Long-Term post-merger announcement performance. A case study of Australian listed real estate

3.1 Buy-and-hold ARs

The results presented by Langetieg (1978) brought forward the focus on the possible disadvantages of the models employed in estimating postmerger performance. Barber and Lyon (1997), Kothari and Warner (1997) and Lyon et al. (1999) have criticised the methodologies employed in the early post-M&A studies and recommend the use of a buy-and-hold methodology. Barber and Lyon (1997, p. 342) argue that earlier methods to calculate long-term excess returns ‘are conceptually flawed and/or lead to biased test statistics’. Early results from employing BHAR methodology returned insignificant excess returns over the 3-year postannouncement period (Higson and Elliott, 1998; Mitchell and Stafford, 2000).

However, subsequent international studies observed significant negative BHARs post-M&A announcement. In the UK, Cosh and Guest (2001), Sudarsanam and Mahate (2003) and Conn et al. (2005) observed negative and significant BHARs over the 3- to 4-year postannouncement period, ranging from −7.50 percent to −16.30 percent. Betton et al. (2007) examined US M&As and found significant negative BHARs of −21.90 percent over the 5-year postannouncement period. Ang et al. (2008) also observed negative and significant BHARs of −5.02 percent. Following on from these US studies, Bouwman et al. (2009) detected negative and significant BHARs of −7.22 percent. Finally, Bessembinder and Zhang (2013) found acquirers earn negative and significant BHARs of −7.90 percent over the 5-year postannouncement period.

3.2 Three-factor model

Fama and French (1993) propose that the three-factor asset-pricing model be employed to examine long-term abnormal performance because the returns can be described by the size and book-to-market factors. Mitchell and Stafford (2000, p. 288) argue ‘the systematic errors that arise with imperfect expected return proxies – the bad model problem – are compounded with long-horizon returns’. In addition, Mitchell and Stafford (2000) discuss that the BHAR method ignores any cross-sectional dependence of the over-lapping excess returns of individual event firms.

Gregory (1997) employed the three-factor model to examine UK M&As. Results showed bidders earn negative and significant mean monthly ARs of −0.75 percent. This result was supported by later studies examining US acquirers (Mitchell and Stafford, 2000; Gaspar et al., 2005) and Canadian M&As (André et al., 2004). All three studies report negative and significant mean monthly ARs ranging from −0.20 percent to −0.75 percent.

However, Moeller et al. (2004) observed insignificant non-negative monthly ARs. Similarly, Croci et al. (2010) study of UK M&As from 1990 to 2005 also reported insignificant mean monthly ARs. Subsequent studies employing the three-factor model display mixed results. For example, Bouwman et al. (2009), Dutta and Jog (2009) and Latorre et al. (2014) all observed positive and significant mean monthly ARs, ranging from +0.52 percent to +0.70 percent, in their studies of the United States, Canadian and Spanish M&A markets, respectively. In contrast, Alexandridis et al. (2006) reports significantly negative excess returns of −1.02 percent in their UK study and Alexandridis et al. (2012) finds US bidders earn negative and significant mean ARs of −0.25 percent.

A number of studies identify the possible differences in methods and therefore to test the robustness of their results the studies employed both methods. Conn et al. (2005) observed negative and significant postannouncement performance across both methods.3 While Croci et al. (2010) and Datta et al. (2013) detected, insignificant excess returns in both methodologies. Dutta and Jog (2009) present positive and significant ARs from the three-factor model, but insignificant positive BHARs. Finally, Bouwman et al. (2009) produced completely contrasting results. Three-factor average monthly ARs were +0.66 percent and significant (equating to a cumulative average AR of +15.84 percent over the 2-year event period), compared to a negative and significant BHAR of −7.22 percent. Unfortunately, the authors did not provide an explanation for the contrasting results. They did, however, cite Loughran and Ritter (2000) that ‘since different methods have different powers of detecting abnormal performance, there should be differences in AR estimates across different methodologies’ (Bouwman et al., 2009, p. 654).

3.3 Australian evidence

Dodd (1976) conducted the first study of Australian post-M&A performance. He examined the 1-year postannouncement CARs for successful acquirers from 1960 to 1970. Results were consistent with early international studies and showed bidders earned negative and significant ARs. Similarly, McDougall and Round (1986), employing a market model method, showed acquirers earn negative and significant CARs of −18.0 percent over the 2-year postevent period. McDougall and Round (1986, p. 189) concluded that M&As ‘appear to have been caused by so-called managerial motives, or by the desire to develop or enhance market power’.

