Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/springer-journals/the-bank-of-japan-s-exchange-traded-fund-purchases-a-help-or-hindrance-t8PcMYkG7R
- Publisher
- Springer Journals
- Copyright
- Copyright © The Author(s) 2023
- ISSN
- 1470-8272
- eISSN
- 1479-179X
- DOI
- 10.1057/s41260-023-00307-2
- Publisher site
- See Article on Publisher Site
Abstract
We examine the impact of the Bank of Japan’s exchange traded fund (ETF) purchases on two aspects of market efficiency— long-range dependence and price delay—of the TOPIX and Nikkei 225 indices. An increase in ETF purchases results in lower long-range dependence for both indices while the impact on the price delay varies according to index and measure. A sub-period analysis shows that the impact on market efficiency varies over time, with the dominant pattern being a delayed harmful effect, followed by a positive impact and thereafter a negative effect. The implications of these findings are discussed. Keywords Exchange traded funds · Market efficiency · Long-range dependence · Price delay · Monetary policy · Asset purchase programmes JEL Classification E52 · E58 · G10 · G14 Introduction purchasing government bonds (Petrov 2017). While some central banks purchase private domestic securities such as As part of their monetary policy, the Bank of Japan (BOJ) corporate bonds, stock market purchases are rare and, if any, began purchasing exchange traded funds (ETFs) tracking the are limited to foreign securities as part of foreign exchange TOPIX and Nikkei 225 indices in December 2010. While the reserves. Although the BOJ’s holdings of ETFs are small in purchasing programme was only intended to last for a year comparison to its holdings of government securities (¥528 with a limit of ¥450 billion, it continues more than a decade trillion as of 30 September 2021), their involvement in the later, with the purchase limit extended eight times with an domestic stock market is unprecedented. Such a strategy is annual limit of ¥12 trillion. However, ETF purchases in risky because ETFs do not mature like bonds meaning that 2021 were the lowest since the introduction of the policy. at some point this policy must be unwound. Since the intro- As of 30 September 2021, the BOJ owns more than ¥36.4 duction of the purchasing policy, the BOJ has not sold any trillion in ETFs, which equates to approximately 80% of ETFs (Lee and Fujioka 2022). The purpose of the BOJ’s all Japanese ETFs and 7% of the value of the first section ETF purchases has been to encourage risk taking by reduc- of the Tokyo Stock Exchange (TSE) (Fujikawa 2021; Lee ing asset risk premiums, lowering the cost of capital for and Fujioka 2022). The BOJ is now the largest shareholder companies, and indirectly boosting economic activity in an on the TSE, overtaking the Japanese Government Pension environment of persistently low or negative ina fl tion (Petrov Fund. Since 2011, Japanese ETF assets have expanded by 2017; Harada and Okimoto 2021). According to Hattori 2222% from ¥2.7 to ¥60 trillion (Lee and Fujioka 2022). For further details on the changes in the BOJ’s asset purchase pro- In an environment of low interest rates, many cen- gramme see Hanaeda and Serita (2017) and Harada and Okimoto tral banks turn to large scale asset purchase programmes (2021). The ¥6 trillion limit was doubled in 2020 due to the Covid-19 (quantitative easing) to increase money supply, typically pandemic. The Hong Kong Monetary Authority purchased domestic stocks during the Asian Financial Crisis in 1998. Although these were large in magnitude, they only lasted two weeks (Hanaeda and Serita 2017). * Ailie Charteris ailie.charteris@uct.ac.za In June 2020, the bank launched a lending scheme whereby it lends its ETF holdings to market participants, but the programme has strug- University of Cape Town, Cape Town, South Africa gled to attract investor interest. Vol.:(0123456789) A. Charteris, C. A. Steyn and Yoshida (2020), this policy was one of the BOJ’s last Chen et al. (2019) present evidence that Japanese stocks with resorts to prevent market turmoil after the Global Financial higher levels of BOJ ownership exhibit lower levels of price Crisis in 2008, having exhausted all other monetary policy informativeness. instruments. In this study, we conduct a comprehensive analysis of An ETF is a basket of stocks that tracks an index. Shares the impact of the BOJ’s ETF purchases on the efficiency in ETFs are traded on an exchange, but if there is excess of the Japanese market, considering both the TOPIX and demand for an ETF, the component stocks can be purchased Nikkei 225 indices. Efficiency is quantified using the Hurst and delivered to the fund manager who creates shares in exponent, which captures long-range dependence in prices, the ETF (the opposite is true if there is excess supply of and the price delay metrics of Hou and Moskowitz (2005), ETF shares) (Charteris 2013). Purchases of the constitu- which measure the speed with which new information is ent stocks drive their prices upwards. The price of the ETF incorporated into stock prices. Using adjusted ordinary and should track the net asset value (NAV) closely, otherwise feasible least squares to account for the estimated nature of arbitrage opportunities arise. If the NAV is below the price, the dependent variable, and controlling for other determi- market participants can purchase the underlying securities nants of market efficiency, we find that an increase in ETF and exchange them for shares in the ETF which can then be purchases by the BOJ results in lower long-range depend- sold at the market price (the opposite is true if the NAV is ence for both indices while the impact on the price delay above the price). Accordingly, if demand for the ETF drives varies according to index and measure. Using the Bai and the price of the ETF above its NAV, this will lead to greater Perron (2003) breakpoint test, we further observe that the demand for the constituent stocks, contributing to their impact of the BOJ’s ETF purchases on market efficiency prices rising (Da and Shive 2018). Theoretically, therefore, varies over time, with the dominant pattern being a delayed substantial purchases of ETFs by the BOJ will contribute to harmful effect, followed by a positive impact and thereafter price increases of the underlying stocks. Barbon and Giani- a negative effect. nazzi (2019) and Harada and Okimoto (2021) confirm that Our study makes several important contributions to the the prices of the stocks contained in the ETFs rise following literature. First, we expand the nascent literature on the study purchases by the BOJ. of the BOJ’s unconventional monetary policy on stock mar- According to the Efficient Market Hypothesis (EMH), an kets. While studies have examined the impact on individual efficient market is one in which share prices fully reflect all stock prices, liquidity, and capital structure (Hanaeda and available information (Fama 1970). The Adaptive Markets Serita 2017; Barbon and Gianinazzi 2019; Harada and Oki- Hypothesis (AMH) of Lo (2004) suggests that market effi- moto 2021), to the best of our knowledge, only Chen et al. ciency is not a static but rather a dynamic concept, mean- (2019), have explicitly examined the impact on market effi- ing that there are periods when prices reflect all available ciency. We build on their study by considering fluctuations information and periods when they do not. For prices to be in efficiency on a daily rather than a quarterly basis and efficient, it is necessary for investors to analyse stocks in using a longer time period that includes the accelerated ETF order to identify mispricing and trade so as to drive prices purchases during the COVID-19 crisis. We also focus on towards their intrinsic value. This is unless there are suf- efficiency at the market level rather than at the individual ficient other market players to offset the price inefficiencies stock level, in line with Samuelson’s (1998) dictum that mar- (Grossman and Stiglitz 1980). ETF investors do not con- ket efficiency may differ at the individual stock level com- sider the fundamentals of the individual companies in the pared to the market level. Second, our analysis sheds light on index (Hanaeda and Serita 2017; Sushko and Turner 2018) the impact of increased passive investment, as several stud- and, consequently, the prices of all stocks in the basket are ies have shown that increased passive investment contrib- likely to be inflated (deflated) by excess demand (supply) utes to market inefficiency (Belasco et al. 2012; Israeli et al. irrespective of whether such price movements are justified or 2017; Zou 2019). Third, we follow the work of Coles et al. not (Zou 2019). Empirical studies confirm that an increase (2022) by considering two different aspects of efficiency— in passive investing distorts market prices (Belasco et al. long-range dependence and speed of adjustment—which are 2012; Zou 2019). The BOJ ETF purchases