相关性下降带来更多超额收益机会

Transcription

1 2012 年 2 月 14 日 欧洲 策略焦点 相关性下降带来更多超额收益机会 证券研究报告 因宏观风险主导市场,2011 年股票间相关性达到了新的峰值 但自年初以来, 欧洲和美国市场一个月股票相关性大幅下降 尽管其中部分原因是季节性因素, 但我们认为欧元区主权风险减弱及宏观前景改善是关键因素 鉴于股市风险溢价上升, 相关性较长期均值可能仍将维持高位 但相关性较 2011 年有所下降应有助于选股, 卖出 EURO STOXX 50 相关性投资组合更具吸引力 1 月份股票相关性大幅下降继 2011 年下半年达到创纪录水平后, 今年 1 月份欧洲和美国市场 1 个月股票相关性大幅下降 此次下降涉及面广泛, 但在 1 月份这种情况很常见, 而且该趋势可能很快扭转 波幅及股市风险溢价下降, 欧元区尾部风险减弱是关键推动因素自 12 月以来, 作为股票相关性主要推动因素的波幅和股市风险溢价下滑, 这带动了相关性的下降 尽管我们认为股市风险溢价与长期均值相比可能将维持高位, 但鉴于欧元区尾部风险下降及宏观前景改善, 我们认为相关性应低于 2011 年水平 选股及卖出 EURO STOXX 50 相关性投资组合更具吸引力相关性下降通常有助于选股, 而且 1 月份板块内部表现差异扩大 因当前 12 个月隐含水平接近 2011 年高点, 卖出 EURO STOXX 50 相关性投资组合更具吸引力 欧元区尾部风险推动 2011 年下半年股票相关性较高 STOXX Europe 个月相关性与欧元 / 美元风险逆转比较 Christian Mueller-Glissmann, CFA +44(20) christian.muellerglissmann@gs.com 高盛国际 彼得 欧品海默 +44(20) peter.oppenheimer@gs.com 高盛国际 Sharon Bell, CFA +44(20) sharon.bell@gs.com 高盛国际 Gerald Moser +44(20) gerald.moser@gs.com 高盛国际 Anders Nielsen +44(20) anders.e.nielsen@gs.com 高盛国际 month correlation EUR/US$ risk reversal (RHS) Matthieu Walterspiler +44(20) matthieu.walterspiler@gs.com 高盛国际 资料来源 : 高盛全球经济 商品和策略研究 高盛与其研究报告所分析的企业存在业务关系, 并且继续寻求发展这些关系 因此, 投资者应当考虑到本公司可能存在可能影响本报告客观性的利益冲突, 不应视本报告为作出投资决策的唯一因素 有关分析师的申明和其他重要信息, 见信息披露附录, 或参阅 由非美国附属公司聘用的分析师不是美国 FINRA 的注册 / 合格研究分析师 高盛集团 高盛全球经济 商品和策略研究

2 Salvation from correlation more alpha opportunities Equity correlations have fallen since the beginning of the year In 2011, equity correlations (average pair-wise correlations of daily returns) in most developed equity markets reached new record levels owing to investor concerns over global growth and tail risks related to the Eurozone sovereign crisis. This was accompanied by record levels of crossasset correlations and correlations across equity markets. The peaks were reached around October, when STOXX Europe 600 one-month correlations were above 0.85 we highlighted the extreme levels of correlation at this time, suggesting that they should ease (Strategy Espresso: Have correlations across equities peaked?, October 13, 2011). Equity correlations matter a lot for stockpickers trying to generate alpha, derivatives markets pricing correlation risk and in general for equities as an asset class, as in high correlation environments, there are limited opportunities for diversification. 1 Exhibit 1: Correlations across equities in Europe dropped significantly in January STOXX Europe 600 correlation 1-month 3-month 12-month Jan-10 Apr-10 Jul-10 Oct-10 Jan-11 Apr-11 Jul-11 Oct-11 Jan-12 Since the start of 2012, one-month correlations for the STOXX Europe 600 have dropped materially, similar to January The same is true in the US, where correlations peaked at 0.85 in August and are now at only 0.15 correlations in Europe and the US have moved closely together since 2010, suggesting similar drivers (Exhibit 2). Does the recent drop indicate salvation from record high equity correlations in 2011? Looking at a longer history and the five-year rolling average of 12-month STOXX Europe 600 correlations shows that correlations have been trending up not only since the financial crisis, but since the tech bubble (Exhibit 3). As discussed in our previous research on correlations 2, we find that the level of volatility and the equity risk premium combined can explain a large part of the variation in equity correlations over the past 25 years. While the increased index and macro investing through ETFs and derivatives might have contributed, we think the primary driver of equity correlations is the perception and pricing of equity risk the increase in equity correlation since 2000 coincided with a major shift up in the equity risk premium from unsustainably low levels during the tech bubble. 1 See Portfolio Strategy: Equity correlations and the ERP; Part 2: Opportunities, September 8, See Portfolio Strategy: Equity Correlations and the ERP; Part 1: The Drivers, July 22, 高盛全球经济 商品和策略研究 2

