全球经济周评：从更广视角来审视新兴市场增长放缓1102.ppt

,高盛国际,2012 年 11 月 1 日Issue No:12/36全球经济周评研究报告从更广视角来审视新兴市场增长放缓,新兴市场增长缓慢引发结构性放缓担忧新兴市场近期的增长放缓引发了市场对于本轮放缓更多是周期性（即放缓势头将在未来几个月逆转）还是结构性（即部分国家将进入增速更低的可持续增长轨道）的担忧。中国是这些担忧的焦点所在，但韩国、巴西和印度等其他主要新兴市场也引发了类似担忧。高速增长经济体的结构性放缓较为普遍从更广阔的全球视角来看，高速增长经济体的结构性放缓相对普遍，而且通常在人均 GDP 处于 1.5 万-2.5 万美元间的经济体出现。类似的结构性放缓也出现在我们的长期增长预测中，这表明增长的趋同性可能自然而然地引发了此类放缓。中国（和新兴市场）所处周期阶段仍是新兴市场资产的关键所在中国等许多大型新兴经济体的收入水平仍未达到普遍存在增速放缓的收入下限。因此对高速增长已至尽头的担心可能为时过早，尽管中国潜在增速的部分放缓已经在我们的长期预测中出现。在研究中国和新兴市场增长（以及新兴市场资产）轨迹时，关注所处周期阶段仍十分重要。宏观环比数据几个月以来首次出现耐人寻味的改善迹象，这表明中国周期（以及中国相关资产）可能企稳。,Kamakshya Trivedi+44(20)7051-4005 Jose Ursua(212)357-2234 高盛集团George Cole+44(20)7552-3779 高盛国际Julian Richers(212)855-0684 高盛集团多米尼克威尔逊(212)902-5924,高盛集团结构性放缓通常出现在人均 GDP 2 万美元左右的经济体结构性放缓启动的人均 GDP 水平；非石油出口国(1950-2010),4540,GDP pc(000)at whichstructural slowdowns,Mean,Median,+/-1SD,35302520151050,begin,5,000,10,000,15,000,Threshold of GDP pc资料来源：WDI、PWT、高盛全球经济、商品和策略研究投资者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,2,2012 年 11 月 1 日,Global Economics Weekly,The EM growth slowdown in a broader perspectiveThe recent slowdown in growth in emerging markets(EM)especially in some of the largereconomies such as China,Brazil and India has become an increasingly important topic ofdiscussion within the global macro investment community.A key question here is the degree towhich this slowdown is largely cyclical(i.e.,it will reverse in the coming months as policyresponds)or more structural(i.e.,some of these countries are moving to a lower sustainablegrowth path than witnessed in recent years).China is perhaps the poster-child of these worries,but similar concerns have been raised about other major EMs,such as Korea,Brazil and India.We discussed some of the cyclical explanations behind the poor recent growth performance insome of the larger EMs in Global Economics Weekly 12/29,Deconstructing our global forecastsand the weakness in large EMs.We focused in particular on the impact of the Great Easing thatfollowed the Great Recession:in some cases,the ability of policy makers to ease further hasbeen constrained,and in others the effectiveness of easing has been limited.Despite substantial heterogeneity in the experience across countries,our overall forecasts for EMgrowth pencil in a reacceleration in 2013 from the lows of 2012(6.3%yoy in 2013 vs 5.6%yoy in2012).This suggests that,at least in aggregate,a large part of the slowdown in EM growth mayprove to be cyclical.Indeed,there are signs that a turn in the EM cycle may be closer at hand,asdiscussed in the inaugural edition of the Emerging Markets Macro Daily:Questions for the EMoutlook in our inaugural Daily,October 29,2012.These include better than expected trade,IPand profits data in China,and an improvement in the flash PMI,as well as better sequential newsfrom Korea and Taiwan.What of the structural considerations?In practice,structural and cyclical explanations arenotoriously difficult to separate definitively,and the answer is almost certainly different for eachcountry.The first step,however,is to understand the nature of such medium-term or structuralslowdowns in previously fast-growing economies.Here we examine the incidence of growthslowdowns in history and in the long-term forward projections that we have constructed andpublished(see Global Economics Paper 208,The BRICs 10 Years On:Halfway Through theGreat Transformation).The main messages that emerge from these exercises are as follows:Structural growth slowdowns in previously rapidly growing economies are relativelycommon,especially at low-and middle-income levels(roughly between per capitaincomes of$15,000 and$25,000).But structural growth slowdowns may,to a large extent,be a function of growth andincome convergence dynamics.Our projections for