may thus impede often examined separately in studies (Al-Yahyaee et al. 2020; the price discovery process as no fundamental analysis is Köchling et al. 2019) but rarely together. Fourth, drawing conducted and the BOJ predominantly purchases when the from the idea of time-varying efficiency under the AMH, market is in a downward swing (Harada and Okimoto 2021). our study contributes to further understanding what factors contribute to variation in market efficiency over time. Allied to this, we also build on prior studies of the efficiency of the Japanese market under the AMH (Noda 2016; Jiang and Li Although some ETFs adopt a synthetic replication strategy, for 2020) and EMH (Efremidze et al. 2021). the purposes of this study, physical replication is assumed given that ETFs purchased by the BOJ follow this approach. The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? Inefficiency reduces the allocative efficiency of the stock find that flows into passive funds cause an increase in the market which lowers economic growth and returns, and trig- price-to-fundamental ratios of the underlying stocks. The gers heightened volatility (Woolley and Bird 2003). As such, results of Coles et al. (2022) reveal that the prices of stocks understanding the impact of the BOJ’s purchases on market with a greater proportion of passive investors deviate more efficiency is important as the BOJ considers the future of from a random walk, a measure of weak-form market effi- its ETF purchase programme and for other central banks ciency. However, they find that price delay and the earn- that seek mechanisms to spur economic activity. Our study ings response coefficient are not impacted by the proportion also provides insights as to the impact of increased pas- of passive investors. In reconciling these disparate results, sive investment management on market dynamics, which is Coles et al. (2022) argue that while increased passive invest- important given the growing trend towards this management ment does impact the price formation process, it does not approach. affect the ability of arbitrageurs to trade and impound infor - The remainder of this study is organised as follows: mation into stock prices. Glosten et al. (2021) also identify Section "Literature review" provides a brief review of the that an increase in ETF ownership results in improved incor- literature on the impact of increased passive management poration of earnings information into prices in the short run, on market efficiency and the effect of the BOJ’s ETF pur - but only for firms with a weak information environment. chases on the Japanese market. Section "Data and method" Several studies have examined the effects of the BOJ’s describes the data and method employed in this study. Sec- ETF purchases on stock prices, liquidity and volatility. tion "Results" presents the results and Section "Conclusion" Results suggest prices of stocks included in the TOPIX concludes. and Nikkei 225 indices have increased more compared to those stocks not part of these indices. There is mixed evi- dence, however, as to whether the effects are temporary or Literature review sustained. Findings also suggest that the price impact has become smaller over time despite the increased purchase Lorie and Hamilton (1973) were the first to suggest that amounts by the BOJ (see Barbon and Gianinazzi 2019; Char- increased passive investing will reduce competition for oenwong et al. 2019; Harada and Okimoto 2021; Aono et al. information in the market and so harm market efficiency. 2021; Shen et al. 2021). Chen et al. (2019) examine whether With the rise in passive investing in the last decade, several holdings by the BOJ impact price efficiency at the stock theoretical models have been developed to understand the level. They find that increased holdings by the BOJ results in consequence of passive investing on market efficiency. The lower price informativeness, as measured by variations from model of Breugem and Buss (2019) infers that an increase a random walk and price delay. Chen et al. (2019) also show in passive investing results in a decline in the number of that increased holdings by the BOJ lowered levels of mar- shares that are price sensitive to new information, resulting ket liquidity. With respect to volatility, Hanaeda and Serita in a decrease in informational efficiency. Similarly, Bond (2017) observe that purchases of ETFs by the BOJ initially and Garcia (2022) conjecture that increased passive owner- increased the return volatility of the constituent stocks but ship contributes to a decline in aggregate market efficiency, thereafter volatility declined. The impact of the BOJ’s pur- meaning that market prices are more divorced from cash chases on company fundamentals, including capital struc- flows. However, through the welfare consequences of the ture, profits and investment, has also been examined (see shift towards passive investing, the relative efficiency of Charoenwong et al. 2019; Gunji et al. 2021; Linh 2021). individual stocks increases. This is consistent with Samuel- Studying the efficiency of the Japanese market, Nagayasu son’s (1998) dictum that markets can be micro-efficient but (2003) documents evidence of long-range dependence in macro-inefficient. Gârleanu and Pedersen ( 2018) propose an stock returns in violation of the weak-form EMH. Efremidze equilibrium model of the optimal active and passive port- et al. (2021) also observe evidence against the weak-form folios that illustrates that a decrease in the cost of passive EMH. However, their results reveal that the BOJ’s ETF pur- investing, a result of increased fund flows to passive invest- chases may have, at least temporarily, made markets more ments, results in a decrease in market efficiency. Therefore, weak-form efficient. Noda (2016) finds that the efficiency these models predominantly conjecture a decline in market of the TOPIX has varied over time, consistent with the efficiency arising from greater passive investment. AMH. Jiang and Li (2020) confirm that the TOPIX is weak- Evidence largely supports the assertions of these theo- form efficient in normal conditions, but not in bull and bear retical models. Israeli et al. (2017) find that increased ETF markets. ownership has contributed to a decline in informational effi- Several key findings emerge from this literature. Across ciency as stocks move more in line with their sector and markets, there is evidence that increased passive investing the broader market and less in line with their own earn- has affected market efficiency. Studies on the BOJ’s unique ings. Similarly, both Belasco et al. (2012) and Zou (2019) ETF purchasing programme show that it has resulted in A. Charteris, C. A. Steyn increases in the prices of the constituents of the TOPIX and which is a measure of the variability of a time series (Cajue- Nikkei 225 indices, with some evidence that this policy has iro and Tabak 2004a, b). The R∕S statistic is the range of distorted the informativeness of stock prices. We build on partial sums of deviations of the return series from its mean this research framework by exploring the extent to which ( R ), rescaled by its standard deviation ( S ). Mat hematically , the BOJ’s ETF purchases have impacted the time-varying this is given by: efficiency of the Japanese market. R∕S ≡ max r t − r − min r t − r ( ) ( ) 1≤t≤ 1≤t≤ t=1 t=1 Data and method (1) where {r(1), r(2) … r( )} is series of continuously com- Data pounded daily returns and r and are the sample mean and standard deviation of r( ) (Lo 1991). R∕S is described The first part of our analysis is exploratory to compare the by the following empirical relationship: measures of market efficiency in the pre-BOJ ETF purchase period to those in the BOJ ETF purchase period. There- R∕S ≈ C (2) after, we set out to examine the impact of the BOJ’s ETF where C is a constant independent of and H is the Hurst purchases on market efficiency after the introduction of the exponent. purchasing programme. If markets exhibit high levels of long-range dependence The BOJ targets its ETF purchases on the TOPIX and between return values, this implies that the movement of Nikkei 225 indices. The TOPIX comprises the first section stock prices is not random, thus contradicting the weak-form of the TSE, tracking approximately 2000 stocks, and the EMH. If 0.5 < H < 1.0 , returns are positively autocorrelated Nikkei 225 comprises 225 stocks from the TSE first section. meaning returns exhibit persistence. If 0 < H < 0.5 , retur ns The TOPIX and Nikkei 225 are market capitalisation- and are negatively autocorrelated meaning returns exhibit anti- price-weighted indices respectively. The BOJ’s purchases persistence. If H is approximately equal to 0.5, returns fol- of ETFs commenced in December 2010 although purchases low a random walk (returns are independent) or returns have accelerated substantially after the introduction of the Quan- only short-range dependence (Barunik and