3 Exhibit 2: Correlations in Europe and the US fell together 1-month correlations Exhibit 3: Correlations have trended up in recent years 12-month correlations 0.90 STOXX Europe 600 S&P 500 STOXX Europe month correlation 5-year average both within and across sectors biggest decline for defensives Since 2012, the declines in equity correlations have been broad-based correlations both across and within sectors have declined in a similar fashion (Exhibit 4). While historically, intra- and inter-sector correlation did show disconnects, especially during the tech bubble, since 2010, one-month correlations have moved together very closely. Looking at intrasector correlations by broader sector groups reveals that commodities and financials remain most correlated, but intra-sector correlations have dropped across sectors. Commodities and financial sectors generally tend to have higher correlations on average, at 0.58 and 0.51 respectively, over the past ten years. Commodities sector returns in particular tend to be closely linked to commodity prices. Defensives and cyclicals have had lower one-month correlations in the past ten years, on average, at 0.39 and 0.43 respectively. Exhibit 4: Correlation declines have been broad-based STOXX Europe month correlation Exhibit 5: Commodities stocks most correlated 1-month correlations for STOXX Europe 600 sectors STOXX Europe 600 Avg. intra-sector (RHS) Inter-sector (RHS) 0.90 Commodity Defensives Cyclicals Financials 0.00 高盛全球经济 商品和策略研究 3

4 Seasonal drops in correlation at the start of the year are common When looking at one-month rolling correlations, we have to consider seasonality. Exhibit 6 shows the average level of equity correlations for different months in the past ten years and since Since 2002, and even more since 2008, equity correlations have tended to drop significantly in January, especially compared with December, when correlations tend to be higher. To some extent, the drop in correlations can be linked to performance in January being generally strong and rotations taking place at the beginning of the year (see Strategy Espresso: January performance: Part 1 Rotation, January 3, 2012 and Strategy Espresso: January performance: Part 2 Seasonality, January 4, 2012). In February, May, August and November, equity correlations tend to be lower; most European companies report their earnings in those months and company-specific risk can play a more important role. Exhibit 6: In January and during earnings season, equity correlations tend to be lower One-month correlation for the STOXX Europe 600 Since 2002 Since January February March April May June July August September October November December Lower volatility has helped ease correlation levels in 2012 so far Not only have correlations dropped significantly since the beginning of the year, but realised volatility has also fallen close to the levels of January 2011, both for the index and on average for single stocks (Exhibit 7). In times of low volatility, correlations also tend to be lower. Equity correlations and volatility tend to be closely linked, especially for shorter periods (one-month) for example, we find a close link between one-month correlation and volatility for the S&P 500 (Exhibit 8). This is usually because during equity market corrections, which tend to result in volatility spikes, equities are closely correlated. Based on our US options analysts economic model for S&P 500 realised volatility, one-month volatility should be trending in the 12%-15% range given current levels of ISM, consumer spending, and employment data (see The Buzz: Where are we in the vol cycle?, January 30, 2012). They highlight that from a business cycle perspective, realized volatility has typically averaged 13.7 in months when the ISM was between 50 and 55 and 14.2 in quarters when GDP growth rates were 0%-2.5%, which is a range consistent with our economists real GDP forecast of 2.1% for the US in Current realised levels for the S&P 500 are only 8.3%, suggesting small increases from current levels. STOXX Europe 600 one-month volatility is at 14% historically, realised volatility of the S&P 500 and STOXX Europe 600 has been relatively close. 高盛全球经济 商品和策略研究 4

5 Exhibit 7: Lower volatility tends to result in lower STOXX Europe month correlation vs. volatility Exhibit 8: correlations, at least for shorter periods Changes in S&P month correlation vs. vol. (since 1989) Correlation Index (RHS) Avg. single stock (RHS) y = x R² = month correlation month volatlity but high equity risk premia might keep correlations elevated However, as highlighted in our previous research, equity correlations and volatility can have significant disconnects over longer periods (12 months); for example, during the tech bubble, volatility remained comparably high but correlations reached all-time lows. Also, in recent years, correlations have regularly been higher than would be expected based on the levels of realised volatility. We think the elevated equity correlations since 2007 can be explained by the increase and continued high level of the equity risk premium (ERP) in developed markets, which in turn can be linked to the weak post-crisis recovery. Exhibit 9: Correlations have increased alongside the ERP, which remains elevated STOXX Europe month correlation vs. European equity risk premium STOXX Europe month correlation European ERP (RHS) The ERP is an aggregate measure of macro risk for equities it is the additional return investors require for holding equities on top of the risk-free rate (as measured by ten-year bond yields). A 高盛全球经济 商品和策略研究 5

