TOP DOWN STOCK SELECTION Incorporating macro views in ...

Finans Yatırım Menkul Değerler A.Ş. Esentepe Mahallesi Büyükdere Caddesi Kristal Kule Kat:5-6-7 Şişli İstanbul Turkey

TOP DOWN STOCK SELECTION Incorporating macro views in stock picking

- Changes in macro variables are always important for equity investors. Macro variables may affect stock

earnings and multiples and thus valuations, through different channels. Therefore equity investors may

benefit from incorporating their macro views in their stock picking process.

- Forming a macro view concerning the future developments of macro variables and then selecting stocks

according to this view may be referred as a top down stock selection process.

- In this research note we have analyzed the relationship between macro variables and stock returns and

propose a systematic approach for top down stock selection to transform macro views into investable

stock ideas.

- Using time series regressions and correlations, we have demonstrated that some of the macro variables

bear a statistically significant and time varying relationship with BIST100 index returns. Furthermore,

with the help of stepwise regressions, we conclude that macro variables also have an important bearing

on the share price as well.

- If a change in a macro variable directly affects the future earnings of a company, then the company is

referred to have a fundamental exposure to that macro variable. Using fundamental exposures may

support profit from a macro view in some cases (see Soda Sanayi case in Figure 9), but may lead to

unintended results in other cases (see cases for Arcelik, Turk Telekom and Garanti in Figures 10-12).

- We directly model the relationship between macro variables and stock performances using dynamic

regressions and obtain macro betas for each stock. By using stocks that have the highest/lowest macro

betas, we then form macro tracking portfolios (see Figure 1), which are market neutral long/short

portfolios, designed to track the macro variables as much as possible.

- We analyze the profitability of the macro tracking portfolio approach first by assuming perfect oversight

on macro variables. The results demonstrate that our macro tracking portfolio approach is a useful tool

in profiting from our macro views, if they turn out to be correct.

- Then by using a business cycle clock, we analyzed the effects of the business cycle on macro

variables and equity returns. Our analysis finds that macro variables and price, earnings and valuation

multiples for the BIST100 index exhibit quite distinct behavior in each of the four phases of the business

cycle. Thus it would be possible to benefit from the information concerning the current phase of the

business cycle in investment decisions.

- On this basis, we propose a systematic business cycle approach to stock selection. Our backtests

demonstrate that this approach may lead to outperformance over the benchmark index (see Figure 2).

Figure 1 – Oil Price and the Macro Tracking Portfolio Figure 2 – Performance of Business Cycle Portfolio

Source: Finansinvest Source: Finansinvest

Quant Blocks

Ayhan Yüksel, CFA

+90 212 336 7271

[email protected]

TOP DOWN STOCK SELECTION / 12 Aug 2015 2

Macro views and stock selection

Changes in macro variables are always important for equity investors. Macro variables may

affect stock earnings and multiples, and thus valuations, through different channels. For

instance improving economic growth may boost a company’s sales, an increase in inflation

may drive input costs higher, and the depreciation of a currency may depress earnings by

increasing the cost of foreign currency denominated debt.

Therefore equity investors may benefit from incorporating their macro views into

their stock picking process. Forming a macro view concerning the future developments

of macro variables and then selecting stocks according to this view may be referred as a

top down stock selection process. Top down stock selection is an alternative means of

stock picking to the bottom up stock selection where stocks are selected by considering

their firm-specific characteristics such as sales growth, EBITDA margins and recent price

momentum.

With the aim of proposing a systematic approach for top down stock selection, in this

research note we have analyzed the following questions from an equity investor’s point of

view:

Does certain set of macro variables affect stock returns?

If so, how can we incorporate this into stock selection?

And, how to profit from such an approach?

Do macro variables affect stock returns?

In short, the answer is “Yes”, they affect. We begin our analysis by considering the

relationship between macro variables and share price performances. To this end, we

consider eight macro variables that potentially affect stock performances and are widely

used by equity investors. The selected variables are shown in Figure 3. We have obtained

monthly data for each variable1 and constructed a monthly return either by differencing the

data (for variables that express an annual growth rate or yield) or taking the percentage

change (for variables that represent a level). Volatility and correlation statistics for macro

variables are provided in Figure 29 in the Appendix.

Figure 3 – Selected macro variables

Macro Theme Variable Used Data Source

Return Calculation

Potential Direct Effects on Fundamentals

Economic Growth Leading Indicator 2 CBRT % Change Sales growth, EBITDA margin

Exports YoY Growth in Exports CBRT Difference Sales growth, EBITDA margin

Confidence Consumer Confidence Index CNBC-e % Change Sales growth, EBITDA margin

Inflation YoY Growth in CPI CBRT Difference COGS, sales price, OPEX

Interest Rates Benchmark Bond Yield Matriks Difference Interest Expenses

USD/TRY Continuous Futures Index 3 Bloomberg % Change FX Revaluation Gain/Loss, COGS, sales price, OPEX

EUR/USD Spot Exchange Rate Bloomberg % Change FX Revaluation Gain/Loss, COGS, sales price, OPEX

Oil Price Spot Brent Oil Price Bloomberg % Change COGS, sales price, OPEX

Source: Finansinvest

We begin analyzing the relationship between macro variables and stock performances by

using time series regressions and correlations, both on the BIST level and the individual

stock level. We first analyzed the lead/lag correlations between returns of macro variables

and BIST100 index returns. The results are presented in Figure 4. In this figure, the x-axis

shows the number of periods used in lagging the macro returns - i.e. a value of -1

1 In our analysis, we assume that all macro data is available at the end of corresponding month, which may not be the case in real life, due to delayed announcements and backward revisions in macro data. 2 The composite leading indicator, as constructed by the CBRT, aims to track and lead the monthly industrial production index for Turkey. Here we use the cyclical component of the composite leading indicator index, which is estimated by stripping out the long term trend and seasonal and irregular components from the original series. 3 For the USD/TRY exchange rate, because of the significant interest rate differentials that may create drifts in the spot exchange rate, we use a continuous futures price index instead of the spot price. We construct the index assuming that we always invest in the front month contract and roll our position into the second month contract 2 days before the expiry of the front month contract. Such a continuous futures price index better captures unexpected changes in exchange rates.

