A study of the relation between U.S. macroeconomic announcements and the turn-of-the-month effect in Latin American countries

(1)

Bachelor thesis Mark Goes 10557636

P. 1

A study of the relation between U.S. macroeconomic

announcements and the turn-of-the-month effect in Latin

American countries.

Abstract

This study tests the macroeconomic announcement hypothesis for the turn-of-the-month effect in Latin American countries. The turn-of-the-month effect has been tested and found for Brazil, Mexico and Chile. The U.S. macroeconomic announcements have an effect on the turn-of-the-month effect in Brazil, Mexico and Chile. However, the hypothesis that U.S. macroeconomic announcements are the only explanation for the turn-of-the-month effect is rejected. Furthermore, results show that the explanatory value of U.S. macroeconomic announcements are stronger on markets that are more integrated with the world market.

Name: Mark Goes

Supervisor: Rob Sperna Weiland

Date of submission: June 28, 2016

Specialization: Economics and Finance

Field: Finance

(2)

P. 2

Statement of Originality

This document is written by Mark Goes who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

P. 3

1. Introduction

The total market capitalization of the stock market for every country in the world is 35.978.198,11 million US dollars according to the MSCI all country world index (MSCI ACWI (USD), 2016). Because the size of the stock market and the frequency of trading, it is reasonable to expect that there are few or none anomalies. These are important characteristics of an efficient market which should not have any anomalies since anomalies would cause a ‘free lunch’, this ‘free lunch’enables investors to make money without increasing their risk (Bodie, Kane & Marcus, 2011).

However several calendar anomalies have been found over the years. Some are the day-of-the-week effect (Basher & Sadorsky, 2006), turn-of-the-month effect (Agrawal & Tandon, 1994; Kunkel, Compton & Beyer, 2003; Lakonishok & Smidt, 1988; Martikainen, Perttunen & Ziemba, 2012), turn-of-the-year effect (Ritter & Chopra, 1989; Ziemba, 1991) and the ‘holiday effect’ (Agrawal & Tandon, 1994; Ziemba, 1991). Since the discovery of these anomalies most of them disappeared or are reduced in their form. The turn-of-the-month effect however, still exists (Martikainen, Perttunen & Ziemba, 2013; Marquering, Nisser & Valla, 2006). Marquering, Nisser and Valla (2006) stated that this might be the case because of the transaction costs needed to gain from the turn-of-the-month effect. On the other hand, different studies did proof that the turn-of-the-month effect is exploitable (Hensel & Ziemba, 1996; Kunkel & Compton, 1998; Zwergel, 2010).

This paper will investigate the turn-of-the-month effect. Several explanations of the turn-of-the-month effect have been found. Like the ‘liquidity effect’, which states that the turn-of-the-month effect is due to increased liquidity during this period (Booth, Kallunki & Martikainen, 2001; Ogden, 1990). A possible explanation for increased liquidity is the ‘preferred habitat’ hypothesis which states that the liquidity effect is caused by the time the salaries are paid (Ogden, 1990; Wiley & Zumpano, 2009). Furthermore, the ‘window dressing’ hypothesis might explain this liquidity effect since it states that during the turn-of-the-month fund managers and other institutional investors might sell stocks that underperform and purchase those that have recently performed well (Lakonishok & Smidt, 1998; Wiley & Zumpano, 2009). These theories exist for quite some time and have been researched thoroughly. However, a recent line of studies investigated the relationship between macroeconomic news announcements and the turn-of-the-month effect (Nikkinen, Sahlström & Äijö, 2007b). This effect has found to be significant in the U.S., U.K, Germany and France (Nikkinen, Sahlström & Äijö, 2007a) which have a highly integrated stock market (Baekart & Harvey, 1995).