6 very low risk premium indicates that perceived risk is low, while a high risk premium indicates investor concerns over growth or changes in risk aversion. Clearly, periods in which the equity risk premium shifts are likely to increase correlations, as discount rate changes should affect all stocks alike. However, we find that the level of the equity risk premium matters as well, and indicates how equities are increasingly sensitive to changes in growth expectations and macro news. If the ERP is elevated and risk aversion is high, we also tend to see less M&A and LBO activity, which is likely to result in higher correlations. The current implied ERP is 7.5%, which is high versus what we would expect based on the current macro environment (the current macro benchmark ERP is 6.8%, which relates the level of the ERP to output gaps). We think a modest further fall in risk aversion is possible, especially if there is a smooth resolution to the Greek debt funding problems. However, over the medium term, Europe continues to face severe economic challenges and our economists expect growth to stay below trend through to the end of Thus, the risk premium is likely to stay elevated relative to history. A two-factor model for STOXX Europe month correlation based on volatility of 15% and an ERP of 7% results in a forecast of An improving macro outlook should help shift the focus on micro As stated earlier, the equity risk premium can in turn be linked to macro factors we find that a key driver tends to be output gaps, which can act as an aggregate measure of the health of the economy. 4 The post-crisis recovery in Europe and the US has been characterised by continued large output gaps and below-trend growth, which have in turn kept equity correlations elevated (Exhibit 10 we use the S&P 500, as historical macro data is more readily available). However, the recent improvement in US macro data suggests that S&P month correlations should decline when benchmarking against Initial Jobless Claims in particular, they appear too high (Exhibit 11). 5 Exhibit 10: Macro conditions and the growth outlook S&P month correlation vs. US output gap S&P month correlation US output gap (inverted, RHS) Source: Haver, Datastream, CBO, Goldman Sachs Global ECS Research Exhibit 11: tend to be key drivers of equity correlations S&P month correlation vs. initial jobless claims S&P month correlation Initial jobless claims (1-month avg., RHS) Source: Haver, Datastream, Goldman Sachs Global ECS Research Clearly, there has not been the same improvement in macro data in Europe, where the recovery, especially in the periphery, has been much weaker than in the US. In addition, equity correlations 3 For details of the two-factor model for equity correlations, see Portfolio Strategy: Equity correlations and the ERP; Part 2: Opportunities, September 8, See Forecasting returns: Fair Value Part 2, March 23, In the US, there has been a divergence recently between an improving labour market and still lacklustre real GDP growth, suggesting a lower potential GDP growth rate and that the output gap could be smaller than originally thought. Our economists estimate that the current US output gap is a little below 4% (see also US Economics Analyst: When Will the Economy Reach its Potential?, February 10, 2012) and thus smaller than that from the Congressional Budget Office (CBO) used in Exhibit 10. 高盛全球经济 商品和策略研究 6

7 can disconnect from the business cycle as a result of financial market shocks. The same is true for the implied ERP, which often overshoots the macro benchmark ERP, our estimated fair value based on the level of output gaps. Fading Eurozone sovereign risks could also lower correlations Financial market shocks due to increases in risk aversion can also cause spikes in equity correlations despite a stable macro outlook and narrowing output gaps. This was the case during the 1998 Russian default/ltcm crisis, the bursting of the tech bubble in 2002 and more recently, the Eurozone sovereign crisis. As shown in Exhibits 10 and 11, the current correlations are above the levels suggested by current macro conditions. This is because since 2010, there have been several shocks in financial markets that are likely to have contributed to higher correlations, in our view. In 1H2010, sovereign CDS spreads for Eurozone periphery countries rose significantly, resulting in the May sell-off in equities, which in turn pushed correlations higher (Exhibit 12). The tail risks related to a potential break-up of the Eurozone remained a key driver of equity markets in 2H2011. Exhibit 13 shows that STOXX Europe 600 one-month correlation has been closely linked to a EUR/US$ risk reversal (short put, long call both 25 delta), which prices the probability of a Eurozone break-up. Assuming progress on the stabilisation of the Eurozone, correlations might stay lower. Exhibit 12: Eurozone sovereign funding risks and STOXX Europe 600 correlation vs. Eurozone sovereign CDS Exhibit 13: Eurozone break-up concerns were a key driver STOXX Europe 600 correlation vs. EUR/US$ risk reversal 12-month correlation 1 year change avg. of periphery CDS (RHS) 700% 600% 500% 400% 300% 200% 100% 0% month correlation EUR/US$ risk reversal (RHS) % Alpha opportunities from stockpicking should increase Lower correlations tend to be supportive for stockpicking. Dispersion of returns, which quantifies the opportunity set for alpha, has started to increase since the beginning of the year (Exhibit 14). In January and February, there has been a significant increase in the dispersion of returns in some sectors (for example, related to M&A in Basic Resources and a rally in periphery banks), but also across sectors (Banks, Insurance and Autos outperforming the market by a large margin). Exhibit 15 shows that in January, the dispersion of returns was above the long-run median for most sectors. Dispersion of stock returns increases both with lower correlation and higher volatility, which is why it is difficult to dissect the influence of the two. If volatility and correlation spike together, 高盛全球经济 商品和策略研究 7