Macro views are important in stock selection ....

... and may be used in a top down stock selection process

We analyzed the effects of eight macro variables on stock returns


corresponds to the correlation of the BIST100 index return with the return of a macro

variable that is lagged by one month. Contemporaneous correlations are provided on the

x-axis value of 0. Our macro stock picking approach assumes that at month t-1, we already

hold a macro view and thus we already have an expectation concerning the month t

realization of the macro variable. Therefore, we are interested in the month t realization of

the stock market performance. This means that the contemporaneous correlations are

the ones that are important in our analysis.

Contemporaneous correlations are high (in absolute terms) and statistically significant4 for

leading indicator, the consumer confidence index, bond yield, oil price and USD/TRY

futures index, while they are low and non-significant for inflation, export growth and the

EUR/USD exchange rate. Additionally the 36 month rolling correlations shown in Figure 5

also confirms that the relation between macro variables and the BIST100 index return is

dynamic and may evolve over time. Our analysis found that some of the macro variables

exhibit a statistically significant and time varying relation with BIST100 index returns.

Figure 4 – Lead/lag correlations with BIST100 index returns Figure 5–36 Month rolling correlations with BIST100 index returns


To analyze the effects at the stock level, we performed a stepwise regression approach

using a 72 stock universe. In this approach, we performed the following steps for each

stock:

Using all available data, we first estimated the following univariate regression using

only the BIST100 index returns as the explanatory variable:

𝑅𝑡𝑖 = 𝛼 + 𝛽𝑅𝑡

𝐵𝐼𝑆𝑇 + 𝜀𝑡

We then deployed a stepwise search procedure to test whether adding each one of

the eight macro variables to the above regression improves the model performance in

terms of the statistical goodness of fit measure5. If the model fitness improves, then

we would add that macro variable to the regression. For example, if the third macro

variable is added, the new regression equation is as follows:


𝐵𝐼𝑆𝑇 + 𝛽𝑀3𝑅𝑡𝑀3 + 𝜀𝑡

We repeat the second step until such a point that the addition or removal of any of the

macro variables does not improve the model performance.

The final model may (or may not) include some of the macro variables. For example,

for a stock, the final model may only include the third and seventh macro variables:


𝐵𝐼𝑆𝑇 + 𝛽𝑀3𝑅𝑡𝑀3 + 𝛽𝑀7𝑅𝑡

𝑀7 + 𝜀𝑡

In Figure 6, the blue bars represent, for each stock, the adjusted-R2 values from the

univariate regression using only the BIST100 index returns. Furthermore the incremental

4 Two sided hypothesis testing is performed at a 90% confidence level. The estimated correlation coefficients and the p-values from the hypothesis testing are provided in Figure 30 in the appendix. 5 We used the Akaike Information Criterion for this step.

Some of the macro variables have a significant and time-varying relation with BIST100 returns

We deploy a stepwise regression approach to analyze the macro effects at the stock level ...


increase in these values when we use the final model (which may also include macro

variables) are shown in grey. For instance for Tupras, the adjusted-R2 values are 0.49 for

the univariate regression and 0.54 for the final model which includes, beside BIST100 index,

oil price and USD/TRY futures index. For 93% of the stocks that we analyzed, the adjusted-

R2 values increase when we add macro variables to the regressions. Our analysis finds

that macro variables have some additional explanatory power (apart from the portion

explained by the general market returns) in explaining the time series variation of stock

returns. The incremental increases in the adjusted-R2 may appear small, especially when

compared to the already high adjusted-R2 values obtained in univariate regressions.

However, in predictive modeling, even small explanatory powers may give rise to significant

investment returns, as we also demonstrate in the following sections. We also report the

number of times a macro variable is selected in our stepwise search procedure in Figure 7.

The leading indicator, oil price, EUR/USD rate and bond yield are the macro variables that

are selected most often.

Figure 6 – Adjusted-R2 values Figure 7 – Frequency of macro variables selected

Macro Variable Frequency

Leading Indicator 29

Inflation 13

Consumer Confidence 18

Export Growth 19

Bond Yield 20

EUR/USD 21

Oil Price 24

USD/TRY Futures Index 18


How to incorporate macro views in stock selection

In this research note, we are focusing on the following investment problem: “Given a view

about the future developments of a macro variable, how can we profit from this view by

using stock selection, if it turns out to be correct?”

The analysis presented in the previous section demonstrates that the relationship between

macro variables and stock performances may not be uniform among different stocks or

over time. Some stocks are more sensitive than others, some stocks may benefit from an

increase in a macro variable while others may suffer from it. Furthermore the effects of a

macro variable over stock performances may weaken or strengthen over time.

Here we assume that certain macro variables affect different stocks in different directions

and magnitudes, thus leading to a cross sectional dispersion among stock performances.

Therefore, as opposed to the absolute stock performances which are affected by general

market movements, we are interested in the relative performances net of market

movements.

In order to profit from a macro view, two conditions need to be satisfied:

1) The first and foremost requirement is of course good forecast accuracy. Forecasting

macro variables (or at least direction of them) accurately in a consistent manner, which

is not an easy task, is the basic prerequisite for profiting from a top down stock

selection approach.