Subsequent studies employing market model methodology have also observed consistent results for bidders post-M&A. Both Bellamy and Lewin (1992) and Nankervis and Singh (2012) find acquires earn negative and significant CARs in the long-term postperiod. In contrast, Dullard and Hawtrey (2008) find bidders earn positive and significant CARs of +10.57 percent over the [0, +36] event window. The authors conclude that M&As in Australia improve the share price performance of acquirers. However, Dullard and Hawtrey (2008) noted their results may be subject to size bias and the market model employed for the long-term analysis.

The study by Brown and da Silva Rosa (1998) was the first Australian investigation to employ the BHAR methodology. The study found that after controlling for the bias described by Barber and Lyon (1997), acquiring firms earn insignificant excess returns over the [+6, +36] postevent period. Brown and da Silva Rosa (1998, p. 36) conclude that ‘the long-term performance of the acquiring firms in the postmerger period is consistent with the proposition that the market for corporate control is informationally efficient’.

Duong and Izan (2012) also employed BHAR methodology to examine Australian M&As from 1980 to 2004. The study found no significant evidence of postacquisition underperformance. However, the authors extended the analysis to examine the impact merger waves may have on bidder performance. Results showed that M&As occurring during a merger wave resulted in negative and significant long-run BHARs of −2.53 percent over the [0, +18] event window. In contrast, M&As occurring outside a merger wave period returned BHARs of +4.58 percent; however, the result was not significant. The authors concluded that acquisitions made during merger waves provide evidence for the hubris and/or agency motives for M&As.

Australian research results show some consistency with international studies, in that the choice of methodology can have an impact on the observed ARs. Studies that have estimated CARs have found, on average, bidder's wealth is adversely impacted post the M&A. However, studies utilising BHAR methodology do not find sufficient evidence to suggest the postbid phenomenon holds for Australian M&As.4

3.4 REIT postannouncement performance

Research into REIT postannouncement shareholder performance has been limited to only two papers, both of which investigated the US REIT market. Sahin (2005) was the first researcher to investigate the long-term performance utilising both BHAR and three-factor model methodologies. Both methods returned insignificant excess returns results over the 3-year postannouncement period.5

Following on from Sahin (2005), Campbell et al. (2009) conducted the second investigation into the long-term wealth effects within the US REIT sector. Campbell et al. (2009, p. 105) tested if ‘the anomaly of postmerger underperformance observed in conventional firms applies to the case of REITs’. In contrast to Sahin (2005), the study observed negative and significant BHARs of −9.9 percent over the 5-year postacquisition period. However, both the one- and 3-year excess returns were statistically insignificant. The authors concluded that the results confirm that postacquisition underperformance of US REITs is consistent with those observed in more general corporate finance studies and provide support for the hubris and/or agency motive.

Examination of the literature makes it difficult to draw any strong conclusions. On average, the results across varying methods of measuring long-term returns suggest market inefficiency. However, Fama (1998) argues that market efficiency should not be discarded. He suggests that an efficient market produces different types of events that individually cause share prices to over or under-react. The under-reaction will be approximately as frequent as the over-reaction in an efficient market. If these anomalies are split randomly between each other, they are consistent with market efficiency.

Furthermore, it appears that the long-term return anomalies that suggest market inefficiency are sensitive to a number of factors. These include the methodology (Martynova and Renneboog, 2008; Bessembinder and Zhang, 2013), the different markets examined (e.g. Unite States versus Europe), the different time periods studied and possibly the different data sets. To add to the existing knowledge and shed more light on postannouncement shareholder performance, we employ both the BHAR and three-factor methodologies. In addition, by examining the postannouncement performance within the same industry decreases possible ‘inaccuracies resulting from missing pricing factors that may have varying effects across industries’ (Campbell et al., 2009, p. 108).

