfed as follows:

8 CHU et al.: Competton n Hong Kong s bankng ndustry Where represents the bank, t represents tme and ε represents the error term whch s the whte nose and vares across banks and tme. However, usng the pooled OLS regresson technque wll not capture the countres unobservable ndvdual effects. Accordng to Bevan and Danbolt (2004), the nherent dfferences across the dfferent panels can nfluence measurements of the estmated parameters. Hence, we use a panel-data model wth fxed or random effects to help account for ndvdual peculartes. By takng nto account countres peculartes and varables used, the panel-data model wth fxed effects s to be estmated as follows: Where, wth representng the banks unobservable ndvdual effects. The panel-data model wth random effects s descrbed n the same way whlst, where wll have zero mean, ndependent of ndvdual observaton error term, has constant varances, and s ndependent of the explanatory varables. s the return on assets (Net Income/Asset Sze). 6. Estmaton Results We begn by reportng the results of the pooled OLS model. OLS s n fact the most restrctve of all models because t does take nto consderaton dfferences n cross-sectonal unts as t assumes a common ntercept for the whole panel. In order to assess whether the pooled OLS model s the correct model to be appled, we conducted the Wald test whose null hypothess s that all ndvdual effects are zero. The result we obtaned rejects the null hypothess and hence, the OLS estmator s based and nconsstent. Therefore, we resort to usng panel-data model wth fxed and random effects (see table below for results). Table 3 Regresson results: what affects the coeffcent on proftablty of a bank Varables Pooled OLS Fxed Effects Random Effects Captal cost (lnpk) *** *** ** Labour cost (dlnpl) Funds cost (dlnpf) ** ** ** 83 Publshed by Dgtal Lngnan Unversty,

9 Lngnan Journal of Bankng, Fnance and Economcs, Vol. 4, Iss. 1 [2013], Art. 6 Rsk of Asset (rskass) *** *** *** Rsk of Asset to the square *** *** *** (rskass*rskass) Asset sze (lnasset) *** *** *** GDP level (lngdp) * *** ** Inflaton rate (nflaton) * * * Constant ^ *** * Hausman Test Prob>ch2 = Wald Test Prob>f= R squared Adjusted R squared Countres 20 banks Observatons 280 Perod legend: ^p<0.10; * p<0.05; ** p<0.01; *** p<0.001 Note: The p-value for dlnpl s for Fxed Effects Despte the fact that the OLS estmators are based, the results are reported snce t does ndcate somehow the relatonshp (postve or negatve) between the ndependent varables and dependent varable. As for the fxed and random effects models, we need to evaluate whch of the two models s the more approprate to use snce they are nherently dfferent as they have dfferent assumptons, as descrbed above. The Hausman specfcaton test s used to choose the better model to be used. Accordng to the test performed, the null hypothess that the ndvdual effects are not correlated wth the other explanatory varables n the model (Hausman 1979) s rejected. Hence, the fxed effect model s more approprate one to employ, as shown above. The above results show that captal costs, labour costs and funds costs have a postve relatonshp wth the proftablty of the banks. Captal costs and funds costs are sgnfcant at 0.1% and 1% level respectvely. Labour costs are sgnfcant at a 15% level. The rsk of asset exerts a postve effect on the proftablty of a bank meanng that the more rsks a bank takes, the more profts are expected. The above results also show that beyond a certan level of rsk, the bank wll begn to ncur losses snce the rsk of the asset to the square reports a negatve fgure and ths s sgnfcant at a 0.1% level. The above results also show that as the level of asset rses, the profts wll be reduced and ths s statstcally sgnfcant. The macro economc varables are also sgnfcant n determnng the level of profts n the banks: the level of GDP

10 CHU et al.: Competton n Hong Kong s bankng ndustry exerts a postve effect on proftablty whereas nflaton rate exerts a negatve effect, as would be expected snce hgher level of GDP would ndcate economc growth and hence banks should eventually experence hgher profts whereas nflaton erodes profts level. We could further determne the level of competton n the bankng ndustry by summng all the αs as mentoned above. The results are demonstrated as follows. Table 4 Regresson results: what affects the coeffcent onproftablty of a bank Varables All Small banks Bg banks Captal cost (lnpk) *** * Labour cost (dlnpl) * Funds cost (dlnpf) ** ** * * Rsk of Asset (rskass) *** *** *** *** *** Rsk of Asset to the square *** *** *** *** *** (rskass*rskass) Asset sze (lnasset) *** *** *** *** GDP level (lngdp) *** *** * Inflaton rate (nflaton) * Constant *** ** R squared Adjusted R squared Countres 20 banks 7 banks 13 banks Observatons Perod H statstc legend: ^p<0.10; * p<0.05; ** p<0.01; *** p<0.001 We used the Fxed Effect model snce the OLS and the Random effects model have been ruled out. We have further tred to determne the level of competton across the szes of banks and across tme. We have categorzed the followng banks (hghlghted n bold) as bg as they were classfed among the 20 largest banks. Table 5 Lst of banks No. Name of bank No. Name of bank 1 HSBC 11 Dah Sng Bank 2 Bank of Chna 12 Chna Constructon Bank (Asa) 3 Hang Seng Bank 13 Shangha Commercal Bank 4 Bank of East Asa 14 Chong Hng Bank 85 Publshed by Dgtal Lngnan Unversty,