2) Even if we assume a perfect oversight on macro variables, we need a predictable

relationship between macro variables and the performance of stocks that we select

through a top down selection approach.

To illustrate this, assume a perfect world where the relative performance of a stock is

strongly related with the level of a macro variable, as depicted in Figure 8. For example, at

time T, an equity investor who is predicting that the macro variable will increase in the future

...which shows that macro variables are also important at the stock level

To profit from a top down stock selection approach, we need accurate forecasts on macro variables and a predictable relationship between macro variables and stock perfornances


may invest in the stock shown in the figure whose relative performance is tightly related to

the level of that macro variable. If his view turn out to be correct and the macro variable

increases after time T, plus the relationship between the macro variable and the relative

stock performance continues to hold after time T, then the investor would profit from his

macro view. However, actual relationships between macro variables and stocks are far

from perfect in real life. Nevertheless the relationship between a macro variable and the

relative performance of a stock is the first step in analyzing macro effects. The main focus

of our research note is to employ a systematic method to form a portfolio of stocks that

closely matches the developments in a macro variable in the Turkish market.

Figure 8 – A Perfect relationship between a Macro Variable and relative performance of a Stock

Fundamental exposure method

One approach to select stocks with a macro view is to find firms that have a direct

“fundamental exposure” to that macro factor. Fundamental exposure refers to the cases

where changes in a macro variable has a direct effect on future earnings of a company.

Examples of such an exposure may include the following:

Accelerating economic growth in export markets may affect the sales revenue of an

exporting firm

An increase in interest rates may increase the interest expenses of a highly indebted

firm

The depreciation of a currency may cause a foreign exchange revaluation loss for a

firm with an unhedged position in that currency

An increase in raw material or energy prices may increase the cost of goods sold for

a firm that uses such materials and energy as inputs in its production process

The fundamental exposures may be positive or negative, meaning that the firm may benefit

or suffer from an increase in that macro variable. In this approach, if we hold a view for a

macro variable, we invest in the firms that have the highest fundamental exposure to that

macro factor. For instance, an investor that expects a depreciation of the TRY against the

USD will screen stocks according to their net unhedged currency position and invests in

firms with the highest short TRY position.

Using fundamental exposures may help profit from a macro view. For instance, Soda

Sanayi, the soda ash producer for a variety of industries, has an operating structure such

that hard currencies has a higher share in its revenues than in its costs. Thus the firm has

a fundamental short exposure on the TRY vs hard currencies, the USD and the EUR.

Because of this, any depreciation (or appreciation) of the TRY would boosts (or harm) the

earnings of the firm. In addition, the recent relative outperformance (or underperformance)

of the stock would generally coincide with the depreciation (or appreciation) of the TRY, as

shown in Figure 9.

0

20

40

60

80

100

120

140

160

180

200

Macro Variable Relative Performance of Stock

T

Using fundemental exposures may help profit from a macro view in some cases ...

If a change in a macro variable directly affects the future earnigns of a company, then the company is referred to have a fundemental exposure to that macro variable


However, using the fundamental exposure method does not always bring the intended

results:

Arcelik also has a currency mismatch in its operations where a significant portion of its

revenues are denominated in EUR terms, while a significant portion of its costs are

based in USD. Therefore the firm has a fundamental long exposure on the EUR/USD.

Figure 10 shows that although there are periods in which we have a positive relation

between stock’s relative performance and EUR/USD, the relationship is not stable

and has switched sign in some time periods. An investor that shorted Arcelik at the

beginning of 2012 with an anticipation of a further depreciation of the EUR vs USD

would have lost money, even the firm continued to hold positive fundamental exposure.

Turk Telekom has had a high short FX position driven by FX denominated debt and

thus has a fundamental short position in the USD/TRY. However, as shown in Figure

11, the relative performance of the stock is positively related to changes in the

USD/TRY futures index, as opposed to the negative relationship suggested by

fundamental exposure. Thus the relationship is the exact opposite of what

fundamental exposures suggest. The main reason behind this is that the periods

when the TRY depreciates against the USD typically coincide with general market sell-

offs where defensive stocks like Turk Telekom tend to outperform the general market.

Garanti Bank, as a commercial bank with no non-financial subsidiary, does not have

any fundamental exposure to oil prices. However, oil prices are typically positively

correlated with the general risk appetite and thus high beta stocks may outperform the

market when oil prices increase. This relationship also holds for Garanti Bank in

general, as depicted in Figure 12. This indicates that a significant relationship between

a macro variable and the relative performance of a stock may arise even the firm

does has no direct fundamental exposure to that macro variable.

Figure 9 – Fundamental exposures may help … Figure 10 – …but the relation may not be stable …


Figure 11 – …sometimes can be just the opposite … Figure 12 – …or may not be the result of fundamental exposure


... but may lead to unintended results in some other occasions ...


These examples reveal that although fundamental exposures are important in analyzing

relationships between macro variables and stock performances, using only these

exposures may miss sight of a bigger picture. To understand this, the conceptual

framework for the linkages between a macro variable and the value of a stock is presented

in Figure 13. The stock value is simply determined by the expected future cash flows,

discounted by an appropriate discount rate which includes two components - the risk free

rate and a risk premium. The fundamental exposure approach considers only the effects of

a macro variable on the expected future cash flows of the firm (i.e. the straight line in the

figure). However, the macro variable may also affect the level of risk free rate and/or risk

premiums. For instance, as per our previous examples, the USD/TRY rate or oil price may

affect the level of risk free rate and risk premiums. Furthermore, the macro variable that we

are interested in may be significantly correlated with another macro variable which may

affect share prices. In some cases, two highly correlated macro variables may influence

the price of a stock in an opposing way, as we discussed in the Turk Telekom example.