Kristoufek 2010). titative and Qualitative Easing (QQE) policy in April 2013 To filter the stock returns of any short-run dependence and (Harada and Okimoto 2021). Closing price data for all vari- thus ensure that if H is approximately equal to 0.5 then the ables (except where noted) are obtained from Bloomberg for series can be considered to follow a random walk, the R∕S all trading days. Returns are defined as the natural logarith- statistic is computed on the standardised residuals of an mic differences in index levels multiplied by 100. We collect AR(1)-GARCH(1,1) model (Cajueiro and Tabak 2004a, b; data from 1 January 2001 to 26 November 2021. Due to the Rejichi and Aloui 2012). use of rolling windows (see Section "Methods"), the pre- The efficiency of the market has also been widely quanti- BOJ ETF purchase period is designated 20 January 2003 to fied using the price delay measures of Hou and Moskowitz 14 December 2010 and the BOJ ETF purchase period from (2005) which capture the speed with which prices reflect 15 December 2010 to 17 November 2020. Both periods are new information (see Hooy and Lim 2013; Bramante et al. used for the comparative analysis while the regression analy- 2015; Köchling et al. 2019). The greater the efficiency of the sis is estimated for the BOJ ETF purchase period. market, the smaller the price delay. We estimate a regression of the daily returns of each index against contemporaneous Methods and lagged market returns as follows: We follow the extant literature (see for example, Cajueiro and Tabak 2004a, b; Rejichi and Aloui 2012; Hiremath and Narayan 2016) by calculating a rolling Hurst Exponent to The conditional mean and variance equations for the AR(1)-GARCH(1,1) models are r = + r + and measure the long-range dependence in the indices. A fixed t t−1 t h = + + h , respectively. The standardised residuals are t t−1 t−1 window length of 500 observations is used. The Hurst expo- √ (t) = (t)∕ h(t). nent is estimated using the rescaled range, denoted R∕S , Although Lo (1991) proposed a modification to the R∕S statis- tic which is robust to short-run dependence, this modification is criticised as it only provides a conservative estimate of short-range The JPX-Nikkei 400 was introduced as a targeted index on 19 dependence (Teverovsky et al. 1999). November 2014. However, purchases of ETFs tracking this index constituted only 4% of the BOJ's total ETF purchases prior to 2016. Although weekly data is frequently used to estimate the price delay Hence, this index is not included in this study (see also Barbon and measure, using daily data results in more precise estimates (Hou and Gianinazzi 2019). Moskowitz 2005; Bramante et al. 2015). The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? where y is the measure of the efficiency of the index; x t t r = + r + r + (3) refers to the value of the BOJ’s purchases of ETFs track- i,t i i m,t i,k m,t−k i,t k=1 ing the index; z is a set of control variables; d is an event t t dummy; and e is the random error term. The measurement where r and r are the returns on the index and market i,t m,t of the variables is outlined in Table 1. respectively and k is the number of lags. The index is either To capture long-range dependence, we compute the devi- the TOPIX or Nikkei 225. For the market index, we follow ation from efficiency using the Hurst exponent, termed the Lim and Hooy (2010) and Hooy and Lim (2013) by using efficiency gap. A value close to zero indicates efficiency an international index to capture the extent to which the whereas a value close to 0.5 denotes inefficiency. This meas- Japanese indices respond to new global information, namely ure does not consider the possibility of anti-persistence the MSCI All Country World Index excluding Japan (MSCI 9 (Vidal-Tomás 2022). A lower value thus indicates a more ACWI ex Japan). If the Japanese market responds imme- efficient market. For the price delay measures (D1 and D2), diately to new information, will be significantly different the lower the price delay, the more informationally efficient from zero and none of the coefficients will differ from i,k the market. zero (Hou and Moskowitz 2005). If there is a delayed market Data on the key explanatory variable, ETF purchases response, then some of the coefficients will differ from i,k by the BOJ, was obtained from the BOJ website. This data zero. Two lags of the market index are found to be optimal. reflects total purchases and does not distinguish between Using the estimates from this regression, we calculate two purchases of ETFs tracking the various indices. We use measures of price delay (Hou and Moskowitz 2005). The 2 the weightings outlined by the BOJ to proportion the pur- first measure, D1, is computed as one minus the ratio of R chases per index over time. To ensure the robustness of from Eq. (3) with the restriction that = 0, ∀k ∈ [1, 2] and i,k 2 the results, we compute a second measure of ETF pur- the original R from Eq. (3): chases by the BOJ for each index, consisting of the daily R net inflows to the ETFs tracking each of the indices (see =0,∀k∈[1,2] i,k (4) D1 = 1 − Barbon and Gianinazzi 2019). If the BOJ’s purchases of ETFs contribute to a decline (improvement) in market effi- As longer lags are more severe for the efficiency of the ciency, then the coefficient in Eq. (6) will be positive market than shorter lags, a second delay measure is com- (negative). puted as follows: Several control variables are specified. Liquidity, which we measure by the turnover ratio (TO), has been i,k k=1 shown to attract arbitrage trading which contributes to D2 = (5) increased efficiency (Chordia et al. 2008; Chung and + k i i,k k=1 Hrazdil 2010; Al-Yahyaee et al. 2020). Hence, a negative The regressions are estimated using a rolling window to coefficient is expected. Arshad and Rizvi (2015) and Al- allow for variation in the delay measure over time (Bramante Yahyaee et al. (2020) illustrate that increased volatility et al. 2015). A window of 500 observations is again used. hampers market efficiency. We thus expect a positive rela- To assess the impact of the BOJ’s purchases of passive tionship between volatility and the efficiency measures. ETFs on Japanese market efficiency, we build on the studies Finally, market size is included as size has been found to that have examined the determinants of long-range depend- affect the level of efficiency, with larger markets exhibit- ence (Hiremath and Narayan 2016; Al-Yahyaee et al. 2020) ing greater efficiency (Hooy and Lim 2013; Arshad et al. and price delay (Hooy and Lim 2013; Köchling et al. 2019). 2016). A negative coefficient is thus expected. A dummy We estimate the following equation for the TOPIX and Nik- variable is included for Japan’s 2011 Fukushima nuclear kei 225 indices: disaster, denoted from 11 March to 11 April 2011 (see Betzer et al. 2013). y = + x + z + d + e (6) t t t t t From December 2010 to 18 November 2014, 46% and 54% were allocated to ETFs tracking the TOPIX and Nikkei 225 respectively. Thereafter until 20 September 2016, 42% of purchases were directed at the TOPIX and 53% to the Nikkei 225. For the next period running This index covers approximately 85% of the global equity opportu- to 30 July 2018, 49% of purchases were allocated to the TOPIX, and nity set outside of Japan. 50% split between the TOPIX and Nikkei 225. Until 19 March 2021, Hou and Moskowitz (2005) propose a third measure of price delay 73.68% of purchases were directed at TOPIX-tracking ETFs and the which adjusts for the precision of the estimates. This was estimated remaining 25.3% split between ETFs tracking the TOPIX and Nikkei but due to its almost perfect correlation with D2 (0.99), results are not 225. For the final period, only ETFs tracking the TOPIX were pur - shown. chased (BOJ 2021). A. Charteris, C. A. Steyn Table 1 Measurement of the variables for each index Variables Acronym Measure Dependent variable Efficiency gap EG The Hurst exponent less 0.5. Price delay D1 The price delay measures of Hou and Moskowitz (2005). D2 Independent variables BOJ ETF purchases PUR The total purchases by the BOJ of ETFs tracking the index. ETF inflows IF The inflows to ETFs tracking the index. Turnover ratio TO The number of trades per share in the index divided by the number of shares outstanding of all constituents of the index. Volatility VOL The 30-day rolling standard deviation of the returns of the index. Change in size ∆SIZE The change in the natural log of the market capitalisation of all the con- stituents in the index. Fukushima disaster D A dummy equal to one from 11 March to 11 April 2011 and zero other- FD wise. The various efficiency measures are based on esti- Figure 1 plots the efficiency measures for the two indi- mates and hence exhibit sampling variance. Employing ces from 2003. The red vertical line on 15 December 2010 ordinary least squares (OLS) to estimate Eq. (6) may denotes the commencement of the BOJ’s ETF purchase thus lead to