4.1 Data collection

To examine the long-term postannouncement performance of A-REIT acquirers over the period of January 1996 to December 2012, we calculate the excess returns over three-event windows; 1, 2 and 3 years. The screening process employed was:

This screening resulted in 65 observations for the 1-year event window. The 2-year window comprises 49 observations, while the 3-year period contains 35 observations. The differences in the number of observations across the three periods are due to the first filtering requirement. For example, if a bidder makes an announcement in July 2001 and another in October 2002, the 2001 announcement would only be included in the 1-year excess return calculations. The ARs would not be calculated for the 2- and 3-year periods, because they overlap the October 2002 announcement. However, the October 2002 announcement has no overlapping postannouncement periods, and therefore, excess returns would be calculated for the 1-, 2- and 3-year event windows.

In relation to the buy-and-hold methodology, the study also required the construction of a matching/control portfolio. The control firms are selected from the A-REIT sector and are subject to the same filtering processes described above, with the additional constraint that the control firm is not involved in a M&A during the sample period.

4.2 Methodology

This section presents the two methodologies employed in the postannouncement study. The first method calculates the buy-and-hold excess returns. The second method is based on the Fama and French (1993) three-factor model, which estimates the average monthly ARs. Gregory (1997) and Limmack (1997) discuss that the choice of event study methodology to access long-term performance can have an important impact on the level of ARs. Utilising two different methods will enable the study to test the robustness of the post-M&A performance of A-REIT acquirers. As identified earlier, this is the first Australian postannouncement study, to the authors' knowledge, to employ the three-factor model. Consistent with short-term event study investigations, the study must firstly identify the date of the M&A announcement. Following Campbell et al. (2009), the event month [t = 0] for the study is set as the month end in which the M&A is announced.

4.3 Buy-and-hold abnormal returns

To calculate BHARs, the study needs to firstly identify an appropriate benchmark (nonevent) control sample. Barber and Lyon (1997) and Lyon et al. (1999) argue that the choice of the control sample has an important impact on the empirical power and the test statistics resulting in biased outcomes. Following Lyon et al. (1999) and Campbell et al. (2009), this study identifies all nonevent A-REIT firms available for the study period. The firms are then ranked on market size and then market-to-book value. Market size is calculated as the number of shares on issue times the closing share price one calendar month before the event occurrence. Market-to-book value is calculated as the firm's market value, divided by the book value of the firm reported in their annual report prior the M&A announcement.

The event firm is then matched to three nonevent firms that comprise the control portfolio that is closely equivalent to size and market-to-book value and the control sample is then matched to the event firm for the full buy-and-hold period.7 Utilising control portfolios from the same industry decreases possible ‘inaccuracies resulting from missing pricing factors that may have varying effects across industries’ (Campbell et al., 2009, p. 108).

The BHARs are then calculated as the difference between the compounded monthly returns over the event period of the event firm less the control portfolio:

(1)

Where:

(2)

BHAR_i is the BHAR for event firm i over the time period T; R_i,t is the monthly total return for event firm i in month t; R_j,t is the monthly total return for nonevent firm j in month t; n is the number of nonevent firms that make up the control portfolio; and BHR_RP,i is the arithmetic average compounded monthly return of the control portfolio.

The statistical significance of the BHARs is calculated for each event window. This study employs two statistical tests to access the robustness of the BHARs significance. The first is a traditional test statistic:

(3)

where is the sample mean BHAR calculated over time period T; σ(BHAR_T) is the cross-sectional sample standard deviation; and N is the number of event observations.

Barber and Lyon (1997) provide evidence that long-term BHARs are positively skewed resulting in negatively biased t-statistics. To adjust for the potential skewness bias when BHARs are calculated using control portfolios, Lyon et al. (1999) advocate the use of a skewness-adjusted t-statistic:

(4)

where

(5)

and,

(6)

The calculation of is an estimate of the coefficient of skewness, and is the standard t-statistic of Equation 3 (Lyon et al., 1999).

4.4 Three-factor model

The second methodology employed to identify postannouncement performance is the three-factor model developed by Fama and French (1993). Asset-pricing research has shown that the three-factor model has strong empirical support in determining share price movements (Brailsford et al., 2012). The three-factor model is implemented by regressing the postannouncement monthly excess returns of the acquiring firm against a market factor, a size factor and a book-to-market factor:

(7)

where R_i,t is the return on security i in month t; R_f,t is the return on the 180-day Bank Accepted Bill Rate in month t; α_i is the intercept term; RMRF_t is the excess return on the value-weighted market index; SMB_t is the return difference between the portfolios of small and large firms in month t; HML_t is the return difference between the portfolios of high and low book-to-market firms in month t; and ε_i,t is the standard error term.