8 dispersion tends to increase, and theoretically, stockpicking should be successful. However, most of the dispersion is macro-driven (e.g. by sector, countries and styles) and thus not easy to capture through traditional stockpicking. The spread between dispersion of returns across sectors and within sectors allows a better assessment of whether it is more important to pick the right sector or pick the right stock this is driven by the level of correlation (Exhibit 16). With a sustained drop in correlations, intra-sector dispersion is likely to increase relative to that across sectors, which should be supportive for stockpickers. Exhibit 14: Dispersion of returns increased in January Dispersion of monthly returns Exhibit 15: Dispersion in sectors picked up in January Dispersion of monthly returns within sectors (since 1988) 21 Dispersion Dispersion (12-month average) 12-month correlation (RHS) % Banks Oil & Gas Retail Stoxx600 Travel & Leis. Cons. & Mat. Tech Basic & Res Indus. Gds & Svs Healthcare Autos & Parts Food & Bev Insurance PHHG Chemicals Media Financial Svs Telecoms Utilities Real Estate Current Long Run Median 5th Percentile/ 95th Percentile Source: Datastream, Goldman Sachs Global ECS Research. Source: Factset, Worldscope, Goldman Sachs Global ECS Research. Exhibit 16: Intra-sector dispersion could increase again Spread intra- vs. inter-sector dispersion vs. STOXX Europe month correlation Exhibit 17: Equity hedge funds suffered from correlations 12-month rolling HF performance vs. STOXX Europe month correlation Spread Spread ex TMT Correlation (RHS, inverted) % 50% 40% % 20% 10% 0% Source: Datastream, Goldman Sachs Global ECS Research. -10% -20% HFR Equity hedge fund Correlation (RHS) -30% Source: Datastream, Goldman Sachs Global ECS Research. Indeed, we find that long/short equity hedge funds struggle to deliver positive returns if correlations are high (Exhibit 17). As a result, in 2H2011, the broad HFR equity hedge index posted negative returns of -9.21%. This illustrates that lower correlations are likely to benefit equity long/short alpha strategies more broadly. 高盛全球经济 商品和策略研究 8

9 Selling EURO STOXX 50 correlation should provide a positive carry Another key opportunity that arises from falling correlations is selling EURO STOXX 50 correlation. 6 EURO STOXX month implied correlation is currently at 0.66, roughly in line with current realised levels (Exhibit 18). EURO STOXX month correlations are sticky and still close to their end-2011 peaks of one-month correlations fell to 0.37 in January. While we do not expect 12-month correlations to drop to similarly low levels, assuming fewer financial market shocks and further improvement in the macro backdrop, we believe realised correlation is unlikely to reach the previous peaks of Exhibit 18: Realised correlation was below that implied 12 months earlier for most of 2011 EURO STOXX month implied correlation vs. realised 1.00 Implied correlation lagged by 1 year 12-month realised 1-month realised 0.90 Jul-12 Jan-13 As realised EURO STOXX 50 correlations are sticky with increasing period length, current implied vs. realised spreads are still low for longer maturities (Exhibit 19). Only implied versus realised spreads for three-month EURO STOXX 50 correlations have increased and are now in line with the five-year average. If implied correlations stay at current levels, we expect implied versus realised spreads to increase, which could increase the positive carry for selling EURO STOXX 50 correlation. Lower correlations should also be supportive for selling index volatility, for example through call overwriting (see Finding alpha: A 16 year study of S&P 500 index overwriting, February 6, 2012). 6 For more details on equity correlation/dispersion trading see Portfolio Strategy: Equity correlations and the ERP Part 2: Opportunities, September 8, 高盛全球经济 商品和策略研究 9