inefficient estimates. Accordingly, follow- programme. The graphs demonstrate that both long-range ing Lewis and Linzer (2005), we use OLS adjusted for dependence and price delay vary substantially over time. heteroscedasticity based on White’s standard errors and This supports the assertions of the AMH of time-varying feasible generalised least squares (GLS) to ensure reli- efficiency, consistent with the findings of Noda (2016) and able results. Jiang and Li (2020) for the Japanese market. From 2012, the efficiency gaps of both indices experience a sharp rise, suggesting greater dependency in prices. This increase dissi- Results pates quite quickly with the efficiency gaps reaching notable lows in early 2014. The improvement in efficiency could be Table 2 presents the descriptive statistics for the efficiency attributable to the greater ETF purchases by the BOJ from measures in the pre-BOJ ETF purchase period, 20 January April 2013 under the QQE policy. This effect would be 2003 to 14 December 2010, and the BOJ ETF purchase somewhat delayed because of the rolling window of histori- period, 15 December 2010 to 17 November 2020. The effi- cal observations used to estimate the Hurst exponent. From ciency gap of each index is close to zero in both periods this point onwards, the long-range dependency continues suggesting low long-range dependence consistent with the to exhibit upward and downward movements but what is weak-form EMH (Cajueiro and Tabak 2004a). For both the noticeable is that the high points are lower than previously. TOPIX and Nikkei 225, the mean efficiency gaps (0.0557 This is consistent with the finding that the average efficiency and 0.0558, respectively) are lower in the BOJ ETF pur- gap for both the TOPIX and Nikkei 225 are lower in the BOJ chase period than in the pre-BOJ purchase period (0.0596 period (Table 2). and 0.0634, respectively). Similarly, the D1 price delay D1 exhibits no sharp movements until early 2014 when measures are lower in the BOJ ETF purchase period than a dramatic spike occurs for both indices (Fig. 1). The price in the prior period (for the TOPIX of 0.5799 and 0.5956, delay measures also rely on a rolling window and thus this and for the Nikkei 225 of 0.5662 and 0.5809 for the respec- spike may be attributable to the acceleration of ETF pur- tive periods). Contrastingly, the average measures for D2 are chases by the BOJ in April 2013. This rise in D1 persists substantially higher in the BOJ ETF purchase period for the only briefly before commencing a decline which continues TOPIX and Nikkei 225 indices (0.5743 and 0.5099, respec- until early 2018 before another sharp rise. D1 predominantly tively) compared to the prior period (0.4238 and 0.4267, fluctuates in the range 0.6–0.7 for the TOPIX and 0.55— respectively). D2 gives greater weightings to longer time 0.65 for the Nikkei 225 thereafter. An increase in D2 also delays than D1 (see Section "Methods") and thus the results occurs from 2014 onwards. However, when the value of D2 suggest that information from two days prior (two lags were declines, it does not decrease as far as seen previously and found to be optimal) takes longer to be incorporated into remains above previous lows as it fluctuates over time. This prices since 2010. confirms the conclusion from Table 2 that D2 for both the The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? Table 2 Descriptive statistics for the variables before and during the BOJ ETF purchasing period TO ∆SIZE VOL EG D1 D2 TO ∆SIZE VOL EG D1 D2 PUR IF Panel A: TOPIX Pre-BOJ ETF Purchase Period Panel B: TOPIX BOJ ETF Purchase Period Mean 6649.35 − 2.42E−05 20.8981 0.0596 0.5956 0.4238 7176.87 0.0003 18.3970 0.0557 0.5799 0.5743 91.60 9584.91 Median 6671.13 0.0006 18.5300 0.0612 0.6225 0.4163 6727.14 0.0006 16.0700 0.0508 0.5844 0.5657 0.0000 48.6025 Std Dev 2268.19 0.0154 11.1715 0.0423 0.0982 0.0635 2822.99 0.0126 8.2780 0.0458 0.1114 0.0716 202.23 47,967.56 Kurtosis 0.1681 17.1415 15.6025 − 0.6288 − 0.9203 0.1666 5.7978 6.0266 2.6345 − 0.1902 − 0.2472 − 0.1108 8.8698 47.90 Skewness 0.3686 − 1.2161 3.3029 − 0.2355 − 0.6129 0.4711 1.7928 −0.4895 1.5668 0.4947 − 0.3906 0.4425 2.7106 2.6542 Min 1989.72 − 0.1886 6.4700 − 0.0676 0.3190 0.2417 2770.6156 −0.0997 5.9400 − 0.0475 0.2813 0.4321 0.0000 − 486,425.70 Max 17,537.84 0.1271 91.9700 0.1590 0.7965 0.6091 28,899.81 0.0758 49.0600 0.1794 0.8253 0.8354 1645.55 751,359.85 SW 0.9820*** 0.8962*** 0.7122*** 0.9803*** 0.8981*** 0.9723*** 0.8782*** 0.9382*** 0.8622*** 0.9727*** 0.9679*** 0.9818*** 0.5186*** 0.5457 Panel C: Nikkei 225 Pre-BOJ ETF Purchase Period Panel D: Nikkei 225 BOJ ETF Purchase Period Mean 67704.27 1.44E−05 22.7422 0.0634 0.5809 0.4267 86414.91 0.0003 19.8580 0.0558 0.5662 0.5099 33.06 3044.67 Median 66693.44 0.0006 20.7650 0.0662 0.5932 0.4134 81210.76 0.0005 17.3850 0.0488 0.5741 0.5158 0.0000 986.15 Std Dev 26266.44 0.0162 12.4975 0.0453 0.0832 0.0670 35996.94 0.0131 8.7445 0.0536 0.1060 0.0787 61.54 39730.90 Kurtosis − 0.0585 9.0717 17.1003 − 0.3490 − 0.8700 1.7802 7.1758 5.4931 2.5893 − 0.6963 − 0.0515 −0.7688 4.4882 38.02 Skewness 0.4754 − 0.5416 3.5145 − 0.3968 − 0.2532 − 0.0668 1.9118 −0.3691 1.5292 0.2859 − 0.5628 −0.2310 2.0155 − 1.4211 Min 15,462.26 − 0.1363 7.5800 − 0.0629 0.3234 0.1947 24397.14 −0.1004 6.8200 − 0.0559 0.2619 0.3195 0.0000 − 533,748.28 Max 183,521.16 0.1377 104.0400 0.1585 0.7480 0.6540 388747.51 0.0802 55.2500 0.1835 0.7977 0.7191 388.49 351,848.71 SW 0.9741*** 0.9200*** 0.6857*** 0.9793*** 0.9604*** 0.9311*** 0.8729*** 0.9438*** 0.8700*** 0.9769*** 0.9541*** 0.9780*** 0.6013*** 0.6606*** This table presents the descriptive statistics for the variables in the pre-BOJ ETF purchase period (20 January 2003–14 December 2010) and during the BOJ ETF purchase period (15 December 2010–17 November 2020). TO refers to the turnover of the index, ∆SIZE refers to the change in the natural logarithm of the market capitalisation of the index, VOL refers to the 30-day rolling return volatility of the index, EG refers to the efficiency gap (defined as H—0.5, where H is the Hurst exponent), D1 and D2 refer to the two price delay measures of Hou and Moskowitz (2005), PUR refers to the value of the BOJ’s ETF purchases and IF refers to the net inflows to ETFs tracking the index. SW is the Shapiro Wilk test of normality. ***indicates significance at 1% A. Charteris, C. A. Steyn Panel A: TOPIX EG PanelB: TOPIX D1 Panel C: TOPIXD2 .15 .8 -.05 -.10 .2 .2 04 06 08 10 12 14 16 18 20 04 06 08 10 12 14 16 18 20 04 06 08 10 12 14 16 18 20 Panel D: Nikkei 225 EG Panel E: Nikkei 225 D1 Panel F: Nikkei 225 D2 .9 .8 .8 .7 .7 .6 .6 .5 .5 .4 .4 .3 -.05 -.10 .2 .1 04 06 08 10 12 14 16 18 20 04 06 08 10 12 14 16 18 20 04 06 08 10 12 14 16 18 20 Fig. 1 Efficiency measures for the TOPIX and Nikkei 225 indices. are the first and second price delay measures of Hou and Moskowitz This figure plots the efficiency measures for the TOPIX (A–C) and (2005). The red vertical line on 15 December 2010 denotes the com- Nikkei 225 (D–F). EG is the Hurst exponent less 0.5 and D1 and D2 mencement of the BOJ’s ETF purchase programme TOPIX and Nikkei 225 is substantially higher in the BOJ due to the BOJ’s purchase programme but also due to other ETF purchase period compared to the prior period. It is thus events/news which impacted Asia’s markets as well such as evident that the long-range dependence and price delay have the 2013 taper tantrum and the bursting of China’s stock changed since the introduction of the BOJ’s ETF purchase market bubble and economic downturn in 2015 which raised programme. Whether these changes can be attributed to the fears that it would spark another financial crisis (Shu et al. BOJ’s monetary policy is investigated further below. 