However, Latin America has a more segmented market (Nikkinen, Omran & Sahlström, 2006) where there are more ambiguous results for the turn-of-the month effect (Agrawal & Tandon, 1994; Batten & Szilagyi, 2011; Giovanis, 2009; Kunkel, Compton & Beyer, 2003; McConnell & Wu, 2008). So despite that the share of the emerging markets in Latin America is approximately 1.42% from the world’s total stock market, it is an interesting market to see if scheduled macroeconomic announcements are indeed a reason for the turn-of-the-month effect (MSCI Emerging Markets Latin America Index (USD), 2016). In this study the following questions will be answered, does the turn-of-the-month effect apply for Latin American countries? Is the turn-of-the-month effect in Latin America related to scheduled macroeconomic announcements from the U.S.? To answer these questions, Brazil, Chile and Mexico are investigated since they account for 93.19% of the total market capitalization of the Latin American countries according to the MSCI (2016).

The results of the study indicate that the U.S. macroeconomic announcements have an effect on the turn-of-the-month effect in Brazil, Mexico and Chile, but the hypothesis that it is the only explanation is rejected. Furthermore, results show that the explanatory value of U.S.

(5)

P. 5

macroeconomic announcements are stronger on markets that are more integrated with the world market. Moreover, during the period of 2008-2014 the turn-of-the-month effect has disappeared in the stock markets of Brazil, Chile and Mexico. In the other periods tested the turn-of-the-month effect does exists.

The remainder of this paper contains the following, section 2 contains the literature review which contains the existing literature on the subject. Section 3 discusses the data used in this paper and section 4 describes the methodology used in the paper. In section 5 the results are provided and section 6 concludes and summarizes the paper.

2. Literature review

2.1 Existence of turn-of-the-month effect and definition

Since the first paper of Ariel (1987), a considerable amount of papers have been published on the turn-of-the month effect and possible explanations of this anomaly. While the work of Ariel (1987) investigated the difference in returns during the first half of the month and the second half of the month, following work narrowed this period down until the turn-of-the month effect that will be discussed in this paper (Agrawal & Tandon, 1994; Kunkel, Compton & Beyer, 2003; Lakonishok & Smidt, 1988; Martikainen, Perttunen & Ziemba, 2012).

The turn-of-the-month effect is defined in most articles as a period of returns during the trading day before the turn of the month until three trading days after the turn-of-the-month where the returns are significant higher compared to the rest of the turn-of-the-month (McConnell & Xu, 2008; Ogden, 1990). Kunkel, Compton and Beyer (2003) even found that a large part of the total return of a stock in a month is caused by the turn-of-the-month effect. However, some articles change this period. Ziemba (1991) found a turn-of-the-month effect during the five trading days before the turn of the month and two trading days after the turn of the month. Wiley and Zumpano (2009) found the turn-of-the-month effect during the four trading days before the turn of the month and the day after the turn of the month. Nikkinen et al. (2006) found the turn-of-the-month effect only the three trading days after the turn of the month. Maher and Pakir (2013) found the turn-of-the-month effect for the period of one trading day before the turn of the month and two trading days after the turn of the month. Therefore, the definition of the turn-of-the-month period in this article is a period around the turn of the month where the returns are significantly higher compared to the rest of the month. Over the years, a lot of articles were written to prove the turn-of-the-month effect is robust. They did so by proving that the effect is significant in many different regions where the turn-of-the-month effect has been found. The turn-of-the-month has been studied for Finland (Booth, Kallunki & Martikainen, 2001), Japan (Ziemba, 1991), India (Maher & Parikh, 2013), early studies which tested the turn-of-the-month effect from regions all over the world (Agrawal & Tandon, 1994; Kunkel & Compton, 2003; Lakonishok & Smidt, 1988) and more recent studies which tested the turn-of-the-month effect from regions all over the world (Martikainen, Perttunen & Ziemba, 2012; Wiley & Zumpano, 2012). These studies have different sample periods, in the case of Lakonishok and Smidt (1988) a sample period of 90-years was used, while McConnel and Xu (2008) and Wiley and Zumpano (2009) studied a more recent period. These studies are done across many markets and have methodologies which are robust to traditional assumptions, which makes the turn-of-the-month effect as a result of data mining unlikely (Wiley & Zumpano, 2009).