10 Exhibit 19: Current implied vs. realised spreads remain low compared to the last five years EURO STOXX 50 implied vs. realised correlation spread (history for the last five years) Min Max +1 sd Average -1 sd Current M 6M 12M 24M 3Y A closer look at implied and realised levels separately indicates that implied correlation levels are well above five-year averages but below their highs of May 2011 (Exhibits 20 and 21). And while realised EURO STOXX 50 correlations for longer periods are still above their five-year average by more than 1 standard deviation, current EURO STOXX 50 onemonth correlations are 0.37, well below the five-year averages. Exhibit 20: Implied correlation is only slightly elevated EURO STOXX 50 implied correlation Exhibit 21: Realised correlations are off their peaks EURO STOXX 50 realised correlations Min Max +1 sd Average -1sd Current Min Max +1 sd Average -1 sd Current 3M 6M 12M 24M 3Y M 6M 12M 24M 3Y Implied EURO STOXX 50 correlation levels sticky throughout 2011 EURO STOXX 50 implied correlation was surprisingly sticky in the sell-off last summer. While in May 2010, EURO STOXX 50 implied correlation increased alongside both implied volatility and skew (Exhibit 22 and 23), it was relatively range-bound last summer, limiting the mark-to-market risk of EURO STOXX 50 dispersion trades. 高盛全球经济 商品和策略研究 10

11 Exhibit 22: Implied correlation lagged volatility EURO STOXX month implied correlation vs. volatility Exhibit 23: and skew in 2011 EURO STOXX month implied correlation vs. skew Implied Volatility (RHS) Implied correlation Normalised skew (RHS) This has been in marked contrast to the S&P 500, for which implied correlation spiked more last summer, and selling correlation subsequently appeared more attractive (Exhibit 24). As highlighted earlier, the stickiness comes from the fact that banks implied volatility for the EURO STOXX 50 spiked more than that for the index (Exhibit 25). As a result, the weighted average single stock implied volatility increased to a similar extent as the implied index volatility. 8 This has in turn limited mark-to-market risk for EURO STOXX 50 dispersion trades. Recently, the gap between EURO STOXX 50 and S&P 500 implied correlation has narrowed again, suggesting that selling correlation is equally attractive in both markets. Exhibit 24: S&P 500 implied correlation increased more 12-month implied correlation Exhibit 25: Eurozone banks implied volatility spiked more EURO STOXX month implied volatility EURO STOXX 50 S&P 500 (RHS) Average single stock Index Eurozone banks Normalised skew = (ATM 25 delta put ATM 25 delta call)/ ATM 50 delta call. 8 Implied correlation can be calculated as I W. 高盛全球经济 商品和策略研究 11