Figure 13 – Macro variables and stock valuation

Our analysis in previous sections and the framework presented in Figure 13 reveal the

need for a more general approach in capturing both the direct and indirect effects of a

macro variable on share price performances. An efficient approach should incorporate the

following properties of the relationship between macro variables and stock performances:

The relation is dynamic which may:

o evolve through time

o exist only for certain time periods

o switch sign

The relationship is not restricted to stocks with fundamental exposure to the macro

variable under question

The relationship should be assessed after filtering out the overall market’s impact on

a stock by beta hedging. In other words, the general market exposure of a stock should

be properly adjusted in analyzing macro effects

In order to account these properties, we take a statistical approach by directly modeling

the relationship between a macro variable and the stock performance and construct

macro tracking portfolios. We will discuss the details of this approach in the following

section.

Macro tracking portfolio approach

Our approach is based on estimating statistical relationships between macro variables and

stock performances and does not assume any fundamental exposure. We first begin by

estimating the sensitivity of stock performances to macro variables, also allowing these

sensitivities to be time varying. In this vein, we employ the following dynamic regression

model:

𝑅𝑡𝑖 = 𝛼𝑡 + 𝛽𝑡𝑅𝑡

𝐵𝐼𝑆𝑇 + 𝛽𝑡𝑀𝑅𝑡

𝑀 + 𝜀𝑡

𝛼𝑡 = 𝛼𝑡−1 + 𝜂𝑡

𝛽𝑡 = 𝛽𝑡−1 + 𝜐𝑡

𝛽𝑡𝑀 = 𝛽𝑡−1

𝑀 + 𝜔𝑡

... and misses some parts of a bigger picture ...

We directly model the relationship between macro variables and stock performances using dynamic regressions ...

...which necessiates a more general approach


where 𝛼𝑡 is the time-varying intercept term, 𝛽𝑡 and 𝛽𝑡𝑀 are the time-varying betas of the

stock with respect to the general market (BIST100 index) and macro factor 𝑀 . All

disturbance terms are independent and have normal distribution. The model is estimated6

separately for each stock and using one macro factor at a time.

The results of these regressions give us macro betas 𝛽𝑡𝑀 (for each stock) which are the

main inputs of our macro tracking portfolio approach. Some examples of macro betas are

given in Figure 14-17. In order to make comparison with general market betas simple, these

figures include the macro betas that are properly scaled7 to match the volatility of BIST100

index and macro variable returns.

Figure 14 – Scaled Beta of EREGL to Leading Indicator Figure 15 – Scaled Beta of KCHOL to Inflation


Figure 16 – Scaled Beta of AYGAZ to USD/TRY Futures Index Figure 17 – Scaled Beta of ARCLK to EUR/USD


After obtaining the general market and macro betas for each stock, we form macro

tracking portfolios by including the stocks that have the highest beta values in absolute

terms to the corresponding macro variable (i.e. those stocks which are most sensitive to

the macro variable in question). We constructed three tracking portfolios for each macro

variable:

Tracking Portfolio (+): This portfolio (hereafter referred to as TP(+)) is composed of

stocks that benefit from an increase in the macro variable. At the end of month t, we

form TP(+) by taking equal weighted long positions in the 10 stocks which have the

6 The parameters of this state-space model is estimated using maximum likelihood and time varying

coefficients (𝛼𝑡, 𝛽𝑡 and 𝛽𝑡𝑀 ) are estimated using Kalman filter recursions.

7 Standard regression arithmetic suggests that regression betas are obtained using a normalization by the volatility of independent variable. In our case, different macro variables have different units of measurement and thus volatility levels, and this makes comparing macro betas difficult. To make different macro betas comparable with each other and with general market beta, we scale them by multiplying the volatility of the corresponding macro variable return and dividing by the volatility of BIST100 returns.

... and obtain macro betas for each stock ...

... and form macro tracking portfolios using stocks that have the highest/lowest macro betas ...


highest macro beta, 𝛽𝑡,𝑖𝑀, for the corresponding macro variable 𝑀. At the same time,

in order to hedge the general market exposure of these stocks, we take short positions

in the BIST100 index with an amount equal to the average of market betas of these

stocks 1/10 ∑ 𝛽𝑖,𝑡. Hence, we are forming a market neutral portfolio of stocks with

the highest positive sensitivity to the macro variable. We maintain our positions until

the end of the following month, and then repeat the same steps for the month thereafter.

For such a portfolio constructed using the macro betas available at time t, the next

period return would be:

𝑅𝑡+1𝑇𝑃(+)

=1

10∑(𝑅𝑡+1

𝑖 − 𝛽𝑖,𝑡𝑅𝑡+1𝐵𝐼𝑆𝑇)

10

𝑖=1

Tracking Portfolio (-): This portfolio is composed of stocks that benefit from a decrease

in the macro variable. We constructed TP(-) with a similar procedure as defined above,

but using the 10 stocks with the lowest macro betas.

Tracking Portfolio: This portfolio is a long/short market neutral portfolio that aims

to capture the effects of the macro variable on cross sectional return differences

of stocks that are highly sensitive to that macro variable. We construct TP by taking a

unit long position in TP(+) and a unit short position in TP(-).

The final tracking portfolio, the TP, bears certain characteristics:

The portfolio includes long and short positions in a total of 20 stocks, selected

according to the macro betas at the beginning of period.

The general market exposure of this portfolio is properly hedged using opposing

positions8 in the benchmark index according to the market betas of the selected

stocks. Thus, the portfolio is a market neutral portfolio.

Since the portfolio has unit long and unit short exposures, it is a cash-neutral

portfolio9.

The portfolio reflects the performance differential between stocks that exhibit a

high positive sensitivity to the macro variable and those exhibit a high negative

sensitivity to the same macro variable.