However, as well as in the early studies which tested the turn-of-the-month effect from regions all over the world as in the recent ones, the studies mainly focussed on the US, Europe and Asia. The reason for the absence of Latin-American countries and African counties in these studies can be explained by the weight that these regions have on world market

(6)

P. 6

capitalization. The weight of Latin-America and Africa is below 2% according to the MSCI (MSCI Emerging Frontier Markets Africa Index (USD), 2016; MSCI Emerging Markets Latin America Index (USD), 2016). To be relevant for research, these markets must have very specific characteristics. Furthermore, reliable information in Latin America and Africa is more difficult to obtain, especially more historical information and information for more corrupt regions.

During the turn of the month a large part of the total return of the month is generated. Kunkel, Compton and Beyer (2003) found in their research that the turn-of-the-month period accounted for 87% of the total monthly return. However, Ogden (1990) states that it might be difficult to exploit the turn-of-the-month effect since the magnitude of the transaction costs makes it hard to gain from this anomaly. He suggested that the turn-of-the-month effect might not disappear because the difficulty of exploiting this anomaly. However, Later studies show however, that it is not impossible to gain from the turn-of-the-month effect. Hensel and Ziemba (1996) showed that institutional investors can exploit the this anomaly by using a switching strategy, where you only invest in an equity portfolio during the turn-of-the-month period. After this period they should switch to an interest bearing cash account for the remainder of the month. They found that this strategy lowered the volatility of the investment and improved the expected returns. Kunkel and Compton (1998) tested the potential effectiveness of a comparable strategy for individual investors by using commingled real estate funds to avoid paying taxes. They found that this is indeed a way to profit from the turn-of-the-month effect. In a more recent study, Zwergel (2010) tested the turn-of-the-month effect on the futures markets. He found that the turn-of-the month effect is still exploitable even after deducting for transaction costs and slippage. Zwergel’s (2010) research is especially interesting since he did deduct for transaction costs which implies that this might not be the reason for the existence of the turn-of-the-month effect.

2.2 Possible explanations for the turn-of-the-month effect

Over the years, different explanations of the turn-of-the-month effect have been conceived and tested. A possible explanation for the turn-of-the-month effect is the ‘news clustering hypothesis’ (Ariel, 1987; Kayacetin & Lekpek, 2016). This hypothesis states that the dissemination of good and bad news is the reason for the excessive returns during the turn-of-the-month. Findings of McNichols (1988) and Penman (1987) support this hypothesis because their paper shows that firms tend to announce good news early in the month. While the announcement of bad news is delayed until the reporting deadlines. A note on this theory is that an empirical relationship between this hypothesis and the turn-of-the-month effect has not been established.

Lakonishok and Smidt (1998) gave the ‘window dressing’ theory as a possible explanation for the turn-of-the-month effect. This states that pension fund managers and other institutional investors look at their stock performance at the end of the month and sell stocks that underperformed and buy stocks that performed well. However, Wiley and Zumpano (2009) pointed out that these disclosure periods, or the period of opportunity for managers to ‘window dress’, may not perfectly coincide with the monthly calendar effect. This is in line with the findings of Lakonishok et al. (1991) who found that this kind of behavior for most of the market happened on a quarterly basis.

Wiley and Zumpano (2009) empirically tested the relationship between institutional investors and the turn-of-the-month effect. They found that institutional investors are not the reason for the turn-of-the-month effect. They found that institutional investors had a reducing effect of the impact that the individual investors have on the turn-of-the-month effect. They

(7)

P. 7

suggested that this might cause the institutional investors to mitigate the turn-of-the-month effect.