global incomes,which illustrate thetransformative role of the BRICs and N-11 in the global economy,envision structuralslowdowns at similar levels of income as observed in history.Many of the larger EMs such as China still lie below the income thresholds wherestructural growth slowdowns are common,suggesting that concerns about theexhaustion of catch-up growth are premature.Nevertheless,our long-term projectionsdo envisage some slowing in Chinas high rates of growth as the rate of capitaldeepening and labour force accumulation slow in future years.Structural slowdowns are common,especially at low-and middle-income levelsOne of the striking features of the recovery from the global financial crisis(GFC)and associatedglobal recession has been the swift return to strong growth across the EM world.With developedmarkets growing at only modest rates,typical of post-housing bust economies,this strong growthin EMs has been a key pillar supporting growth in the global economy.But growth has slowed inthe EM world since 2011,partly as some of the unprecedented fiscal and monetary stimulus afterthe GFC was withdrawn.As that slowdown has persisted,and deepened in some cases through高盛全球经济、商品和策略研究,0,3,2012 年 11 月 1 日,Global Economics Weekly2012,there are more questions about whether the slowdowns represent something morefundamental or structural,and therefore are longer-lasting than in the typical economic cycle.Given the multi-year nature of any structural slowdown in economic growth,it may be useful toplace the present economic slowdown within historical cross-country perspective.Therefore,wetake a step back to examine the incidence of economic slowdowns over the past 50 years.Economic slowdowns are relative phenomena:literally,a comparison of growth rates acrossperiods divided at a certain point.A systematic approach to identifying and understanding suchslowdowns requires drawing lines over fixed thresholds,which are reasonable and intuitive butcannot avoid some degree of arbitrariness.Our baseline definition follows that of a paper byEichengreen,Park and Shin(When Fast Growing Economies Slow Down:International Evidenceand Implications,Asian Development Bank,June 2011),which defines slowdowns as episodessatisfying three conditions:Growth in the seven years prior to the structural slowdown should be at least 3.5%.A structural slowdown requires that the average growth during the subsequent sevenyears be at least 2ppt lower.Slowdowns are restricted to countries with GDP per capita of at least$10,000(inconstant 2005 international Dollars).In simple terms,we are interested in capturing instances of slowdowns when growth deceleratesfrom a rapid rate,by a meaningful amount,and over a sizeable length of time.The incomethreshold ensures that only countries that are on the path to development are included(so as torule out very low-income countries with a high degree of output volatility).Given that many of theEMs we care about today have per capita incomes below$10,000,we consider lower(and higher)thresholds of GDP per capita as well.Our historical sample comes from the latest version of the Penn World Tables,covering almost190 countries from 1950 through 2010.We focus primarily on non-oil exporting countries,occasionally referring to a narrow sample where,in addition to the historical data,we haveprojections through to 2050(based on Global Economics Paper 208).,Exhibit 1:Number of structural slowdowns declines athigher income levelsIdentified episodes by threshold of GDP pc and sample,Exhibit 2:Examples of structural slowdown episodesidentified by our algorithmCases according to GDP pc at time of slowdown(2005 Intl.