12 信息披露附录 申明 我们,Christian Mueller-Glissmann, CFA 彼得 欧品海默 Sharon Bell, CFA Gerald Moser Anders Nielsen Matthieu Walterspiler, 在此申明, 本报告所表述的所有观点准确反映了我们对上述公司或其证券的个人看法 此外, 我们的薪金的任何部分不曾与, 不与, 也将不会与本报告中的具体推荐意见或观点直接或间接相关 高盛信息披露 评级分布 / 投资银行关系 高盛投资研究部的全球研究覆盖范围 评级分布 投资银行关系 买入 持有 卖出 买入 持有 卖出 全球 30% 55% 15% 47% 42% 34% 截至 2012 年 01 月 16 日, 高盛全球投资研究部对 3,593 种股票评定了投资评级 高盛给予股票在各种地区投资名单中的买入和卖出评级 ; 未给予这些评级的股票被视为中性评级, 根据纳斯达克 / 纽约证券交易所的披露要求, 这些评级分别对应买入, 持有及卖出 详情见以下 公司评级, 研究行业及评级和相关定义 部分 美国法定披露 任何本报告中研究企业所需的特定公司法定披露见上文 : 包括即将进行交易的承销商或副承销商,1% 或其他股权, 特定服务的补偿, 客户关系种类, 之前担任承销商或副承销商的公开发行, 担任董事, 担任股票做市及 / 或专家的角色 高盛通常担任本报告中涉及的固定收益证券的做市商, 并常作为这些证券的交易对手 以下为额外要求的披露 : 股权及重大利益冲突 : 高盛的政策为禁止其分析师 分析师属下专业人员及其家庭成员持有分析师负责研究的任何公司的证券 分析师薪酬 : 分析师薪酬部分取决于高盛的盈利, 其中包括投资银行的收入 分析师担任高级职员或董事 : 高盛的政策为禁止其分析师 分析师属下人员及其家庭成员担任分析师负责研究的任何公司的高级职员 董事 顾问委员会成员或雇员 非美国分析师 : 非美国分析师可能与高盛无关联, 因此可以不受纳斯达克 2711 条 / 纽约证券交易所 472 条对于与所研究公司的交流 公开露面及持有交易证券的限制 美国以外司法管辖区规定的额外披露 以下为除了根据美国法律法规规定作出的上述信息披露之外其他司法管辖区法律所要求的披露 澳大利亚 : Goldman Sachs Australia Pty Ltd 及其相关机构不是澳大利亚经授权的存款机构 (1959 年 银行法 所定义 ), 因此不在澳大利亚境内提供银行服务, 也不经营银行业务 本研究报告或本报告的其他形式内容只可分发予根据澳大利亚公司法定义的 " 批发客户 ", 在事先获得高盛许可的情况下可以有例外 巴西 : 与 CVM Instruction 483 相关的信息披露请参阅 根据 CVM Instruction 483 第 16 条, 在适用的情况下, 对本研究报告内容负主要责任的巴西注册分析师为本报告开头部分标明的第一作者, 除非报告末另有说明 加拿大 : 如果本报告与加拿大股票发行人有关, 高盛集团已批准本报告, 并同意承担有关责任 分析师可进行实地考察, 但不可收受公司对此等差旅支付的任何款项或偿付 香港 : 可从高盛 ( 亚洲 ) 有限责任公司获取有关本报告中所研究公司的证券的额外资料 印度 : 有关本研究报告中的研究对象或所提及的公司的进一步信息可能来自高盛 ( 印度 ) 证券私人有限公司 日本 : 见下文 韩国 : 可从高盛 ( 亚洲 ) 有限责任公司首尔分公司获取有关本报告所研究公司的证券的额外资料 俄罗斯 : 在俄罗斯联邦分发的研究报告并非俄罗斯法律所定义的广告, 而是不以产品推广为主要目的的信息和分析, 也不属于俄罗斯法律所界定的评估行为 新加坡 : 可从高盛 ( 新加坡 ) 私人公司 ( 公司编号 : W) 获取有关本报告中所研究公司的证券的额外资料 台湾 : 本信息仅供参考, 未经允许不得翻印 投资者应当谨慎考虑他们自身的投资风险, 投资结果由投资者自行负责 英国 : 在英国根据金融服务局的定义可被分类为私人客户的人士参阅本报告的同时应当参阅高盛以往对本报告研究企业的研究报告, 并应当参考高盛国际已经发给这些客户的风险警告资料 该风险警告资料复本, 以及本报告中采用部分金融辞汇的解释可向高盛国际索取 欧盟 : 与欧盟指令 2003/126/EC 第四章 (1)(d) 和第六章 (2) 有关的披露信息可参见 其中列明了欧洲在管理投资研究方面利益冲突的政策 日本 : 高盛证券株式会社是依据 金融工具与交易法 在关东财务局注册 ( 注册号 :No. 69) 的金融工具交易商, 同时也是日本证券业协会和日本金融期货业协会的成员 股票买卖需要缴纳与客户事先约定的佣金及消费税 关于日本证券交易所 日本证券交易商协会或日本证券金融公司所要求的适用的信息披露, 请参见与公司有关的法定披露部分 公司评级 研究行业及评级和相关定义 买入 中性 卖出 : 分析师建议将评为买入或卖出的股票纳入地区投资名单 一只股票在投资名单中评为买入或卖出由其相对于所属研究行业的潜在回报决定 任何未获得买入或卖出评级的股票均被视为中性评级 每个地区投资评估委员会根据 25-35% 的股票评级为买入 10-15% 的股票评级为卖出的全球指导原则来管理该地区的投资名单 ; 但是, 在某一特定行业买入和卖出评级的分布可能根据地区投资评估委员会的决定而有所不同 地区强力买入或卖出名单是以潜在回报规模或实现回报的可能性为主要依据的投资建议 潜在回报 : 代表当前股价与一定时间范围内预测目标价格之差 分析师被要求对研究范围内的所有股票给出目标价格 潜在回报 目标价格及相关时间范围在每份加入投资名单或重申维持在投资名单的研究报告中都有注明 高盛全球经济 商品和策略研究 12

