1、1 19 March 2019|DEEP DIVEEM Strategy:Inferring credit cycles from EM asset prices Proposed approach Economic cycles,also known as business or trade cycles,refer to the economy-wide fluctuations in production,trade and general economic activity.From a conceptual point of view,a cycle is often defined
2、 by the upward or downward long-term trend of real GDP and other macro indicators such as industrial production,real-income,employment,sales,etc.Bottom out cycles,for example,are usually characterized by periods of significant decline in economic activity across the whole economy,or by two consecuti
3、ve quarters of decline in real GDP.However,we believe that market agents can sometimes react faster than real GDP when assessing the impact of lower economic activity,thereby directly impacting asset prices.Unlike the standard approach to economic cycles,which uses macroeconomic variables,we used on
4、ly emerging markets asset prices to measure(i)the stage of the economic cycle in each country,(ii)the duration of the current cycle,(iii)the pace of acceleration/deceleration,and(iv)how long the current cycle is likely to last.Since our goal was to extract market perception on where selected emergin
5、g markets are in terms of economic cycles,we only used mathematical methods to define the current cycle instead of using regressing variables or making historical assumptions.LATIN AMERICA STRATEGYPlease refer to important information at the end of the report DEEP DIVE|EM STRATEGY 19 March 2019 Gabr
6、iel Gersztein Global Head of Emerging Markets Strategy+55 11 3841 3421 Luca Maia FX&IR Latin America Strategy+55 11 3841 3447 Banco BNP Paribas Brasil S.A.KEY MESSAGES When it comes to assessing economic cycles,the conventional approach is to gather and sort out economic variables to determine the p
7、hase of the credit cycle.What if we could model the data in such a way that we are able to infer the phase of the cycles from market prices?With that objective in mind,we have designed a quantitative tool that uses the performance of selected financial assets to extract the implied stage of the econ
8、omic cycle.Results suggest that current market prices imply that Brazil is in the early stage of the credit cycle.South Africa,Indonesia,USA,and Turkey have already entered into the late phase.Indonesia is still in the mid of the cycle whereas South Korea,Poland and China are bottoming out/slowing d
9、own.We plan to update the model on a monthly basis and use the results to further complement our recommendations.The results are in line with the conclusions of the BNPP Global Outlook;link:Global Outlook:Q2 2019:Beige-ilocks Weak economy,strong carry 2 19 March 2019|DEEP DIVEModelling economic cycl
10、es We started with three countries:Brazil,Mexico and South Africa and have now expanded our framework to include the US,China,Indonesia,South Korea,Turkey and Poland.In order to determine the part of the economic cycle each country is in,we developed a unique framework,which uses only market variabl
11、es related to interest rates,equities,bonds,break-evens and FX performance(Table 1).We already use most of the selected variables in the models we have in place such as BEER,interest rates factor model and debt dynamics for CDS.Table 1:Variables used to extract the implied cycle Source:BNP Paribas W
12、e applied a normalization process to make all our series comparable,varying inside the interval-1,+1.All sub-indices were individually adjusted in a way that positive numbers indicate a higher perceived economic growth and negative values a failing perceived economic activity(see Technical Appendix
13、for details)Our index took the average of all normalized sub-indices using the same positive and negative numbers criteria on a monthly frequency.The first column of the figures below shows our proposed index.Considering that GDP data releases usually lag by a quarter,our indicator appeared to work
14、as a leading indicator for economic activity in most cases.China was the only exception,as we used the first monetary aggregate change YoY with our index instead of GDP(Fig.9)although the calculations were the same.We applied a Hodrick-Prescott filter(HP filter),to isolate the trend component from n
15、oise and cyclical factors(second column of the figures).Fig.1-6:Brazil,Mexico and South Africa economic cycle index before and after HP filter Sources:Bloomberg LLP,BNP Paribas Local MSCI financialsLocal currency-EM currency returnLocal MSCI(growth-value)MSCI local-MSCI global1y ex-ante real rate2y
16、ex-ante real rateIR Slope 5x11y breakeven inflationEquity-Fixed income returnsLocal MSCI(Small Caps-Large Caps)-6%-2%2%6%-0.70-0.40-0.100.20Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Cycle IndexSouth Africal GDP YoY(rhs)-6%-2%2%6%-0.50-0.250.000.25Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Cycle IndexIndex with H
17、P filterSouth Africa GDP YoY(rhs)-9%-4%2%7%-0.60-0.300.000.300.60Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Cycle IndexMexico GDP YoY(rhs)-9%-4%2%7%-0.60-0.300.000.300.60Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Cycle IndexIndex with HP filterMexico GDP YoY(rhs)-8%-4%0%4%8%-0.75-0.45-0.150.150.45Jan-03Jan-06Jan-
18、09Jan-12Jan-15Jan-18Cyclel IndexBrazil GDP YoY(rhs)-6%-2%2%6%10%-0.60-0.300.000.300.60Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Cycle IndexIndex with HP filterBrazil GDP YoY(rhs)3 19 March 2019|DEEP DIVEFig.7-18:US,China,Indonesia,South Korea,Turkey,Poland economic cycle index before and after HP filter S
19、ources:Bloomberg LLP,BNP Paribas-8%-4%0%4%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexUS GDP YoY(rhs)-8%-4%0%4%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndex with HP filterUS GDP YoY(rhs)-4%6%16%26%36%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Ja
20、n-15Jan-18Original IndexChina M1 YoY(rhs)-4%6%16%26%36%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndex with HP filter2%4%6%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndonesia GDP YoY(rhs)2%4%6%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan
21、-18Original IndexIndex with HP filterIndonesia GDP YoY(rhs)-4%0%4%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexKorea GDP YoY(rhs)-4%0%4%8%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndex with HP filterKorea GDP YoY(rhs)-15%-11%-7%-3%1%5%9%13%-0.75-0.