$),140,Number,All countries,$5,000Belize,1990-95,$10,000Argentina 1997-98,$15,000Bahamas 1982-85,12010080604020,Non-oil exporters(broad sample)Non-oil exporters(narrow sample),BrazilBulgariaDom.Rep.SlovakiaGuatemalaLebanonMalaysiaPolandTonga,1973-801987-891998-9919981977-781995-971993-981977-791982-88,AustriaCyprusGabonHungaryIranIsraelMexicoSpainSuriname,19611989-9319941977-801971-771971-761979-8119661977-81,BarbadosBelgiumDenmarkHong KongIcelandJapanPortugalSingaporeSeychelles,1969-731974-751969-701990-9419821989-921990-921980-841998,5,000,10,000,15,000,Uruguay,1995-98,Taiwan,1992-93,USA,1968-69,Threshold of GDP pc,Source:WDI;PWT;Goldman Sachs Global ECS Research.高盛全球经济、商品和策略研究,Source:WDI;PWT;Goldman Sachs Global ECS Research.,4,2012 年 11 月 1 日Exhibit 3:Structural slowdowns tend to occur at GDP pclevels of around$20,000GDP pc of slowdown starts;non-oil exporters(1950-2010),Global Economics WeeklyExhibit 4:Growth shortfall during structural slowdownsaverages around 3pptAvg.growth before and during slowdowns;non-oil exporters,454035,GDP pc(000)at whichstructuralslowdownsbegin,Mean,Median,+/-1SD,6.05.0,%,Previous,During Slowdown,30252015105,4.03.02.01.0,3ppt,0,5,000,10,000,15,000,0.0,5,000,10,000,15,000,Threshold of GDP pcSource:WDI;PWT;Goldman Sachs Global ECS Research.,Threshold of GDP pcSource:WDI;PWT;Goldman Sachs Global ECS Research.,After running the algorithm as described above,the first observation is that structuralslowdowns are quite frequent;put differently,it is not that common for countries to sustain highrates of growth for prolonged periods.Exhibit 1 shows the number of slowdown episodes in ourthree basic samples and by the threshold of GDP per capita after which slowdowns areaccounted for.We found 70 episodes in our$10,000-onward baseline case,but this number risesto almost 120 when the threshold is lowered to$5,000.These include many episodes that areoften thought to embody the notion of EM economies slowing down after a period of rapid growth,such as Brazil(1973-80),Argentina(1997-98),Poland(1977-79),and Mexico(1979-81).Exhibit 2provides more examples of slowdown episodes.A second observation is that,while the average GDP per capita at which structural slowdownstend to occur depends,obviously,on the threshold one chooses,it hovers around$20,000(in2005 international PPP Dollars).For our baseline$10,000 threshold,the mean and median are at$22,513 and$19,693,respectively(Exhibit 3).For lower thresholds,say$5,000,the mean andmedian fall to$16,435 and$12,878.In contrast,for the higher thresholds,the average incomelevel at which slowdowns start can be higher than$26,000.The third observation relates to the average pre-slowdown and during-slowdown growth ratescaptured by our algorithm.Once more,for the$10,000 threshold baseline,the average pre-slowdown growth rate is 5.3%(mean)and 4.8%(median),which means that growth is indeedhigh before slowdowns begin.During slowdowns,growth hovers around 2.6%(mean)and 2.3%(median).Therefore,on average,structural slowdowns imply a difference with respect to previousgrowth rates of almost 3ppt.Unlike the number or starting income level of slowdowns,theirseverity in terms of the difference between pre-and during-slowdown growth rates does notdepend much on the threshold one chooses to account for these phenomena(Exhibit 4).Structural slowdowns tend to occur between$15k and$25k GDP pcTo gain a firmer sense of whether todays large EM economies are at risk of a structural growthslowdown,as we have defined it,we would like to know if there is any tendency for suchslowdowns to bunch around particular income levels.In Exhibit 5 we depict the relativefrequencies of slowdowns for the benchmark case where the sample is restricted to slowdownsabove$10,000.This does suggest that slowdowns are centred around the$20,000 mark,andthat the danger zone for countries falling into structural growth slowdowns lies between$15,000and$25,000 GDP per capita levels.