Another possible explanation is the ‘preferred habitat’ theory. This theory implies that the turn-of-the-month is a preferred habitat to invest. Ogden (1990) states that ‘for major economic entities, the turn of each calendar month is a typical payoff date for accrued real wages, dividends, interest, principal payments, and other liabilities’. Because the cash flow during this period is higher than during the rest of the month, investors have more money to invest into the stock market. This might be the reason for the increased trading volume during the turn of the month found by Booth, Kallunki and Martikainen (2001). Therefore, it could be the reason for the turn-of-the-month effect. In the same paper, Ogden (1990) finds that the ‘preferred habitat’ theory is actually a reason for the turn-of-the-month effect. He was the first one who was able to obtain an empirical relationship between this hypothesis and the turn-of-the-month effect. Furthermore, Ziemba (1991) provides a paper which is in line with the ‘preferred habitat’ theory. In Japan most salaries are paid at least five days before the end of the month and the excessive results which are usually found during the last trading day of the month, start 5 trading days before the turn of the month in Japan. Which implies that investors tend to reinvest their income quickly. Furthermore, bonuses are paid in June in Japan and Ziemba (1991) found a June effect in Japan. This indicates that the moment when people get paid and calendar anomalies are related which is in line with the ‘preferred habitat’ theory.

Booth, Kallunki and Martikainen (2001) tested a direct relation between liquidity and the turn-of-the-month effect and found its existence in Finland. The finding of Booth, Kallunki and Martikainen (2001) is vital for the ‘preferred habitat’ and the ‘window dressing’ theory since both these theories are built on the tendency that increased liquidity moves together with increased returns.

Zhao, Liano and Hardin (2004) investigated the influence of the presidential election cycle in the U.S. and found a relation with the turn-of-the-month effect. They found evidence of higher turn-of-the-month returns in the second half of the presidential terms. These higher returns were linked to fiscal and administrative policies that increased household liquidity prior to elections, which is also in line with the earlier findings of Ziemba (1991) in Japan and strengthens the ‘preferred habitat’ theory.

However, a paper of Maher and Parikh (2013) did not find any support for the ‘preferred habitat’ theory due to erratic salary payments through the month but did find a turn-of-the-month effect in India. However, they did find a turn-turn-of-the-month effect. This does not mean that the ‘preferred habitat’ theory has no influence on the turn-of-the-month effect. However, it does show that it is likely that there is also another explanation for the turn-of-the-month effect.

2.3 Macroeconomic announcements of U.S.

Graham, Nikkinen and Sahlström (2003) investigated whether scheduled macroeconomic news announcements have significant influence on the stock valuation. They found that the Employment Report, NAPM (manufacturing), Producer Price Index, Import and Export Price Indices, and Employment Cost Index have a significant influence. Furthermore, they found that the announcements that exert the greatest influence are the Employment Report and the NAPM (manufacturing). However, since 2003 several things have changed and according to the Bureau of Labour and Statistiscs (Bureau of Labour and Statistics, 2016) Consumer Price Index, Employment Cost Index, Employment Situation, Producer Price Index, Productivity and Costs, Real Earnings and U.S. Import and Export Price Indices are the most important macroeconomic announcements.

(8)

P. 8

Nikkinen, Sahlström and Äijö (2007b) were the first researchers to link the turn-of-the-month effect to scheduled macroeconomic announcements of the U.S.. The reasoning behind this hypothesis was that the timing of the scheduled macroeconomic releases is well known in advance. Thereby affecting investors’ expected risks and hence expected returns, as well as realized volatilities and returns (Jones, Lamont & Lumsdaine, 1998). Furthermore, and most significantly, important macroeconomic news announcements are systematically clustered on particular days of each month, mostly in the first half of the month. In addition, previous literature shows that the macroeconomic news announcements released at the beginning of the month are the most important news announcements since they have the highest information content for investors (Bollerslev, Cai & Song, 2000; Graham, Nikkinen & Sahlström, 2003). Furthermore, an increase in trading activity around these important announcements has been found as investors trade according to their opinions before and after these announcements (Chordia, Roll & Subrahmanyam, 2001; Fleming & Remolona, 1999; Nofsinger & Prucyk, 2003) which increases the liquidity during this period. Taking into account that Karpoff (1987) states that the increase in liquidity is positively associated with price changes and that this relationship is largely driven by the arrival of information. The information arrival is generally integrated in the market within one minute after an announcement (Balduzzi, Elton & Green, 2001). Because of this fast integration Nikkinen, Sahlström and Äijö (2007b) concluded that linking the turn-of-the-month effect to macroeconomic news announcements is consistent with earlier empirical findings on the increased trading activity at the turn-of-the-month (Booth, Kallunki & Martikainen, 2001).