22、350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexTurkey GDP YoY(rhs)-15%-11%-7%-3%1%5%9%13%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndex with HP filterTurkey GDP YoY(rhs)-1%1%3%5%7%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexPoland GDP Y
23、oY(rhs)-1%1%3%5%7%-0.75-0.350.050.45Jan-03Jan-06Jan-09Jan-12Jan-15Jan-18Original IndexIndex with HP filterPoland GDP YoY(rhs)4 19 March 2019|DEEP DIVEThe next step was to divide an economic cycle into four stages:early cycle,mid cycle,late cycle and bottom out cycle.For each cycle,we identified how
24、the main financial indicators react during each phase.Our conclusions are shown in Fig.19.Table 2 and Fig.19-22:Breakdown of an economic cycle and examples of variables for Brazil Source:BNP Paribas Using the cycle index as the base,our next step was to identify the part of the cycle each emerging m
25、arket is in.We concluded that the best way to do this was to fit the cycle index using a sinusoid function and calculate the value of the angle in each point(month)of the series.We divided the sinusoid function into four regions,each with a 90 degrees interval,as shown in Fig.23 below.Fig.23:Dividin
26、g the sinusoid function into four regions Sources:BNP Paribas To ensure that the fit does not overshoot,we included a constraint for each new observation the modelled angle will not move more than 10%of a full cycle of 360.We also estimated the initial frequency of the economic cycle by extracting t
27、he first harmonic using a Fourier transformation(see Technical Appendix for a detailed explanation).Activity Rebounds(GDP,IP)Growth peaking Growth moderatingFalling Activity Inflation Credit begins to grow Credit growth gains pace Credit tightensCredit dries upProfits grow rapidly Profit growth peak
28、s Earnings under pressure Profits declinePolicy stimulative Policy neutral Policy contractionaryPolicy easing nearingInventories low Inventories,sales grow Inventories grow;Inventories fall;Sales improve Inflation Sales growth fallsSales fallVery high growthPos GrowthNeg Growth orVery low growthYiel
29、d Bull Steepening Bear Flattening Bear Steepening Bear FlatteningCurve V-Shaped sharp recovery fromTypically the longest phase of Monetary policy starts to beContraction in recession or near-recessionthe economic cycle restrictive.Inflation picks upeconomic activity Small Caps;Value x GrowthActivity
30、 gathers momentum Inventories tend to build as salesCredit scarce and Equities over fixed incomeStrong credit growth,healthy growth declines.profits profitability against an Fixed Income over EquitiesMonetary policyaccommodative(through Large Caps;Value x Growthbecomes moreincreasingly neutral)monet
31、ary backdropaccommodativeEquities over fixed incomeEARLYMIDLATEBOTTOM OUTRECOVERYEXPANSIONBOTTOM OUTGROWTH BELOW POTENTIAL-300-1500150300-40%-13%15%43%Brazil Slope 5x1-LT averagePerformance Equity-Fixed IncomeIbovespa outperformingFixed Income outperforming-300-1500150300450-150-3090210330 YoYBrazil
32、 Slope 5x1 since 2002Bull SteepeningBear Flattening-200-50100250-30%10%50%90%130%Brazil Slope 5x1MSCI Brazil (Small-Large)YoYFlatteningSteepening B-EARLYC-MIDD-LATEA-BOTTOM OUT/GROWTH BELOW 45135225315315Step 1:Construction of the cycle index using market asset prices Step 2:Smoothing process using
33、HP filter Step 3:Sinusoid fit and calculation of its angle at each point(month)of the smoothed cycle index Step 4:Definition of the cycles using the calculation in Step 3 Step 5:Analysis and calculation of the historical duration of each phase of the cycle index Strategy recommendations based on the
34、 difference between the phase of the cycle(extracted from market prices)and the macro view from BNPPs economics team 5 19 March 2019|DEEP DIVEConclusions and implications United States We applied the methodology for the US.Our index showed that it was in line with GDP activity in 2008(Fig.24),turnin
35、g negative almost a quarter before GDP started to decline substantially.It is interesting to note that US economic cycles follow a well-defined sequence compared to the other countries listed below.Our current value indicates that the US is in a late stage cycle,where the peak of the current cycle w
36、as reached during the second half of 2018.In order to have a clearer picture of market views and expectations of the US cycle,we included the historical duration of each cycle and the percentage of the current cycle(see Fig.24).Fig.24:Historical results and economic cycles in the US Sources:Bloomber