Thus, by construction, we expect the TP to closely track the original macro

variable as much as possible.

For macro variables that have significant influence on the cross sectional return

differences among stocks, TP’s closely track the original macro variables in most

time periods. Two examples of such a case are depicted in Figures 18 and 19. The figures

include the spot oil price and the USD/TRY futures index as well as corresponding TP’s. In

these cases, the TP’s closely track the related macro variables.

However the relationships between macro variables and TP’s are not as strong as these in

all cases. The TP’s for all macro variables are presented in Figures 31-38 in the appendix.

Indeed, the strength of the relationship depends on two variables:

The strength of the macro effects on the cross sectional return differences: If a macro

variable has a weak effect on stock performances, the corresponding TP may not

closely track the macro variable.

Other systematic or firm specific issues that influence stock performances: TP’s are

formed by assuming that the stock performances are driven only by the general market

index (the BIST100 index), and the corresponding macro factor. However, there may

be other systematic factors10 that affect stock performances but are omitted from our

analysis. Such a situation may degrade the tracking performance of TP’s.

8This hedging can be implemented using BIST30 index futures. 9To ensure equal contributions from the TP(+) and TP(-), it is also possible to construct the TP by taking a unit long exposure in TP(+), plus a short exposure in the TP(-) with an amount equal to the ratio of the macro betas of these two portfolios. Such macro beta adjusted TP’s better track the corresponding macro variable in some cases. However, they would not be cash neutral, which may create additional hurdles in real life implementations. 10 For instance, as shown in our previous paper entitled A Factor Based Stock Selection Model for Turkish Equities, style factors such as value, momentum, profitability, growth, size and sentiment strongly influence relative stock performances.

...which are market neutral long/short portfolios, designed to track the macro variable as closely as possible

https://www.finansonline.com/DosyaYoneticisi.ashx?FileCode=A0THANA7KYFLLILCEC8D

https://www.finansonline.com/DosyaYoneticisi.ashx?FileCode=A0THANA7KYFLLILCEC8D


Nevertheless, in systematic investing, we do not always require very strong relationships.

Since we can adjust our portfolio periodically, a modest level of correlation between returns

of TPs and macro variables is the only thing that we need in practice. The

contemporaneous correlations are in the range of 0.15 to 0.54 and such levels may give

rise to significant investment returns. We have analyzed the profitability of the macro

tracking portfolio approach in the next section.

Figure 18 – Oil Price vs Tracking Portfolio Figure 19 – USD/TRY Futures Index vs Tracking Portfolio


How to profit from a top down stock selection approach?

After establishing TP’s that track macro variables, our next question would be how to exploit

the benefits of such portfolios within a systematic investment process. In this section, we

consider two cases and analyze the profitability of such a process. While the first case

assumes perfect forecasting ability for macro variables, the latter encompasses a business

cycle investing approach that requires the knowledge of the current phase in the business

cycle. In our analysis, we ignore the transaction fees and costs associated with shorting

stocks, so it does not represent a simulation of a real-world trading strategy.

Part I - A timing strategy with a perfect oversight on macro variables

In the first case, we assume that we have a perfect oversight on the future developments

of a macro variable. Specifically, we assume that we precisely know the turning points for

the macro variable and adjust our positions at those dates. We have illustrated this idea in

Figure 20, where we plot the leading indicator along with its turning points.

To analyze the benefits of such an approach, we performed backtests for each macro

variable. Since TP’s are cash-neutral portfolios, we combine them with a unit exposure in

the BIST100 index. Such a portfolio may represent the portfolio of a benchmark-relative

investor where the investor over or underweights certain stocks. The process of

constructing the portfolio process is as follows:

If we expect the macro variable to increase in the future, we take a unit long position

in the BIST100 index, plus a unit long position in the corresponding TP

If we expect the macro variable to decrease in the future, we take a unit long position

in the BIST100 index, plus a unit short position in the corresponding TP

The results of such a timing strategy for the TP that tracks the leading indicator are

presented in Figure 21. We compared this strategy with a simple buy and hold long position

in the BIST100 index and found that the strategy significantly outperformed the benchmark

in this period. The strategy performance for all macro variables is shown in Figures 39-46

in the appendix.

We analyzed the profitability of the macro tracking portfolio approach first by assuming perfect oversight on macro variables ...


Figure 20 – Leading Indicator with Turning Points Figure 21 – Timing Strategy Using TP for Leading Indicator


The performance results of our timing strategy for all macro variables are provided in Figure

22. For all macro variables, the strategy outperformed the benchmark by a significant

margin, leading to high information ratios and hit ratios. The results demonstrate that our

macro tracking portfolio approach is useful in profiting from our macro views - if

they indeed turn out to be correct.

Figure 22 – Performance of macro timing strategy

Macro Variable Annualized

Outperformance vs BIST100

Annualized Information

Ratio

% of Months With

Outperformance

Leading Indicator 10.6% 0.54 58.5%

Inflation 9.1% 0.63 63.4%

Consumer Confidence 3.5% 0.32 55.8%

Export Growth 9.4% 0.63 58.5%

Bond Yield 31.0% 1.42 66.7%

EUR/USD 13.7% 0.81 60.2%

Oil Price 14.9% 0.88 62.9%

USD/TRY Futures Index 28.2% 1.38 70.5%

Source: Finansinvest

Part II - Adjust your clocks: Business cycle investing

Economic activity typically presents a cyclical characteristic that includes alternating

periods of relatively rapid growth and slowdowns/contractions. This fluctuating nature of

the economy is characterized with business cycles in four phases: recovery, expansion,

slowdown and contraction. Most macro variables are correlated, so that their behavior is

typically modestly synchronized, but with some variables tending to lead or lag others. Thus

the prevailing phase of the business cycle usually has profound effects on the behavior of

different macro variables.