2015; Kaletsky 2015). Similarly, changing efficiency in 2018 As a comparative analysis, we examine the efficiency across the Japanese and Asian indices may reflect events of the MSCI Asia excluding Japan index (MSCI Asia ex such as the launch of missiles by North Korea and the Chi- Japan). Figure 2 shows that this index also exhibits time- nese government’s regulation of internet finance companies varying long-range dependence and price delay. D1 and during 2017 (Mundey 2017). The fact that the timing differs D2 exhibit several peaks and troughs but, overall, there is a between Asian and Japanese markets, however, may reflect decline in price delay over the period, i.e., faster incorpora- the role of the BOJ’s purchase programme. tion of new information across Asian markets. This contrasts As all variables are stationary based on the Augmented with the trend for the Japanese indices seen in Figure 1 of Dickey-Fuller test, we proceed to estimate Eq. (6). The increasing price delay post 2010. The efficiency gap of the results for OLS with White’s standard errors are shown in Asia ex Japan index is lower from 2010 to 2014 than from Table 3. The findings using FLS (Table 6 in the “Appen- 2003 to 2009. Thereafter, long-range dependence increases dix”) confirm those obtained using OLS. The BOJ’s ETF notably although wide fluctuations are evident. In compari- purchases have a significant negative impact at the 1% level son to the Japanese market, the lower ec ffi iency gap of Asian on both the TOPIX and Nikkei 225 efficiency gap, with the markets occurs earlier than seen in Japan and is more per- impact larger for the latter (coefficients of −1.14E−05 and sistent. Thus, the improvement in long-range dependence −0.0002, respectively). This suggests that an increase in seen in Figure 1 for Japan may not necessarily be entirely ETF purchases by the BOJ contributes to an improvement This index comprises stocks from two developed and eight emerg- ing markets in Asia. It was chosen given the findings of prior stud- The taper tantrum refers to the response by markets to the unex- ies showing similarity in price delay across Asian markets (see Lim pected news that the United States Federal Reserve was reducing its & Hooy 2010; Hooy & Lim 2013), thus allowing for a comparison quantitative easing policy. to the efficiency of Japan’s market since the commencement of the BOJ’s ETF purchase programme. We thank an anonymous reviewer The results of the unit root tests are available from the authors for the suggestion of a comparative analysis. upon request. The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? Panel A: EG -.05 -.10 04 06 08 10 12 14 16 18 20 Panel B: D1 Panel C: D2 .4 .4 .2 .2 .1 .1 04 06 08 10 12 14 16 18 20 04 06 08 10 12 14 16 18 20 Fig. 2 Efficiency measures for the MSCI Asia ex Japan index. This first and second price delay measures of Hou and Moskowitz (2005) figure plots the efficiency measures for the MSCI Asia ex Japan (B and C, respectively) index. EG is the Hurst exponent less 0.5 (A) and D1 and D2 are the in efficiency due to lower long-range dependence. This fol- impact for the Nikkei 225 (9.13E−05). ETF inflows have no lows the findings of Efremidze et al. (2021) that the BOJ significant impact on D2 for both indices. The negative sign purchase programme may have temporarily improved weak- on the BOJ’s ETF purchases for the TOPIX suggests that form efficiency of the Japanese market. Net ETF inflows the BOJ’s purchasing programme decreases the price delay have no impact on the efficiency gap. meaning that information is incorporated into prices faster. Differences are observed across the indices for the price In contrast, for the Nikkei 225, the purchasing programme delay measures. The BOJ’s ETF purchases have a signifi- results in an increase in the price delay when longer delays cant negative impact on D1 for TOPIX but not the Nikkei are given greater weighting (D2 vs. D1). 225 (−7.44E−05 and −2.68E−05, respectively). The same This difference across indices could be attributable to is seen with net ETF inflows (−1.41E−07 and −4.84E−08, the different weightings of the TOPIX and Nikkei 225 respectively). The BOJ’s ETF purchases have a negative indices. The Nikkei 225 index is price weighted whereas impact on D2 of the TOPIX (−2.35E−05) but a positive the TOPIX is market capitalisation weighted. According A. Charteris, C. A. Steyn Table 3 Results from the adjusted OLS regressions Dependent variable EG EG D1 D1 D2 D2 Panel A: TOPIX Intercept 0.0335*** 0.0321*** 0.5345*** 0.5255*** 0.5913*** 0.5884*** PUR − 1.14E−05*** − 7.44E−05*** − 2.35E−05*** IF − 1.84E−08 − 1.41E−07*** − 4.48E−08 ∆SIZE −0.0242 0.0389 − 0.4202*** − 0.0049 − 0.3444*** − 0.2133* TO 3.62E−06*** 3.74E−06*** 3.17E−06*** 3.94E−06*** − 2.97E−06*** − 2.73E−06*** VOL −0.0002 − 0.0002* 0.0016*** 0.0015*** 0.0003* 0.0003 D 0.0032 0.0040 − 0.0342*** − 0.0293*** 0.0370*** 0.0385*** FD Adjusted R 0.0492 0.0511 0.0303 0.0425 0.0120 0.0149 Panel B: Nikkei 225 Intercept 0.0419*** 0.0394*** 0.4885*** 0.4881*** 0.5424*** 0.5438*** PUR − 0.0002*** − 2.68E−05 9.13E−05*** IF − 3.08E−08 − 4.84E−08 2.49E−08 ∆SIZE − 0.2850*** 0.0505 − 0.0447 0.0179 − 0.0682 − 0.2599** TO 1.88E−07*** 1.65E−07*** 5.98E−07*** 5.94E−07*** − 2.64E−07*** − 2.51E−07*** VOL 0.0002 0.0001 0.0014*** 0.0014*** − 0.0007*** − 6.48E−04*** D − 0.0127*** − 0.0073* − 0.0252*** − 0.0245*** 0.0856*** 0.0826*** FD Adjusted R 0.0124 0.0390 0.0665 0.0663 0.0271 0.0310 This table presents the results of the regressions of the BOJ’s ETF purchases against various measures of market efficiency using OLS with White’s heteroscedasticity-consistent standard errors. Panels A and B present the results for the TOPIX and Nikkei 225 respectively. TO refers to the turnover of the index, ∆SIZE refers to the change in the natural logarithm of the market capitalisation of the index, VOL refers to the 30-day rolling return volatility of the index, EG refers to the efficiency gap (defined as H—0.5, where H is the Hurst exponent), D1 and D2 refer to the two price delay measures of Hou and Moskowitz (2005), PUR refers to the value of the BOJ’s ETF purchases and IF refers to the net inflows to ETFs tracking the index. ***, ** and * indicate significance at 1%, 5% and 10%, respectively to the EMH, fundamental information should be based for the differing findings on the TOPIX relative to the Nik - on market capitalisation and, as such, the allocation of kei 225. Differing levels of liquidity between stocks in funds by the BOJ based on non-market weights may exac- the Nikkei 225 versus the TOPIX may be a contributing erbate the negative effects on market efficiency (Barbon factor. The Nikkei 225 selects stocks which are highly liq- and Gianinazzi 2019). To investigate the impact of index uid (Chen et al. 2019). Thus, more liquid stocks may be weighting, we created a synthetic Nikkei 225 index, negatively impacted by the trades of the BOJ whereas less accounting for all index additions/ deletions over the liquid stocks (such as some of those in the TOPIX) are period, where the constituents are weighted according to positively impacted (see also Chen et al. 2019). We recom- market capitalisation rather than price. Results using mend the analysis of liquidity of the index constituents as adjusted OLS are shown in Table 4. Consistent with the an avenue for further research. findings for both the TOPIX and Nikkei 225 indices, the The finding that the BOJ’s ETF purchasing programme BOJ’s purchases contribute to a decline in the efficiency has largely contributed to improved speed in the incor- gap of the synthetic index. The BOJ’s purchases have no poration of information into prices and lower long-range impact on D1 but a positive and significant impact on D2. dependence in prices differs from the result of Chen et al. The significant positive impact on D2 resembles that seen (2019) of an adverse effect on the efficiency of Japanese for the canonical Nikkei 225. ETF inflows have no impact stocks. The differences may reflect varying methodologies on the efficiency measures of the synthetic index, as with and time periods studied. Chen et al. (2019) examined effi- the Nikkei 225. This analysis suggests that the different ciency at the firm level across quarters whereas our analy - weighting mechanisms do not appear to directly account sis focuses on indices daily. Also, their sample ended in 2016 where our sample extends to November 2021, captur- ing further acceleration in the ETF purchasing programme including that driven by the COVID-19 pandemic. The We would like to thank an anonymous reviewer for this valuable differences in findings are consistent with Samuelson’s suggestion. The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? Table 4 Results for the synthetic Nikkei 225 Dependent variable EG EG D1 D1 D2 D2 Intercept 0.0234*** 0.0212*** 0.5200*** 0.5194*** 0.5724*** PUR − 0.0001*** − 3.50E−05 9.55E−05*** IF − 2.51E−08 − 6.44E−08 1.71E−08 ∆Size − 0.2295*** 0.0794 − 0.0728 0.0118 − 0.0012 − 0.2066* TO 3.14E−07*** 2.93E−07*** 4.4E−07*** 4.36E−07*** − 1.1E−08 2.96E−09 VOL 0.0002* 0.0002 0.0014*** 0.0014*** − 0.0004* − 0.0004* D − 0.0136*** − 0.0086** − 0.0371*** − 0.0362*** 0.0377 0.0344*** FD Adjusted R 0.0758 0.0499 0.0382 0.0385 0.0073*** 0.0022 This table presents the results of the regressions of the BOJ’s ETF purchases against various measures of market efficiency using OLS with White’s heteroscedasticity-consistent standard errors for the synthetic Nikkei 225 index, which weights the constituents of the Nikkei 225 according to market capitalisation rather than price. TO refers to the turnover of the index, ∆SIZE refers to the change in the natural logarithm of the market capitalisation of the index, VOL refers to the 30-day rolling return volatility of the index, EG refers to the efficiency gap (defined as H—0.5, where H is the Hurst exponent), D1 and D2 refer to the two price delay measures of Hou and Moskowitz (2005), PUR refers to the value of the BOJ’s ETF purchases and IF refers to the net inflows to ETFs tracking the index. ***, ** and * indicate significance at 1%, 5% and 10%, respectively (1998) dictum that market efficiency may vary at the firm For the price delay, an oscillating