After Nikkinen, Sahlström and Äijö (2007b) accounted for the scheduled macroeconomic news announcements the returns where no longer statistically significant during the turn-of-the-month period. This indicates that the clusterization of important macroeconomic announcements of the U.S. explains the intramonth return patterns (Nikkinen, Sahlström & Äijö, 2007b). Gerlach (2007) also found that important macroeconomic announcements influence the turn-of-the-month effect. Gerlach (2007) made a distinction between turn-of-the-month days with and without macroeconomic announcements and finds that on turn-of-the-month days with macroeconomic announcements the returns a were significantly higher. This is in line with the paper of Savor and Wilson (2013) which stated that 60% of the total returns in a month were on days of scheduled macroeconomic announcements.

Nikkinen et al. (2006) found that G7 countries, other European countries than the G7 countries, developed Asian countries and emerging Asian countries are closely integrated with respect to scheduled U.S. macroeconomic news announcements. Latin American countries however, are more segmented. After these findings Nikkinen, Sahlström and Äijö (2007a) tested if scheduled macroeconomic announcements of the U.S. applied for other regions like the U.K., France, and Germany and found a significant result. Nikkinen et al. (2009) also found for the much smaller market of Finland that their hypothesis did still hold which made the findings more robust. A note on the work of Nikkinen, Sahlström and Äijö (2007a; 2007b) is that they have also regressed macroeconomic news announcements that are not confirmed by Graham, Nikkinen and Sahlström (2003). This might have influenced the outcome.

2.4 The aim of the research

All of the existing evidence supporting the theory of U.S. macroeconomic news announcements is from markets which are highly integrated (Nikkinen et al., 2006). However, since integrated markets tend to move together, the results found so far by Nikkinen et al. (2006) and Nikkinen et al. (2009) can be challenged by testing their hypothesis for a more

(9)

P. 9

segmented market. Segmented markets are defined to not or barely react to information from the world market. Consequently, it is reasonable to assume that these markets do not tend to move with the world market. Testing this hypothesis for Latin American countries can therefore be interesting since these countries are segmented (Nikkinen, Omran & Sahlström, 2006) from the world market, which might lead to different results. A note on this theory is that the segmentation of a market is not unalterably, Bekaert and Harvey (1995) found that the country specific integration in relation to the world market is time-varying. This makes it possible that over time the markets of Latin America became more or less segmented from the world market.

McConnel and Xu (2008) found a turn-of-the-month effect in Mexico and Chile after the findings of Nikkinen, Omran and Sahlström (2006) that Latin America was segmented from the rest of the world market. The data investigated in these papers were roughly from the same period which makes the finding of the turn-of-the-month effect odd. If a segmented market did not or barely react to the information from the world market, it is unlikely that the macroeconomic news announcements from the U.S. were the reason for the turn-of-the-month effect in Latin American countries. This might indicate that the macroeconomic announcements from the U.S. were not the reason for the turn-of-the-month effect in the period measured by McConnel and Xu (2008).

Domestic macroeconomic releases might be the reason for this phenomenon. However, the reasoning of Nikinnen, Sahlström and Äijö (2007b) that macroeconomic announcements are the reason for the turn-of-the-month effect is largely based on the fact that U.S. macroeconomic announcements are very reliable and well known in advance. These characteristics are important because people make expectations for the macroeconomic announcements and trade according to their opinions before and after the announcements (Chordia, Roll & Subrahmanyam, 2001; Fleming & Remolona, 1999; Nofsinger & Prycyk, 2003). However, Veronesi (2000) showed in his paper that the quality of public information arrival is essential for the effect it has on the market. They found that the reliability of public information is important no matter the state of the market. Brazil, Chile and Mexico all experienced different political scandals and the reliability of the information given by these governments is questionable.