37、g LLP,BNP Paribas We went a step further to extract the most out of our framework.We developed a monitor to keep a watch on the current stage of the cycle(Fig.25)and included how long the economic cycles have lasted historically(Tables 2-3).Apart from the deep recession of 2008,the average duration
38、for bottom out cycles in the US has been eight months,due to the short periods of growth below potential cycles.The US has recently entered on a late stage cycle,still at 12.6%of its duration.In order to estimate the expected time remaining during this cycle,we took into account the historical avera
39、ge as well as the speed at which it has moved until now.Our results indicate that the US economy may enter a bottom out cycle in 7.5 months from now,that should last approximately eight months,according to the historical average.Fig.25 and Tables 2-3:US economic cycle watch and historical table Sour
40、ces:Bloomberg LLP,BNP Paribas 024681 0-0.6-0.4-0.20.10.30.5200020032005200720092012201420162018A-Growth pontetialB-EarlyC-MidD-LateInitial indexHP filtered indexSinusoidal model indexA408.0204B326.475C9515.8224D518.5134D12.6%2.927%7.5Current Cycle%of the cycleMonths elapsedExpected remaining monthsS
41、peed of current cycle to averageCycleTotal time(months)Average timeLongest timeShortest time-1.0-0.50.00.51.0-1.0-0.50.00.51.0C-MidD-LateB-EarlyCurrently at 12.6%of a D cycleA-Growth below potential/Bottom out 6 19 March 2019|DEEP DIVEChina Using the same method for China,we got results that were si
42、milar to US cycles.During 2001-07 China went through a long period of recovery with only a short bottom out period in 2005.However,in 2008 China went through a stronger bottom out stage,but with a shorter duration than in the US.Since then,China has been through short economic cycles,most of them sh
43、owing an economic recovery.What caught our attention is that our index decreased to negative levels last year and has now moved to the beginning of a bottom out cycle(see Fig.26).Fig.26:Historical results and Chinas economic cycles Sources:Bloomberg LLP,BNP Paribas Similar to the US,Chinas mid stage
44、 cycles last longer than other cycles,usually lasting for almost two years.Our results show that China has already entered into a bottom out cycle,according to market prices.We took the opportunity to also estimate how long it may last.Our results indicate that during the next three to seven months,
45、Chinas economy is likely to go through a bottom out period.Fig.27 and Tables 4-5:Chinas economic cycle watch and historical table Sources:Bloomberg LLP,BNP Paribas 024681 0-0.8-0.4-0.10.30.7200020032005200720092012201420162018A-Growth pontetialB-EarlyC-MidD-LateInitial indexHP filtered indexSinusoid
46、al model indexA369.0116B338.3106C9819.6319D518.5117A31.4%3.966%6.6Current Cycle%of the cycleMonths elapsedSpeed of current cycle to averageExpected remaining monthsCycleTotal time(months)Average timeLongest timeShortest time-1.0-0.50.00.51.0-1.0-0.50.00.51.0C-MidD-LateB-EarlyCurrently at 31.4%of a A
47、 cycleA-Growth below potential/Bottom out 7 19 March 2019|DEEP DIVEBrazil According to our model,the market perception is that Brazil went through its longest bottom out cycle since 2003 during 2016-2017 and it is in its early stage phase,but in a more advanced stage than our last report,with the in
48、dex already moving to the positive side.It is also interesting to note that during the Temer tantrum of 2016 there were signs of an early stage,but given the fast pace of the change,our framework identified it as a long bottom out period.Fig.28:Historical results of Brazils economic cycles Sources:B
49、loomberg LLP,BNP Paribas Table 7 also estimates how long the current cycle is likely to last;assuming the cycle speed will be a compounded average of the historical speed and current cycle speed.According to our results,the current cycle should last for five more months,and we could expect economic
50、activity to peak during the middle of the mid-stage cycle,12 months from now(4.6+14.32).Fig.29 and Table 6-7:Brazils economic cycle watch and historical table Source:Bloomberg LLP,BNP Paribas 024681 0-0.6-0.4-0.20.00.20.40.6200420052006200820092010201220132014201620172018A-Growth pontetialB-EarlyC-M