The intercept, α_i, is the variable of interest in the model and measures the mean monthly AR of the event firm. A positive intercept indicates the sample firm has outperformed, after controlling for market, size and book-to-market factors (Barber and Lyon, 1997).

Because the focus of the study is on the A-REIT sector, calculation of the factors is developed from the A-REIT universe. Excluding conventional firms from the estimation of the factors removes any possible noise within the factors that may not be relevant to the A-REIT sector. The market index employed is the S&P/ASX200 A-REIT index. This index captures the price movements of those entities classified as A-REITs on the ASX.

The calculation of the SMB and HML factors follows Brailsford et al. (2012) and Fama and French (1993). Mitchell and Stafford (2000) establish that there is cross-sectional correlation of individual event firms when estimating long-term ARs. Bouwman et al. (2009) discuss that by employing the three-factor model automatically accounts for cross-sectional correlations in the portfolio variance at each point in time. In addition, the three-factor model does not require size and book-to-market data for the event firms, as is the case with the BHAR method does (Barber and Lyon, 1997). Barber and Lyon (1997) demonstrate large firms or firms with low book-to-market values may have common share returns that more closely follow the returns of small firms or high book-to-market firms.

5.1 Buy-and-hold abnormal returns

The BHAR results for A-REIT acquirers are presented in Table 1. Panel A displays the results for the full sample period. A-REIT bidders earn negative and significant BHARs over the 2- and 3-year periods of −8.21 percent and −12.27 percent, respectively. These outcomes imply the existence of hubris and/or agency issues. However, the influence of the financial crisis needs to be examined before any strong conclusions can be drawn. Panels B and C of Table 1 partitions the BHARs for announcements occurring prior and post-December 2007 to investigate any differences between the two periods. Pre-GFC results show that A-REIT bidders earn positive and significant BHARs of +3.77 percent over the 1-year period. Both the 2- and 3-year periods are positive but insignificant. This result provides support for the synergy motive for A-REIT M&As in the pre-GFC period.

In contrast, the post-GFC BHARs [2008–2012] are highly significant and negative across all three periods. As shown in Table 1, these range from −16.07 percent for the 1-year window to −23.86 percent for the 3-year period. This outcome highlights the structural change in the A-REIT sector as a result of the GFC. These results support the claims of Betton et al. (2007) who suggest that negative postbid ARs may be due to a negative industry shock. However, one would expect a shock like the GFC to impact nonmerged firms as well. This outcome may be due to the high uncertainty and volatility in the A-REIT sector during this period, making it more difficult to integrate the assets of the target firm and achieve any possible synergistic benefits.

In addition, Rose (2011) highlights that Australian commercial property values were exceptionally high in 2006/2007. With the onset of the GFC, a large number of A-REITs re-valued their property portfolios. In 2009, approximately 40 percent of A-REITs reported property valuation write-downs of greater than 20%, with an average sector write-down of 16 percent (BDO, 2009). This suggests that, in addition to the problems with integrating the targets assets, over-payment for property assets prior to the GFC and subsequent revaluations compounded the under-performance.

Figure 1 presents the indexed returns for the acquirers and nonacquirers portfolio over the study period. The figure shows that the bidders portfolio closely tracks the control portfolio from 1996 to October 2005. Between October 2005 and October 2007, the bidders' portfolio slightly outperforms the control sample. However, it is clearly evident that after December 2007 both portfolios sharply declined with the acquirers' sample falling at a greater rate. Figure 1 highlights a structural break in the sector as a result of the financial crisis.

The impacts of the GFC appear to continue until March 2009, after which both the acquirers and nonbidder control portfolios begin to recover. BDO (2010) observed a dramatic fall in A-REIT volatility in 2009 compared to 2008. PCA (2010) suggest that the recovery of the sector was due to A-REITs focusing on their core business of rental returns and lower risk exposures. A large number of A-REITs rebalanced their balance sheets through assets sales and capital raisings during the GFC (Newell and Peng, 2009; Dimovski and O'Neill, 2012). This restructuring resulted in a large fall in leverage levels, decrease in exposure to international property markets and a reduction in higher risk activities (e.g. property development), resulting in a decrease in the risk profile of the sector (BDO 2010).