13 研究行业及评级 : 每个行业研究的所有股票名单可登陆 通过主要分析师 股票和行业进行查询 分析师给出下列评级中的其中一项代表其根据行业历史基本面及 / 或估值对研究对象的投资前景的看法 具吸引力 (A): 未来 12 个月内投资前景优于研究范围的历史基本面及 / 或估值 中性 (N): 未来 12 个月内投资前景相对研究范围的历史基本面及 / 或估值持平 谨慎 (C): 未来 12 个月内投资前景劣于研究范围的历史基本面及 / 或估值 暂无评级 (NR): 在高盛于涉及该公司的一项合并交易或战略性交易中担任咨询顾问时并在某些其他情况下, 投资评级和目标价格已经根据高盛的政策予以除去 暂停评级 (RS): 由于缺乏足够的基础去确定投资评级或价格目标, 或在发表报告方面存在法律 监管或政策的限制, 我们已经暂停对这种股票给予投资评级和价格目标 此前对这种股票作出的投资评级和价格目标 ( 如有的话 ) 将不再有效, 因此投资者不应依赖该等资料 暂停研究 (CS): 我们已经暂停对该公司的研究 没有研究 (NC): 我们没有对该公司进行研究 不存在或不适用 (NA): 此资料不存在或不适用 无意义 (NM): 此资料无意义, 因此不包括在报告内 全球产品 ; 分发机构 高盛全球投资研究部在全球范围内为高盛的客户制作并分发研究产品 高盛分布在其全球各办事处的分析师提供行业和公司的股票研究, 以及宏观经济 货币 商品及投资组合策略的研究 本研究报告在澳大利亚由 Goldman Sachs Australia Pty Ltd(ABN ) 分发 ; 在巴西由 Goldman Sachs do Brasil Banco Múltiplo S.A. 分发 ; 股票及其他研究在加拿大由高盛集团分发 ; 在香港由高盛 ( 亚洲 ) 有限责任公司分发 ; 在印度由高盛 ( 印度 ) 证券私人有限公司分发 ; 在日本由高盛证券株式会社分发 ; 在韩国由高盛 ( 亚洲 ) 有限责任公司首尔分公司分发 ; 在新西兰由 Goldman Sachs New Zealand Limited 分发 ; 在俄罗斯由高盛 OOO 分发 ; 在新加坡由高盛 ( 新加坡 ) 私人公司 ( 公司号 : W) 分发 ; 在美国由高盛集团分发 高盛国际已批准本研究报告在英国和欧盟分发 欧盟 : 高盛国际 ( 由英国金融服务局监管 ) 已批准本研究报告在英国和欧盟分发 ;Goldman Sachs AG( 由联邦金融监管局监管 ) 可能也会在德国分发 一般性披露 本研究报告仅供我们的客户使用 除了与高盛相关的披露, 本研究报告是基于我们认为可靠的目前已公开的信息, 但我们不保证该信息的准确性和完整性, 客户也不应该依赖该信息是准确和完整的 我们会适时地更新我们的研究, 但各种规定可能会阻止我们这样做 除了一些定期出版的行业报告之外, 绝大多数报告是在分析师认为适当的时候不定期地出版 高盛是一家集投资银行 投资管理和证券经纪业务于一身的全球性综合服务公司 高盛全球投资研究部所研究的大部分公司与我们保持着投资银行业务和其它业务关系 美国证券经纪交易商高盛是 SIPC 的成员 ( 我们的销售人员 交易员和其它专业人员可能会向我们的客户及我们的自营交易部提供与本研究报告中的观点截然相反的口头或书面市场评论或交易策略 我们的资产管理部门 自营交易部和投资业务部可能会做出与本报告的建议或表达的意见不一致的投资决策 本报告中署名的分析师可能已经与包括高盛销售人员和交易员在内的我们的客户讨论, 或在本报告中讨论交易策略, 其中提及可能会对本报告讨论的证券市场价格产生短期影响的推动因素或事件, 该影响在方向上可能与分析师发布的股票目标价格相反 任何此类交易策略都区别于且不影响分析师对于该股的基本评级, 此类评级反映了某只股票相对于报告中描述的研究范围内股票的回报潜力 我们以及我们的关联机构 高级职员 董事和雇员, 不包括股票分析师和信贷分析师, 将不时地对本研究报告所涉及的证券或衍生工具持有多头或空头头寸, 担任上述证券或衍生工具的交易对手, 或买卖上述证券或衍生工具 在任何要约出售股票或征求购买股票要约的行为为非法的司法管辖区内, 本报告不构成该等出售要约或征求购买要约 本报告不构成个人投资建议, 也没有考虑到个别客户特殊的投资目标 财务状况或需求 客户应考虑本报告中的任何意见或建议是否符合其特定状况, 以及 ( 若有必要 ) 寻求专家的意见, 包括税务意见 本报告中提及的投资价格和价值以及这些投资带来的收入可能会波动 过去的表现并不代表未来的表现, 未来的回报也无法保证, 投资者可能会损失本金 外汇汇率波动有可能对某些投资的价值或价格或来自这一投资的收入产生不良影响 某些交易, 包括牵涉期货 期权和其它衍生工具的交易, 有很大的风险, 因此并不适合所有投资者 投资者可以向高盛销售代表取得或通过 取得当前的期权披露文件 对于包含多重期权买卖的期权策略结构产品, 例如, 期权差价结构产品, 其交易成本可能较高 与交易相关的文件将根据要求提供 在撰写研究报告期间,Goldman Sachs Australia 全球投资研究部的职员可能参与本研究报告中所讨论证券的发行人组织的现场调研或会议 在某些情况下, 如果视具体情形 Goldman Sachs Australia 认为恰当或合理, 此类调研或会议的成本可能部分或全部由该证券发行人承担 所有研究报告均以电子出版物的形式刊登在我们的内部客户网上并向所有客户同步提供 并非所有研究内容都转发给我们的客户或者向第三方整合者提供, 高盛也并不对由第三方整合者转发的我们研究报告承担任何责任 有关个股的所有研究报告, 请联络您的销售代表或登陆 披露信息可以查阅 或向研究合规部索取, 地址是 200 West Street,New York,NY 高盛版权所有 2012 年未经高盛集团公司事先书面同意, 本材料的任何部分均不得 (i) 以任何方式制作任何形式的拷贝 复印件或复制品, 或 (ii) 再次分发 高华证券信息披露 一般披露 本报告在中国由高华证券分发 高华证券具备证券投资咨询业务资格 本研究报告仅供我们的客户使用 本研究报告是基于我们认为可靠的目前已公开的信息, 但我们不保证该信息的准确性和完整性, 客户也不应该依赖该信息是准确和完整的 我们会适时地更新我们的研究, 但各种规定可能会阻止我们这样做 除了一些定期出版的行业报告之外, 绝大多数报告是在分析师认为适当的时候不定期地出版 高盛高华为高华证券的关联机构, 从事投资银行业务 高华证券 高盛高华及它们的关联机构与本报告中涉及的大部分公司保持着投资银行业务和其它业务关系 我们的销售人员 交易员和其它专业人员可能会向我们的客户及我们的自营交易部提供与本研究报告中的观点截然相反的口头或书面市场评论或交易策略 我们的自营交易部和投资业务部可能会做出与本报告的建议或表达的意见不一致的投资决策 高盛全球经济 商品和策略研究 13