Identifying the phases and turning points of the business cycles is not straightforward. One

commonly used variable to define the business cycle phases is the smoothed deviation of

gross domestic product (or any other variable that represents current economic activity)

from its long term trend.

The leading indicator index11 that we used in previous sections aims to track and lead the

cyclical component of the monthly industrial production index for Turkey. The index is found

to be useful in finding turning points in economic activity12. We took a systematic approach

11 For the details of index and its performance, see the CBRT research paper entitled “Methodological Changes in the Composite Leading Indicator for Turkish Economic Activity”. 12 The turning points for GDP and the industrial production index are typically similar and thus the cyclical component of the industrial production index is considered as a sound proxy for the fluctuation of overall economic activity. The leading indicator leads the turning points of industrial production index with a median lead time of 3 months.

...and found that our macro tracking portfolio approach may enable profits from macro views

The current phase of the business cycle usually has profound effects on the behavior of different macro variables

We use a business cycle clock to define the current phase of business cycle ...

http://www.tcmb.gov.tr/wps/wcm/connect/44ad4ca1-c266-4acd-a27e-826448def6d8/EN1404eng.pdf?MOD=AJPERES&CACHEID=44ad4ca1-c266-4acd-a27e-826448def6d8

http://www.tcmb.gov.tr/wps/wcm/connect/44ad4ca1-c266-4acd-a27e-826448def6d8/EN1404eng.pdf?MOD=AJPERES&CACHEID=44ad4ca1-c266-4acd-a27e-826448def6d8


and employed a business cycle clock derived from leading indicator index to define

the phases of business cycles13. We use the level and growth of leading indicator index to

define the following four phases of the business cycle:

Expansion: The index level is above 100 and has increased in the last month

Slowdown: The index level is above 100 and has decreased in the last month

Contraction: The index level is below 100 and has decreased in the last month

Recovery: The index level is below 100 and has increased in last month

Employing the business cycle clock approach, we identify the phases of the cycle for each

month from February 2000 to May 2015. The business cycle clock for the period between

January 2013 and May 2015 is presented in Figure 23. As shown from the clock, Turkey

was in an expansionary phase in the third quarter of 2014 before slipping into the slowdown

phase in October 2014 and rapidly passing through a contraction phase thereafter.

Although the contraction then decelerated towards the end of the period, the May 2015

level still points to a contraction phase. The transition matrix for the four phases is provided

in Figure 24 which shows that the contraction phase was followed by a recovery in 6 out of

30 cases (or 20% of the time).

Figure 23 – Business Cycle Clock for Turkey Figure 24 – Transition Matrix for Phases of Business Cycle

Next Phase

Recovery Expansion Slowdown Contraction

Curr

ent

Phase

Recovery 28 4 0 2

Expansion 0 47 10 0

Slowdown 0 5 52 5

Contraction 6 0 0 24

Source: CBRT, Finansinvest Source: Finansinvest

We then analyzed the behavior of six domestic macro variables as well as price, earnings

and the valuation multiple of the BIST100 index in different phases. Figure 25 includes the

average monthly returns for macro variables in different phases:

For the leading indicator, the consumer confidence index and export growth, beginning

from a positive level, returns decay during the transition from recovery to contraction.

For the USD/TRY futures index, a consistent increase is found during the transition.

For inflation and interest rates, we find a different behavior where the variables have

the lowest returns in the recovery phase, while returns peak in the slowdown phase

and weaken thereafter.

The growth rates in the price, earnings and the price/earnings (P/E) ratio for the BIST100

index in different phases are presented in Figure 26. As shown from the figure:

Recovery Phase: The BIST100 index returns are highest in this phase. Trailing

earnings still continue to contract while expected forward earnings are improving. On

the other hand, valuation multiples were found to start to grow rapidly, possibly from a

depressed level in the previous contraction phase.

Expansion Phase: In this phase, both the earnings and valuation multiples grow, giving

rise to high BIST100 index returns. Although the highest level of earnings growth is

seen in this phase, the expansion in multiples is more muted than the rapid expansion

in the recovery phase.

13 More detailed discussion on the business cycle clock approach for Turkey may be found in the CBRT research paper entitled “Analysis of Economic Activity in the Perspective of Composite Indices and the Business Cycle Clock”. The OECD has a web site (http://stats.oecd.org/mei/bcc/default.html) on business cycle clocks for different countries which includes an instructive animation of how the clocks have evolved over time.

... and find that macro variables and price, earnings and BIST valuation multiples all have exhibit distinct behavior in each of the four phases of the business cycle ....

http://www.tcmb.gov.tr/wps/wcm/connect/2ed63608-9679-4db0-880a-ef0d106d6498/EN1202eng.pdf?MOD=AJPERES&CACHEID=2ed63608-9679-4db0-880a-ef0d106d6498

http://www.tcmb.gov.tr/wps/wcm/connect/2ed63608-9679-4db0-880a-ef0d106d6498/EN1202eng.pdf?MOD=AJPERES&CACHEID=2ed63608-9679-4db0-880a-ef0d106d6498

http://stats.oecd.org/mei/bcc/default.html


Slowdown Phase: This phase is characterized by continued growth in earnings while

valuation multiples fall from their high levels in the previous expansion phase. Average

BIST100 index return is negative in this phase.

Contraction Phase: A typical contraction phase includes contractions in both earnings

and the valuation multiple, leading to losses for the BIST100 index.

Our analysis finds that macro variables and price, earnings and valuation multiples for

the BIST100 index all exhibit some distinct behavior in each of the four business

cycle phases. Therefore we should consider the fact that these four phases have quite

distinct investment implications, and thus we cannot treat them as parts of a smooth and

uninterrupted process over time. The four phases of the business cycle may be considered

as “four seasons” for macro variables and investment returns. Consequently we may

benefit from using the information concerning the current phase of business cycle in our

investment decisions.