impact exists over sub- level versus the stock market level. periods. With the TOPIX, ETF purchases initially have no Given the variability in the measures of market effi- impact on D1 followed by a positive impact and then a nega- ciency over time (Figure 1), it may be that the relation- tive impact. In contrast for the Nikkei 225, the impact is ship between the BOJ’s ETF purchases, and the efficiency initially positive, then negative and then reverts to positive of the Japanese indices has varied. We therefore test for again. ETF net inflows show a similar positive, then nega- breakpoints in this relationship in Eq. (6) using the Bai and tive effect on the Nikkei 225 but the negative effect remains Perron (2003) test. The results are presented in Table 5. thereafter. No breakpoints are identified for the Nikkei 225 Four breakpoints in the relationship between the efficiency when ETF purchases are used. For the TOPIX the impact gap and the BOJ’s ETF purchases occur for both indices. on D2 is insignificant, negative, positive, negative and insig- ETF purchases have a significant impact on the efficiency nificant across the five subperiods respectively whereas for gap across all periods, but the signs vary. The coefficient is the Nikkei 225, the impact on D2 is positive, negative, posi- positive for the first two subperiods; negative for the sec- tive negative and positive respectively. ETF net inflows have ond two subperiods; and positive for the final subperiod. an insignificant effect followed by a negative impact on the This suggests that initially the BOJ’s purchases resulted in TOPIX but an initial positive impact on the Nikkei 225 fol- a decline in efficiency but from approximately December lowed by a negative, positive and then insignificant impact. 2013 efficiency improved before a fall again after 2018 and Overall, for D1 and D2, the results largely indicate that the 2019 for the TOPIX and Nikkei 225, respectively. When BOJ ETF purchasing programme initially has no impact on net ETF inflows are used, three breakpoints are identi- the speed of information incorporation into prices of the fied. Up until 2012/ 2013, the BOJ’s programme harms TOPIX, but thereafter the purchases slow information incor- efficiency (a significant positive coefficient), followed poration before enhancing price informativeness. However, by an improvement in efficiency (a significant negative as seen with the efficiency gap, a harmful effect appears to coefficient) until the end of 2016/ 2017, while there is no materialise later in the sample period. For the Nikkei 225, impact in the final sub-period. Efremidze et al. ( 2021) also the BOJ’s purchases immediately hinder price delay before noted that the improvement in weak-form efficiency of the then reducing price delay. Japanese market due to the BOJ’s monetary policy may What emerges from the subperiod analysis is of a have been temporary. Similarly, Hanaeda and Serita (2017) complex and dynamic relationship between the BOJ’s showed that the impact of the purchasing programme on ETF purchases and market efficiency. There appears to market volatility varied over time. be evidence that in the early years, the purchases harmed A. Charteris, C. A. Steyn Table 5 Results from the OLS breakpoint regressions Dependent variable EG EG D1 D1 D2 D2 Panel A: TOPIX Breaks 12/6/2012 20/11/2012 20/5/2014 30/6/2016 19/3/2013 23/5/2019 17/12/2013 10/10/2017 14/6/2016 13/2/2018 11/11/2014 8/9/2016 2/7/2018 6/3/2018 5/3/2018 Intercept 0.0355*** 0.0315*** 0.5530*** 0.5317*** 0.5927*** 0.5905*** PUR 0.0003*** 3.51E−05 0.0001 0.0005*** 0.0009*** − 0.0006*** − 0.0003*** − 0.0004*** 0.0007*** − 3.95E−05*** 5.09E−05*** − 0.0001*** 3.74E−07*** 2.52E−06 IF 2.21E−06*** 2.54E−07*** 1.89E−08 − 1.47E−07*** − 1.29E−06*** − 5.43E−07*** 1.69E−08 − 4.81E−09*** ∆SIZE − 0.0580 0.0401 0.2576* − 0.0186 − 0.1135 − 0.2113* TO 3.67E−06*** 3.87E−06*** 2.85E−06*** 4.12E−06*** − 2.64E−06*** − 3.35E−06*** VOL − 0.0002*** − 0.0002** 0.0004* 0.0012*** 1.60E−05 0.0005*** D − 0.0026 0.0048 0.0061 − 0.0231* 0.0452*** 0.0353*** FD Adjusted R 0.1507 0.0682 0.2709 0.0739 0.1628 0.0269 Panel B: Nikkei 225 Breaks 12/6/2012 13/112013 20/5/2014 No breaks 18/2/2013 5/11/2012 26/12/2013 11/10/2016 16/9/2016 19/1/2015 16/9/2014 5/8/2016 2/7/2018 12/7/2016 23/3/2016 22/5/2019 6/2/2018 Intercept 0.0422*** 0.0389*** 0.5173*** 0.5411*** 0.5442*** PUR 0.0001** 9.71E−05** 0.0002*** 0.0005*** 0.0003*** − 0.0007*** − 0.0003*** − 0.0015*** 0.0004*** − 1.14E−04*** 2.16E−04*** − 0.0001*** 4.07E−04*** 1.74E−04** IF 9.57E−07*** 8.18E−07*** − 4.35E−07*** − 1.01E−06*** 4.09E−09 8.66E−07*** − 2.47E−08 ∆SIZE − 0.0553 0.0618 0.0927 − 0.0628 − 0.2614 TO 2.02E−07*** 1.79E−07*** 4.97E−07*** − 2.19E−07*** − 2.64E−07** VOL − 0.0001 0.0001 0.0005** − 7.27E−04*** − 0.0006*** − 0.0106** − 0.0072* − 0.0031 0.0798*** 0.0838*** FD Adjusted R 0.1488 0.0349 0.2608 0.1417 0.0511*** This table presents the results of the regressions of the BOJ’s ETF purchases against various measures of market efficiency using OLS break - point regressions with White’s heteroscedasticity-consistent standard errors. The breakpoints are determined using the Bai and Perron (2003) test. Panels A and B present the results for the TOPIX and Nikkei 225 respectively. TO refers to the turnover of the index, ∆SIZE refers to the change in the natural logarithm of the market capitalisation of the index, VOL refers to the 30-day rolling return volatility of the index, EG refers to the efficiency gap (defined as 1—H, where H is the Hurst exponent), D1 and D2 refer to the two price delay measures of Hou and Moskowitz (2005), PUR refers to the value of the BOJ’s ETF purchases and IF refers to the net inflows to ETFs tracking the index. ***, ** and * indicate significance at 1%, 5% and 10%, respectively market efficiency, with both long-range dependence and and Yoshida 2020). This may ref lect market participants price delay increasing. Thereafter, however, the ETF that have overreacted to negative news driving prices purchases contributed to an improvement in price effi- too low. As such, the ETF purchases by the BOJ drive ciency. The BOJ purchases usually occur following a prices back to their intrinsic value thus acting counter- substantial negative return over the previous evening and cyclically. However, from approximately 2018 onwards, morning market, especially when the return is below the a harmful effect of the BOJ’s purchases on efficiency is third decile of the historical return distribution (Hattori observed. This may ref lect that by this point in time, the The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? BOJ’s purchases ref lected little fundamental informa- rapidly accelerated from 2013 onwards but tapered off tion but were simply an attempt to stimulate the mar- substantially in 2021. Several studies have examined ket, especially with the impact of COVID-19 (Lee and the impact of these ETF purchases on stock returns, Fujioka 2022). liquidity and volatility. However, very little attention As noted in Figure 2, the efficiency gap and price delay has been given to the impact on efficiency. In this study, measures of the MSCI Asia ex Japan index exhibited notable we investigate whether increased ETF purchases by the trend changes from 2014 to 2016, possibly driven by the taper BOJ impact the efficiency of the TOPIX and Nikkei tantrum (2013) and fears of an impending financial crisis due 225 indices. to China’s economic downturn and stock market crash (2015). Our findings show that the ETF purchase programme The breakpoints observed in the relationship between ETF pur- has enhanced weak-form market efficiency as measured chases by the BOJ and Japanese market efficiency in the period by the efficiency gap across both the TOPIX and Nikkei thereafter (late 2015 to 2016) may also reflect these events (the 225 indices. The purchase programme has also contrib- rolling window accounting for the delayed impact). Likewise, uted to improved speed of information incorporation for the breakpoints witnessed in 2018 may reflect changing trading the TOPIX but not for the Nikkei 225 where the price behaviour in light of repeated North Korean missiles as well as delay worsens especially with longer delays. Results sug- the Chinese government’s regulation of internet finance com- gest that this difference is not due to the different weight- panies in 2017. The comparative analysis thus reveals that the ings of the two indices. A breakpoint analysis reveals a impact of the BOJ’s purchases on Japanese market efficiency, time-varying impact of ETF purchases on the efficiency which vary over time, may also be driven by external events measures. The general trend that emerges is that the BOJ beyond those in Japan only. programme had no initial impact on market efficiency The results in Tables 3 and 4 show that turnover has a but thereafter hampered efficiency before improving effi- significant impact on the efficiency measures but with a ciency although