Figure 2 shows the A-REIT M&A announcements by year employed in the study. It can be seen that consolidation activity within the sector gathered momentum in 1999. However, there were no M&A observations from 2008 to 2009 as the impacts of the financial crisis took effect on A-REITs. Martynova and Renneboog (2008) identify that M&A activity ‘is usually disrupted by a steep decline in stock markets and a subsequent recession’. This study therefore concludes that there are three distinct subperiods of activity within the overall analysis. These are pre-December 2007, January 2008 to March 2009 and April 2009 onwards.

Given the observed structural breaks in the data set, the post-December 2007 subsample was further divided into two periods. Any postannouncement ARs calculations that extend over the period of January 2008 to March 2009 are classified as being during the GFC subsample, while calculations occurring after March 2009 are classified as the post-GFC subsample. Table 2 presents the results for these two subsamples.

Panel A shows the BHAR results for postannouncement returns that extend past December 2007. The excess returns are negative and significant across all three-event windows, ranging from −20.27 percent in the 1-year model to −29.53 percent for the 3 years. This result is consistent with Duong and Izan (2012) who provides evidence that M&As occurring late in a merger wave earn lower excess returns. Panel B results show that acquirers earn negative and significant BHARs in the 2-year window although one outlier drives this significance. Real Estate Capital US Property Trust was identified to have a BHAR value of −120.02 percent, more than three standard deviations away from the mean BHAR. After removing this outlier observation, the BHAR result was a negative but insignificant value of −14.47 percent. Both 1- and 3-year periods are insignificant.8

Comparing Panel A and B results suggests that the negative BHARs in the 2008 to 2012 subperiod of Table 1 are being driven by postannouncement returns that extend across the financial crisis period. Panel C displays the BHAR results over the full study period, excluding the GFC. The excess returns are insignificant across all three-event periods.9

Results for A-REIT bidders prior the onset of the GFC suggest that the reason for M&As was the synergy motive. One-year BHARs are positive and significant, while the 2- and 3-year BHARs are positive, but insignificant. Conversely, it is evident that A-REIT acquirers were severely punished by investors after the onset of the GFC. The results from the BHAR modelling show that, after accounting for the GFC, the A-REIT sector is consistent with informational efficiency.

5.2 Three-factor model abnormal returns

Three-factor model ARs for A-REITs is presented in Table 3. Panel A displays the full study period, A-REIT bidders exhibit negative and significant ARs in the 2-year event window. The intercept is negative, but insignificant, for the 1- and 2-year periods. This outcome varies slightly from the BHAR results, which displayed significantly negative BHARs over the 3-year model.

Results for bidders separated by pre- and post-GFC are displayed in Panels B and C of Table 3, respectively. Pre-GFC ARs are positive across all three-event periods, and both the 1- and 2-year models are statistically significant. This result supports the BHAR conclusions that prior to the onset of the GFC, A-REIT M&As were driven by the synergy motive. Post-GFC results show, consistent with the BHAR approach, that acquirers earn negative and significant ARs across both the 1- and 2-year event periods. However, in contrast, the 3-year window is now insignificant.

We further examine the excess returns over the GFC period. The results are presented in Table 4. Consistent with the BHAR model, acquirers earn negative and significant excess returns during the GFC period in the 1-year event window. Conversely, the 2- and 3-year periods are insignificant. Post-March 2009 long-term ARs are comparable with Table 2, with only the 2-year event period displaying a negative and significant intercept. However, again the significance disappears after removing the outlier observation. Panel C shows the ARs for the full sample, excluding the GFC period. As with the BHAR results, both the 2- and 3-year periods are insignificant. It is observed, however, that the 1-year period AR is positive and significant. The mean monthly AR over the twelve-month period is +0.4 percent.

Overall, the results from the three-factor model support the BHAR findings. Prior to the GFC, there is evidence that A-REIT acquisitions were not value destroying for shareholders. However, it is observed that the period of the financial crisis is a significantly different period in the evaluation of the long-term performance of A-REIT bidders. The post-March 2009 period shows some indication that the M&A market in the A-REIT sector is recovering, possibly due to the refocussing of A-REITs on their core business. However, due to a low number of observations, it is difficult to obtain a strong conclusion for this subperiod.