Figure 25 – Macro Variables in Different Business Cycle Phases Figure 26 – BIST100 Index in Different Business Cycle Phases

Recovery Expansion Slowdown Contraction

Leading Indicator

0.37% 0.15% -0.18% -0.47%

Inflation -0.29% -0.03% 0.06% 0.02%

Consumer Confidence

6.69% 1.61% -2.42% -3.49%

Export Growth

0.83% 0.24% -0.53% -1.32%

Bond Yield

-0.67% -0.21% 0.16% -0.17%

USD/TRY Futures

-1.42% -0.73% 0.27% 1.89%


To analyze the profitability of a business cycle approach to stock selection, we

performed a backtest14 using portfolios constructed as follows:

At the end of each month, we identify the current phase of the business cycle by using

business cycle clock approach.

Then, using the results presented in Figure 25, we have identified the macro variables

that exhibit a positive average return in that phase as those that need to be

overweighted in our portfolio. Likewise, macro variables with a negative average return

in that phase need to be underweighted in the portfolio. Such an approach implicitly

assumes that we expect the current phase of business cycle to continue in the next

month which is in most cases true, as shown in the transition matrix in Figure 24.

Next, we form a portfolio with:

o A unit long exposure in the BIST100 index

o A unit long exposure in an equally weighted basket of TP’s for the macro

variables that are overweighted.

o A unit short exposure in an equally weighted basket of TP’s for the macro

variables that are underweighted.

In contrast with the backtests that we presented in the previous section, where we assume

a perfect oversight on macro variables, our business cycle approach assumes that we only

have knowledge concerning a) the current phase of the business cycle (which can be

deduced from the final reading of the leading indicator) and b) which macro variables we

would over- or under-weight in our portfolio, depending on the current phase.

The performance of our business cycle portfolio is provided in Figures 27 and 28. The

portfolio outperformed the benchmark with an annualized excess return of 7.7% and

14 We ignore the transaction fees and costs associated with shorting stocks; thus our backtest does not correspond to the simulation of a real-world trading strategy.

... which shows that we may benefit from using the information concerning the current phase of the business cycle in our investment decisions

To test this idea we backtested a business cycle approach to stock selection ...

...and found that this approach may lead to an outperformance over the benchmark index


tracking error of 19.4%. The information ratio was materialized at 0.48. The portfolio

outperformed the benchmark in 53% of the months tested.

Figure 27 – Business Cycle Portfolio vs BIST100 Index Figure 28 – Performance of Business Cycle Portfolio

Annualized Excess Return 7.7%

Annualized Tracking Error 19.4%

Annualized Information Ratio 0.48

% of Months with Outperformance 53%


Conclusion

In this research note, we analyzed the relations between a certain set of macro variables

and stock performances, and propose a systematic approach for top down stock selection

to transform macro views into investable stock ideas. Our analysis reveal that our macro

tracking portfolio approach is a useful tool in profiting from our macro views, if they turn out

to be correct. It is also possible to use these tracking portfolios within a business cycle

investing approach.

Macro themes that have emerged in the recent months include a depreciation in TRY (along

with other emerging market currencies), rising interest rates and a further down leg in oil

prices. In the coming periods, we plan to combine our macro views with corresponding

macro tracking portfolios to make top down stock selection calls.


Appendix

Figure 29 – Volatility and correlations for macro variables Figure 30–Return Correlations

Lead Ind

Inf Conf Exp Bond Yield

EUR/ USD

Oil USD/ TRY

Volatility

Lead Ind 100% -12% 24% 14% -18% 25% 36% -32% 1%

Inflation -12% 100% -21% -13% 21% 3% 9% 0% 3%

Confidence 24% -21% 100% -5% -31% 4% 5% -29% 32%

Exports 14% -13% -5% 100% -9% 12% 25% -9% 45%

Bond Yield -18% 21% -31% -9% 100% -19% 8% 59% 4%

EUR/USD 25% 3% 4% 12% -19% 100% 52% -48% 11%

Oil 36% 9% 5% 25% 8% 52% 100% -38% 30%

USD/TRY -32% 0% -29% -9% 59% -48% -38% 100% 15%

Correlation with BIST100 (%)

p-value (%)

Lead Ind 25.1 0.06

Inflation -7.9 36.76

Confidence 27.5 0.04

Exports 8.3 26.06

Bond Yield -56.5 0.00

EUR/USD 8.7 23.90

Oil 14.7 7.71

USD/TRY -63.0 0.00


Figure 31 – Leading Indicator vs Tracking Portfolio Figure 32 – Inflation vs Tracking Portfolio


Figure 33 – Consumer Confidence Index vs Tracking Portfolio Figure 34 – Export Growth vs Tracking Portfolio



Figure 35 – Bond Yield vs Tracking Portfolio Figure 36 – EUR/USD vs Tracking Portfolio


Figure 37 – Oil Price vs Tracking Portfolio Figure 38 – USD/TRY Futures Index vs Tracking Portfolio


Figure 39 – Timing Strategy Using TP for Leading Indicator Figure 40 – Timing Strategy Using TP for Inflation



Figure 41 – Timing Strategy Using TP for Consumer Confidence Figure 42 – Timing Strategy Using TP for Export Growth


Figure 43 – Timing Strategy Using TP for Bond Yield Figure 44 – Timing Strategy Using TP for EUR/USD


Figure 45 – Timing Strategy Using TP for Oil Price Figure 46 –Timing Strategy Using TP for USD/TRY Futures Index