the timing thereof differs across indices positive sign for the efficiency gap and D1 but a negative and efficiency measures. But what also emerges is that sign for D2. The negative sign is consistent with expecta- in the latter years, the purchase programme has hindered tions that increased turnover attracts more arbitrageurs market efficiency again. and thus supports market efficiency (Chordia et al. 2008; These findings have important implications for other Al-Yahyaee et al. 2020). However, Chordia et al. (2008) central banks that may be forced to directly intervene in and Chung and Hrazdil (2010) also demonstrate that a stock markets in unprecedented times. Moreover, the find- negative relationship between liquidity and efficiency ings of this study also have important lessons for the BOJ may arise due to adverse selection whereby those partici- when they unwind their ETF holdings. For these policy pants remaining in the market when liquidity is low, have makers our results show that large scale purchases impact greater incentives to gather more information and trade market efficiency, both positively and negatively. Adopting on this information. Chen et al. (2019) also observe vary- a counter-cyclical policy of buying (selling) when prices ing signs for turnover across the efficiency measures in are abnormally low (high) appears to better support market the Japanese market. The impact of volatility on market efficiency as it counteracts potential overreaction caused efficiency is mostly significant but the signs are incon- by market participants. Creative solutions could be imple- sistent. Accordingly, this suggests that volatility does mented by the BOJ to minimise market disruptions due to not have a persistent adverse effect on Japanese market the sale of ETFs. For example, Koll (2021) recommends efficiency (Arshad and Rizvi, 2015). Size has only lim- encouraging retirees to purchase ETFs by making the ETFs ited impact on market efficiency but when significant, exempt from inheritance tax in Japan, which is the highest the coefficient is negative confirming that increased size in the world. results in an improvement in efficiency (Hooy and Lim As shown in this study, a time-series analysis allows for 2013). The Fukushima nuclear disaster largely results in a more granular approach and thus we suggest expanding an increase in efficiency (as ref lected by negative coef- on the work of this study and Chen et al. (2019) by look- ficients) except for D2. ing at individual stocks using data of a daily frequency to further understand the impact of the BOJ’s ETF purchases on market efficiency. Conclusion Appendix The BOJ’s unconventional monetary policy of purchas- ing ETFs was initially introduced as a temporary pro- See Table 6. gramme yet remains more than a decade later. Purchases A. Charteris, C. A. Steyn Table 6 Results from the FLS regressions Dependent variable EG EG D1 D1 D2 D2 Panel A: TOPIX Weight 0.0394*** 0.0408*** 0.5310*** 0.5246*** 0.5890*** 0.5858*** PUR − 8.46E−06** − 3.69E−05*** − 2.08E−05*** IF − 2.22E−08 − 1.09E−07*** − 6.87E−09 ∆Size − 0.0185 0.0043 − 0.1747* 0.0020 − 0.2051*** − 0.1475*** TO 2.30E−06*** 2.44E−06*** 4.41E−06*** 5.22E−06*** − 3.22E−06*** − 2.89E−06*** VOL − 1.85E−05 − 0.0002 0.0013*** 0.0011*** 0.0005*** 0.0005*** D 0.0062 0.0101** − 0.0351*** − 0.0326*** 0.0360*** 0.0375*** FD Adjusted R 0.2461 0.2261 0.9327 0.9481 0.9698 0.9542 Panel B: Nikkei 225 Weight 0.0513*** 0.0486*** 0.5017*** 0.4951*** 0.5313*** 0.5315*** PUR − 0.0001*** − 3.03E−05 0.0001*** IF − 1.31E−08 − 2.46E−08 3.31E−08 ∆Size − 0.2420*** − 0.0408 − 0.0711 − 0.0127 − 0.0661 − 0.1685*** TO 3.90E−08 7.56E−08** 5.67E−07*** 6.53E−07*** − 3.09E−07*** − 2.90E−07*** VOL 0.0001 2.53E−05 0.0009*** 0.0008*** 1.39E−05 0.0002 D 0.0002 − 0.0009 − 0.0140* − 0.0178** 0.0698*** 0.0624*** FD Adjusted R 0.1530 0.0767 0.8620 0.8972 0.9647 0.9477 This table presents the results of the regressions of the BOJ’s ETF purchases against various measures of market efficiency using FLS. Panels A and B present the results for the TOPIX and Nikkei 225 respectively. Weight is the weighting used under FLS. TO refers to the turnover of the index, ∆SIZE refers to the change in the natural logarithm of the market capitalisation of the index, VOL refers to the 30-day rolling return volatility of the index, EG refers to the efficiency gap (defined as H—0.5, where H is the Hurst exponent), D1 and D2 refer to the two price delay measures of Hou and Moskowitz (2005), PUR refers to the value of the BOJ’s ETF purchases and IF refers to the net inflows to ETFs tracking the index. ***, ** and * indicate significance at 1%, 5% and 10%, respectively Acknowledgements The authors would like to thank Tessa Glyn for are inefficient: The impact of liquidity and volatility. The North assistance on previous drafts of this research. American Journal of Economics and Finance 52: 101168. Aono, Kohei, Hiroshi Gunji, and Hayato Nakata. 2021. Did the bank Funding Open access funding provided by University of Cape Town. of Japan’s purchases of exchange-traded funds affect stockprices? This research did not receive any specific grant from funding agencies A synthetic control approach. Applied Economics Letters 29(20): in the public, commercial, or not-for-profit sectors. 1859–1863. Arshad, Shaista, and Syed Aun R. Rizvi. 2015. The troika of business cycle, ec ffi iency and volatility. An East Asian perspective. Physica Declarations A: Statistical Mechanics and Its Applications 419: 158–170. Arshad, Shaista, Syed Aun R. Rizvi, Gairuzazmi Mat Ghani, and Jar- Conflict of interest None. ita Duasa. 2016. Investigating stock market efficiency: A look at OIC member countries. Research in International Business and Open Access This article is licensed under a Creative Commons Attri- Finance 36: 402–413. bution 4.0 International License, which permits use, sharing, adapta- Bai, J., and P. Perron. 2003. Computation and analysis of multiple tion, distribution and reproduction in any medium or format, as long structural change models. Journal of Applied Econometrics 18(1): as you give appropriate credit to the original author(s) and the source, 1–22. provide a link to the Creative Commons licence, and indicate if changes Bank of Japan (BOJ). 2021. Bank of Japan Accounts. https:// www. were made. The images or other third party material in this article are boj. or. jp/ en/ stati stics/ boj/ other/ acmai/ index. htm/. Accessed 21 included in the article's Creative Commons licence, unless indicated December 2021. otherwise in a credit line to the material. If material is not included in Barbon, Andrea, and Virginia Gianinazzi. 2019. Quantitative easing the article's Creative Commons licence and your intended use is not and equity prices: Evidence from the ETF program of the Bank permitted by statutory regulation or exceeds the permitted use, you will of Japan. The Review of Asset Pricing Studies 9 (2): 210–255. need to obtain permission directly from the copyright holder. To view a Barunik, Jozef, and Ladislav Kristoufek. 2010. On Hurst exponent copy of this licence, visit http://cr eativ ecommons. or g/licen ses/ b y/4.0/ . estimation under heavy-tailed distributions. Physica A: Statistical Mechanics and Its Applications 389 (18): 3844–3855. Belasco, Eric, Michael Finke, and David Nanigian. 2012. The impact of passive investing on corporate valuations. Managerial Finance References 38 (11): 1067–1084. Betzer, André, Markus Doumet, and Ulf Rinne. 2013. How policy Al-Yahyaee, Khamis Hamed, Walid Mensi, Hee-Un. Ko, Seong-Min. changes affect shareholder wealth: The case of the Fukushima Yoon, and Sang Hoon Kang. 2020. Why cryptocurrency markets The Bank of Japan’s exchange traded fund purchases: a help or hindrance to market efficiency? Daiichi nuclear disaster. Applied Economics Letters 20 (8): Hanaeda, Hideki, and Toshio Serita. 2017. Effects of Nikkei 225 799–803. ETFs on stock markets: Impacts of purchases by Bank of Japan. Bond, Philip, and Diego Garcia. 2022. The equilibrium consequences 30th Australasian Finance and Banking Conference. Available of indexing. The Review of Financial Studies 35 (7): 3175–3230. at SSRN 3019769. Bramante, Riccardo, Giovanni Petrella, and Diego Zappa. 2015. On Harada, Kimie, and Tatsuyoshi Okimoto. 2021. The BOJ’s ETF the use of the market model R-square as a measure of stock price purchases and its effects on Nikkei 225 stocks. International efficiency. Review of Quantitative Finance and Accounting 44 (2): Review of Financial Analysis 77: 101826. 379–391. Hattori, Takahiro, and Jiro Yoshida. 2020. The Bank of Japan’s inno- Breugem, Matthijs, and Adrian Buss. 2019. Institutional investors vation in ETF purchase during COVID-19. Available at SSRN and information acquisition: Implications for asset prices and 3575835. informational efficiency. The Review of Financial Studies 32 (6): Hiremath, Gourishankar S., and Seema Narayan. 2016. Testing the 2260–2301. adaptive market hypothesis and its determinants for the Indian Cajueiro, Daniel O., and Benjamin M. Tabak. 