14 本报告中署名的分析师可能已经与包括高盛销售人员和交易员在内的我们的客户讨论, 或在本报告中讨论交易策略, 其中提及可能会对本报告讨论的证券市场价格产生短期影响的推动因素或事件, 该影响在方向上可能与分析师发布的股票目标价格相反 任何此类交易策略都区别于且不影响分析师对于该股的基本评级, 此类评级反映了某只股票相对于报告中描述的研究范围内股票的回报潜力 高华证券及其关联机构 高级职员 董事和雇员, 不包括股票分析师和信贷分析师, 将不时地对本研究报告所涉及的证券或衍生工具持有多头或空头头寸, 担任上述证券或衍生工具的交易对手, 或买卖上述证券或衍生工具 在任何要约出售股票或征求购买股票要约的行为为非法的地区, 本报告不构成该等出售要约或征求购买要约 本报告不构成个人投资建议, 也没有考虑到个别客户特殊的投资目标 财务状况或需求 客户应考虑本报告中的任何意见或建议是否符合其特定状况, 以及 ( 若有必要 ) 寻求专家的意见, 包括税务意见 本报告中提及的投资价格和价值以及这些投资带来的收入可能会波动 过去的表现并不代表未来的表现, 未来的回报也无法保证, 投资者可能会损失本金 外汇汇率波动有可能对某些投资的价值或价格或来自这一投资的收入产生不良影响 某些交易, 包括牵涉期货 期权和其它衍生工具的交易, 有很大的风险, 因此并不适合所有投资者 投资者可以向高华销售代表取得或通过 取得当前的期权披露文件 对于包含多重期权买卖的期权策略结构产品, 例如, 期权差价结构产品, 其交易成本可能较高 与交易相关的文件将根据要求提供 在撰写研究报告期间,Goldman Sachs Australia 全球投资研究部的职员可能参与本研究报告中所讨论证券的发行人组织的现场调研或会议 在某些情况下, 如果视具体情形 Goldman Sachs Australia 认为恰当或合理, 此类调研或会议的成本可能部分或全部由该证券发行人承担 北京高华证券有限责任公司版权所有 2012 年未经北京高华证券有限责任公司事先书面同意, 本材料的任何部分均不得 (i) 以任何方式制作任何形式的拷贝 复印件或复制品, 或 (ii) 再次分发 高盛全球经济 商品和策略研究 14