高盛全球经济、商品和策略研究,10-12.5,12.5-15,15-17.5,17.5-20,20-22.5,22.5-25,25-27.5,27.5-30,30-32.5,32.5-35,35-37.5,37.5-40,40-42.5,42.5-45,45-47.5,47.5-50,7.5-10,5-7.5,10-12.5,12.5-15,15-17.5,17.5-20,20-22.5,22.5-25,25-27.5,27.5-30,30-32.5,32.5-35,35-37.5,37.5-40,40-42.5,42.5-45,45-47.5,47.5-50,5,2012 年 11 月 1 日,Global Economics WeeklyThis conclusion is pretty robust to altering the criteria for defining structural slowdowns in variousways.Varying income thresholds so that we include slowdowns at lower income levels starting at$5,000 increases by construction the sample size and the frequency of occurrence at low incomelevels.But the tendency to cluster at GDP per capita between$15,000 and$25,000 is visiblehere too(Exhibit 6).Other aspects of the criteria can also be varied.In Exhibit 7,we leave theincome threshold fixed at our benchmark level of$10,000,but alter the required size of pre-slowdown growth(which we bump up to 5%from 3.5%),and the length of the pre-slowdowngrowth period(which we raise to 10 years instead of 7 years).In both cases,slowdowns are stillcentred around the$20,000 GDP per capita level and the deceleration in growth still amounts toapproximately 3ppt.So,our results on growth slowdowns are robust to considering slowdownswhere the pre-slowdown growth was stronger,or if the pre-slowdown period of strong growthlasted longer.Our long-run projections envision similar structural slowdownsThe structural slowdowns that we have identified in the historical data may be triggered by shocksor crises that dislocate capital or labour markets,or downward shifts in productivity growth thatcountries are unable to reverse either because of policy failures or simply because as countriesconverge towards the frontier they have limited scope for catch-up growth.One way of examiningthis question is to run our slowdown algorithm over our long-run income projections through to2050 to see whether a similar pattern of slowdowns is envisaged.The projections are based onestimates of total factor productivity growth,capital accumulation and labour force dynamics,soshould capture slowdowns that are based solely on their equilibrium growth and convergencedynamics.Combining our baseline sample of non-oil exporters with our growth projections through 2050reduces the sample to about 70 countries,but it allows us to cover 100 years of data,half ofwhich are observed and half of which are projected.The most notable result is that the mean and median threshold levels for projected figures aresimilar to those from past observations.For the$10,000 threshold,the mean is slightly lower($19,047 vs.$20,901)but the median is slightly higher($18,721 vs.$17,729),Exhibit 8.Therefore,in both cases,the average GDP per capita at structural slowdown starts is still close to$20,000.Because our projections abstract from cyclical fluctuations,the standard deviation bandsare narrower in this case($7,535 vs.$12,366).Moreover,the average decline in growth is closerto the prescribed condition of 2ppt,going on average from about 5.0%before the slowdown to3.0%.,Exhibit 5:Starts of slowdowns bunch at GDP pc between$15k and$20k.Relative frequencies of slowdowns at GDP pc starting at$10k,Exhibit 6:.even if lower levels of GDP pc are included inthe sampleRelative frequencies of slowdowns at GDP pc starting at$5k,0.160.14,Rel.Freq.,0.25,Rel.Freq.,Slowdowns starting at 5kSlowdowns starting at 10k,0.12,0.100.080.060.040.020.00GDP pc(000)at which structural slowdowns beginSource:WDI;PWT;Goldman Sachs Global ECS Research.高盛全球经济、商品和策略研究,0.200.150.100.050.00GDP pc(000)at which structural slowdowns beginSource:WDI;PWT;Goldman Sachs Global ECS Research.,6,2012