11 Lngnan Journal of Bankng, Fnance and Economcs, Vol. 4, Iss. 1 [2013], Art. 6 5 ICBC (Asa) 15 Fubon Bank (Hong Kong) 6 DBS Bank 16 Chyu Bankng Corporaton 7 Nanyang Commercal Bank 17 Publc Bank HK 8 Wng Hang Bank 18 Orx Asa Lmted 9 Chna CITIC Bank Internatonal 19 Alled Bankng Corporaton 10 Wng Lung Bank 20 Mevas Bank Source: We also tred to analyse the effect of competton across tme by splttng the tme perods nto 2 parts: and to see the effects before and after the Fnancal Tsunam of The above table summarses the fndngs. We found that the H-stat ranged between and across both the bg and small banks, and across the dfferent tme spans. Ths ndcates that the level of competton n the bankng sector s of a monopolstc competton or a partally contestable equlbrum, whereby total revenue rses less than proportonally to the changes n nput prces. Ths does show that the bankng sector n Hong Kong s nether a monopoly nor a perfect competton stuaton as revealed by the statstcs. As such, we can deduce that the market s somehow very compettve whle t s domnated by a large amount of banks at the same tme. 7. Concluson Ths paper determnes the man factors that affect the level of proftablty n a bank. We have used the Panzar- Rosse approach n order to acheve ths. The bankng sector n Hong Kong s under a monopolstc stuaton over the perod under revew. The estmaton results showed that the compettve pressures were equal across tme and across the dfferent szes of the banks. Ths may suggest that currently banks are equally exposed to the same knd of challenges n mantanng ther proftablty. Compettve pressures may heghten n the near future as banks seeks to gan compettve edge and they can do so through mergers and acqustons. Regulatons are also slowly beng relaxed after the Fnancal Tsunam n order to boost the economy and more advances n technology are also antcpated. Hence, the compettve pressures need to be closely montored n the future to avod any surprses. Reference Allen, F., & Gale, D. (2004). Competton and Fnancal Stablty. Journal of Money, Credt and Bankng, 36(3), Avral, K. T., & Mha, M. (2011). Economc Growth and FDI n Asa: A Panel-DataAapproach. Economc Analyss and Polcy, 41(2),

12 CHU et al.: Competton n Hong Kong s bankng ndustry Berger, A. N., & Bouwman, C. H. S. (2009). Fnancal Crses and Bankng Lqudty Creaton. Unversty of South Carolna workng Paper. Bevan, A., & Danbolt, J. (2004). Testng for Inconsstences n the Estmaton of UK Captal Structure Determnants. Appled Fnancal Economcs, 14, Boyd, J., De Ncolo, G., & Jalal, A. M. (2006). Bank Rsk Takng and Competton Revsted: New Theory and Evdence. IMF Workng paper 06/29. Jang, G., Wong, J., & Sze, A. (2004). Bankng Sector Competton n Hong Kong Measurement and Evoluton over tme. HKMA Research Memorandums, Hong Kong Monetary Authorty. Panzar, J. C., & Rosse, J. N. (1987). Testng for Monopoly Equlbrum. Journal of Industral Economcs, 25, Shaffer, S. (1982). A Non-structural Test for Competton n Fnancal Markets. Bank Structure and Competton, Conference Proceedngs, Federal Reserve Bank of Chcago, Uhde, A., & Hemeshoff, U. (2009). Consoldatton n Bankng and Fnancal Stablty n Europe: Emprcal Evdence. Journal of Bankng & Fnance, 33, Wong, J., Wong, E., Fong, T., & Cho, K. F. (2006). Competton n Hong Kong s Bankng Sector: A Panzar-Rosse Assessment. HKMA Research Memorandums, Hong Kong Monetary Authorty, 87 Publshed by Dgtal Lngnan Unversty,