Finansinvest Rating System We employ a relative scale in our rating system (i.e. Market Outperform, Neutral, Underperform) in order to better present relative value propositions and more

actively pursue long vs. short ideas at the BIST. The relevant benchmark is the broader Turkish stock market, using the BIST-100 index as a basis. The ratings also incorporate a certain degree of relativity within the analyst’s own stock coverage universe due to asymmetric return expectations among the industries under our BIST coverage. The rating system combines analysts’ views on a stock relative to the sectors under coverage, and the sector call relative to the market, together providing a view on the stock relative to the market. Individual ratings reflect the expected performance of the stock relative to the broader market over the next 6 to 12 months. The assessment of expected performance includes a function of near-term company fundamentals, industry outlook, confidence in earnings estimates and valuation, and other factors. An essential element of our rating methodology involves benchmarking a 12-month expected return against the cost of equity. We set a required rate of return for each stock, calculated from our risk-free rate and equity risk premium assumptions. The price target for a stock represents the value that the stock is expected to reach or sustain over the performance horizon of 12 months, according to the view of the analyst. We have separated the stocks under our coverage into two groups, mainly with respect to their liquidity (market cap, free float market cap and historical average daily trading volume) as small-cap stocks exhibit different risk/return characteristics to more-liquid large-caps. For the purposes of the relative stock rating, however, stocks within each group will be considered on an unweighted basis with regard to their market capitalization. For a stock to be assigned an Outperform rating, the implied return must exceed the required rate of return by at least 5 percentage points over the next 12 months for our larger-cap stock coverage, or by 10 percentage points for the small-cap group. For a stock to be assigned an Underperform rating, the stock must be expected to underperform its required return by at least 5 percentage points over the next 12 months. Stocks between these bands will be classified as Neutral. When the potential upside of an average stock in our coverage exceeds its required rate of return (i.e. the market upside exceeding the implied average cost of capital), a greater number of stocks would fall into the aforementioned Outperform (Buy) category, illustrating the significance of the “relative return” concept (vis-à-vis absolute return) in picking better investment ideas with a positive alpha. The same holds true when the potential upside of an average stock in our coverage falls short of its required rate of return. In this regard, as a supplemental methodology, we rank the stocks in our coverage according to their notional target price with respect to their current market price, and then categorise the top group (approximately 40-50% of the companies under coverage) as Outperform, the next 40-50% as Neutral and the lowest 10-20% (and no less than 10%) as Underperform. It should be noted that the expected returns on some stocks may at times fall outside the relevant ranges of the applicable respective rating category because of market price movements and/or other short-term volatility or trading patterns. Such interim deviations from specified ranges are permitted but becomes subject to review. Also note that the analyst’s short-term view may occasionally diverge from the stock’s longer-term fundamental rating. Outperform. We expect the stock to outperform the BIST-100 over the next 6 to 12 months. Neutral (Market Perform). We expect the stock to broadly perform in line with the BIST-100 index over the next 6 to 12 months. (Although we would normally have a neutral assessment of stocks in this category, if a stock has gone through a period of market underperformance, it would be an indication that the stock may be expected to improve its performance relative to market averages in the coming period, and vice versa). Underperform. We expect the stock to underperform the BIST-100 over the next 6 to 12 months. N/R. Not Rated. U/R. Under Review.

Analyst Certification The following analysts hereby certify that the views expressed in this research report accurately reflect their own personal views regarding the securities and issuers referred to therein and that no part of their compensation was, is, or will be directly or indirectly related to the specific recommendations or views contained in the research report: Ayhan Yüksel, CFA Unless otherwise stated, the individuals listed on the cover page of this report are research analysts.

Disclaimer All information enclosed in this document has been obtained from sources believed to be reliable. While FINANSINVEST has spent reasonable care in verifying the accuracy and completeness of the information presented herein, it cannot be held responsible for any errors, omissions or for consequences arising from the use of such information. This document is published for purposes of providing information to investors who are expected to make their own investment decisions without undue reliance on this report. Therefore, no article or statement can be construed as an investment advice or a solicitation to buy or sell the securities mentioned herein. The affiliates, officers, partners and employees, including persons involved in the preparation or issuance of this material may have a direct or indirect position in any security mentioned in this report.

Disclosure For U.S. persons only: This research report is a product of Finans Yatirim Menkul Degerler A.S. (“FINANSINVEST”), which is the employer of the research analyst(s) who has prepared the research report. The research analyst(s) preparing the research report is/are resident outside the United States (U.S.) and are not associated persons of any U.S. regulated broker-dealer and therefore the analyst(s) is/are not subject to supervision by a U.S. broker-dealer, and is/are not required to satisfy the regulatory licensing requirements of FINRA or required to otherwise comply with U.S. rules or regulations regarding, among other things, communications with a subject company, public appearances and trading securities held by a research analyst account. This report is intended for distribution by FINANSINVEST only to "Major Institutional Investors" as defined by Rule 15a-6(b)(4) of the U.S. Securities and Exchange Act, 1934 (the Exchange Act) and interpretations thereof by U.S. Securities and Exchange Commission (SEC) in reliance on Rule 15a 6(a)(2). If the recipient of this report is not a Major Institutional Investor as specified above, then it should not act upon this report and return the same to the sender. Further, this report may not be copied, duplicated and/or transmitted onward to any U.S. person, which is not the Major Institutional Investor. In reliance on the exemption from registration provided by Rule 15a-6 of the Exchange Act and interpretations thereof by the SEC in order to conduct certain business with Major Institutional Investors, FINANSINVEST has entered into an agreement with a U.S. registered broker-dealer, Marco Polo Securities Inc. ("Marco Polo"). Transactions in securities discussed in this research report should be effected through Marco Polo or another U.S. registered broker dealer.

TOP DOWN STOCK SELECTION Incorporating macro views in ...

Documents