2004a. The Hurst expo- stock markets. Finance Research Letters 19: 173–180. nent over time: Testing the assertion that emerging markets are Hooy, Chee-Wooi., and Kian-Ping. Lim. 2013. Is market integra- becoming more efficient. Physica A: Statistical Mechanics and Its tion associated with informational efficiency of stock markets?. Applications 336 (3–4): 521–537. Journal of Policy Modeling 35 (1): 29–44. Cajueiro, Daniel O., and Benjamin M. Tabak. 2004b. Evidence of long- Hou, Kewei, and Tobias J. Moskowitz. 2005. Market frictions, price range dependence in Asian equity markets: The role of liquidity delay, and the cross-section of expected returns. The Review of and market restrictions. Physica A: Statistical Mechanics and Its Financial Studies 18 (3): 981–1020. Applications 342 (3–4): 656–664. Israeli, Doron, Charles Lee, and Suhas A. Sridharan. 2017. Is there a Charoenwong, Ben, Randall Morck, and Yupana Wiwattanakan- dark side to exchange traded funds? An information perspective. tang. 2019. Asset prices, corporate actions, and Bank of Japan Review of Accounting Studies 22 (3): 1048–1083. equity purchases. National Bureau of Economic Research Jiang, Jinjin, and Haiqi Li. 2020. A new measure for market effi- (NBER), Working paper no. 25525. http:// www. nber. org/ papers/ ciency and its application. Finance Research Letters 34: 101235. w25525. Kaletsky, A. 2015, 12 October. China is not collapsing. Project Syn- Charteris, Ailie. 2013. The price efficiency of South African exchange dicate. https:// www . pr oje ct- syndi cate. or g/ comme nt ar y/ wh y - traded funds. Investment Analysts Journal 42 (78): 1–11.china-i s-no t-c ollap sing-b y-a natol e-k alets ky-2 015-10 . Accessed Chen, Zhuo, Keiichi Ito, Takeshi Yamada, and Bohui Zhang. 2019. 21 December 2021. Monetary policy through the stock market: Central bank purchase Köchling, Gerrit, Janis Müller, and Peter N. Posch. Does the intro- of equity index ETFs. Available at SSRN 3458250. duction of futures improve the efficiency of Bitcoin?. Finance Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam. 2008. Research Letters 30 (2019): 367–370. Liquidity and market efficiency. Journal of Financial Economics Koll, Jesper. 2021, March 29. How Japan can escape financial 87 (2): 249–268. socialism. Official Monetary and Financial Institutions Forum. Chung, Dennis Y., and Karel Hrazdil. 2010. Liquidity and market effi- https:// www . omfif. or g/ 2021/ 03/ ho w - japan- can- escape- finan ciency: Analysis of NASDAQ firms. Global Finance Journal 21 cial- socia lism/. Accessed 27 June 2022. (3): 262–274. Lee, Min J., and Toru Fujioka. 2022, April 10. A $430bn caution- Coles, Jeffrey L., Davidson Heath, and Matthew C. Ringgenberg. ary tale inside Japan’s central bank. Bloomberg. https:// www. 2022. On index investing. Journal of Financial Economics 145 bloom berg. com/ news/ artic les/ 2022- 04- 07/a- 430- billi on- habit- (3): 665–683.got-j apan-s-c entra l-b ank-h ooked-o n-e tfs#x j4y7v zkg. Accessed Da, Zhi, and Sophie Shive. 2018. Exchange traded funds and asset 27 June 2022. return correlations. European Financial Management 24 (1): Lewis, Jeffrey B., and Drew A. Linzer. 2005. Estimating regression 136–168. models in which the dependent variable is based on estimates. Efremidze, Levan, Darrol J. Stanley, and Clemens Kownatzki. 2021. Political Analysis 13 (4): 345–364. Entropy trading strategies reveal inefficiencies in Japanese stock Lim, Kian-Ping., and Chee-Wooi. Hooy. 2010. The delay of stock market. International Review of Economics & Finance 75: price adjustment to information: A country-level analysis. Eco- 464–477. nomics Bulletin 30 (2): 1609–1616. Fama, Eugene F. 1970. Efficient market hypothesis: A review of theory Linh, Nguyen T. 2021. The impact of the Bank of Japan’s exchange and empirical work. Journal of Finance 25 (2): 28–30. traded fund and corporate bond purchases on firms’ capital Fujikawa, Megumi. 2021, March 7. The bank of Japan is sitting on structure. Working Paper No. DP21-1. Research Center for Eco- a pile of profit but isn’t sharing the wealth. Wall Street Journal. nomic and Social Risks, Institute of Economic Research, Hitot- https:// www. wsj. com/ artic les/ the- bank- of- japan- is- sitti ng- on-a- subashi University. https:// www. efmae fm. org/ 0EFMA MEETI pile-of- pr ofit-but- isnt- shar ing- t he-w ealth-11615 118413 . Accessed N G S / E F M A % 2 0 A N N U A L % 2 0 M E E T I N G S / 2 0 2 1- L e e d s / 27 June 2022.papers/ EFMA% 202021_ stage- 2049_ quest ion- Full% 20Pap er_ Gârleanu, Nicolae, and Lasse Heje Pedersen. 2018. Efficiently inef-id- 101. pdf. ficient markets for assets and asset management. The Journal Lo, Andrew W. 1991. Long-term memory in stock market prices. of Finance 73 (4): 1663–1712. Econometrica 59 (5): 1279–1313. Glosten, Lawrence, Suresh Nallareddy, and Yuan Zou. 2021. ETF Lo, Andrew W. 2004. The adaptive markets hypothesis. The Journal activity and informational efficiency of underlying securities. of Portfolio Management 30 (5): 15–29. Management Science 67 (1): 22–47. Lorie, James H., & Mary T. Hamilton. 1973. The Stock Market: Grossman, Sanford J., and Joseph E. Stiglitz. 1980. On the impos- Theory and Evidence. Irwin: Homeland, IL. sibility of informationally efficient markets. The American Eco- Mundey, S. (2017). Asian markets in 2017: five moves and trends. nomic Review 70 (3): 393–408. Financial Times. https:// www. ft. com/ conte nt/ 02585 908- ec41- Gunji, Hiroshi, Kazuki Miura, and Yuan Yuan. 2021. The effect of the 11e7- 8713- 513b1 d7ca8 5a. Accessed 24 October 2022. Bank of Japan’s ETF purchases on firm performance. Available Nagayasu, Jun. 2003. The efficiency of the Japanese equity mar - at SSRN 3798098. ket. In: J. Jay Choi, and Takato Hiraki. (Eds.) The Japanese A. Charteris, C. A. Steyn Finance: Corporate Finance and Capital Markets. Emerald, Teverovsky, Vadim, Murad S. Taqqu, and Walter Willinger. 1999. A Bingley: 155-171. critical look at Lo’s modified R/S statistic. Journal of Statistical Noda, Akihiko. 2016. A test of the adaptive market hypothesis using Planning and Inference 80 (1–2): 211–227. a time-varying AR model in Japan. Finance Research Letters Vidal-Tomás, David. 2022. All the frequencies matter in the Bitcoin 17: 66–71. market: An efficiency analysis. Applied Economics Letters 29(3): Petrov, Alexander. 2017. ETFs in Monetary Policy: A Case Study 212–218. of Japan. Policy Research Unit. State Street Global Advisors. Woolley, Paul, and Ron Bird. 2003. Economic implications of pas- https:// swfi- cdn. sfo2. cdn. digit aloce anspa ces. com/ partn erCon sive investing. Journal of Asset Management 3 (4): 303–312. tent/ ID115 48- INS T - 8354_ ETFs- in- Mone t ar y - P olicy_ w eb_ Zou,Yuan. 2019. Lost in the rising tide: ETF flows and valuation. FINAL. pdf. Working paper, Columbia Business School. https:// www. ander Rejichi, Imen Zgueb, and Chaker Aloui. 2012. Hurst exponent son. ucla. edu/ docum ents/ ar eas/ fac/ accou nting/ wor ki ngpap ers/ behavior and assessment of the MENA stock markets effi-worksh op%2 0pape rs/Y uan%20 Zou%20 JMP_J an22%2 0%28 002% ciency. Research in International Business and Finance 26 (3): 29. pdf. 353–370. Samuelson, Paul A. 1998. Summing upon business cycles: Opening Publisher's Note Springer Nature remains neutral with regard to address. Conference Series - Federal Reserve Bank of Boston jurisdictional claims in published maps and institutional affiliations. 42: 33–36. Shen, Steven, Sayad Baronyan, Vikram Srimurthy, and Cameron Brandt. 2021. A rising tide lifts some (Japanese) boats: The Bank of Japan's ETF purchases and their impact on market Ailie Charteris is an associate professor in the Department of Finance signals for individual stocks. Available at SSRN 3803864. and Tax, University of Cape Town, South Africa. Her research interests Shu, C., He, D., Wang, H., & Dong, J. 2015. The influence of Chi- include capital markets, asset pricing, exchange traded funds, passive nese and US financial markets on Asia-Pacific. Bank of Interna- investing and behavioural finance. tional Settlement (BIS) Papers, (82i). https://www .bis. or g/publ/ bppdf/ bispa p82i_ rh. pdf. Conrad Alexander Steyn is a Masters’ candidate in the Department Sushko, Vladyslav, and Grant Turner. 2018, March. The implications of Finance and Tax, University of Cape Town, South Africa and an of passive investing for securities markets. Bank of International analyst at a private equity firm. His research interests include passive Settlements (BIS) Quarterly Review. https:// www. bis. org/ publ/ and active investing and the implications of the changing dynamics qtrpdf/ r_ qt180 3j. htm. between the two investment approaches.
Journal
Journal of Asset Management – Springer Journals
Published: May 1, 2023
Keywords: Exchange traded funds; Market efficiency; Long-range dependence; Price delay; Monetary policy; Asset purchase programmes; E52; E58; G10; G14