Furthermore, important macroeconomic news announcements of the U.S. are systematically clustered on particular days of each month, mostly in the first half of the month. However, the release date of important economic indicators of Chile (Economic Indicators, n.d.), Mexico (Estadícas Económica, 2016) and Brazil (Short-term indicators in 2016, 2016) are not released around the same dates as the important U.S. economic indicators. Additionally, the release dates of important economic indicators of Brazil, Chile and Mexico are not clustered during the first half of the month. This can be explained by the fact that these countries have different economic calendars. Consequently, domestic macroeconomic releases in these countries are not likely to be the reason for the turn-of-the-month effect.

Furthermore, Gay (2011) tried to link macroeconomic variables like the exchange rate and oil price for Brazil to the stock price and found no significant relationship. Muradoglu et al. (2000) found in their study that the relationship between macroeconomic variables on the stock market is dependent of the relative size of the stock market and their integration with world markets. The Latin American countries account for a small part of the world stock market capitalization (MSCI ACWI (USD), 2016). Consequently, the relative size of the stock market of Latin American countries is small. Furthermore, the Latin American markets can be seen as segmented from world markets (Nikkinen, Omran & Sahlström, 2006). Therefore, the relationship between macroeconomic variables and the Latin American countries is small, according to the findings of Muradoglu et al. (2000). Consequently, it is questionable that U.S. macroeconomic announcements have a large impact on the stock market returns in Latin

(10)

P. 10

American countries. This indicates that U.S. macroeconomic announcements by itself are unlikely to be the main reason for the turn-of-the-month effect in Latin American countries. This is confirmed by Verma and Ozuna (2005) who found that stock market movements in Latin America cannot be attributed to movements in cross-country macroeconomic variables.

There is strong evidence for the turn-of-the-month effect in integrated markets (Agrawal & Tandon, 1994; Kunkel, Compton & Beyer, 2003; Lakonishok & Smidt, 1988; Martikainen, Perttunen & Ziemba, 2012). Ambiguous results have been found for the turn-of-the-month effect in Latin American countries (Agrawal & Tandon, 1994; Batten & Szilagyi, 2011; Giovanis, 2009; Kunkel, Compton & Beyer, 2003; McConnell & Wu, 2008). A possible explanation for this phenomenon is that the turn-of-the-month effect is for a large part caused by scheduled macroeconomic announcements from the U.S. and when the market becomes more segmented, the turn-of-the-month effect disappears. Kunkel, Compton and Beyer (2003) tested the turn-of-the-month effect for Brazil during 1988-2000 and they did not find a turn-of-the-month effect for Brazil. However, Agrawal and Tandon (1994) tested the the-month effect for Mexico and Brazil during 1971-1987 and they did find a month effect for Mexico and Brazil. McConnell and Xu (2008) tested the turn-of-the-month effect for Mexico and Chile during 1983-2006 and they found a turn-of-the-turn-of-the-month effect for Mexico and Chile. This could be the result of time-varying integration of the Latin American countries which would be in line with the findings of Bekaert and Harvey (1995).

3. Data

We focus on Brazil, Chile and Mexico since they account for 93.19% of the market capitalization in Latin America according to the MSCI. The BOVESPA, IPSA and IPC are used as proxies for the general market indices. The sample period measured is 2000-2014 for Brazil and Mexico and 2002-2014 for Chile. Bekaert and Harvey (1995) have found that the country specific integration in relation to the world market is time-varying and to take this into account the period measures is divided into two periods. Consequently, the periods tested for Brazil and Mexico are January 1, 2000 until December 31, 2007, January 1, 2008 until December 31, 2014 and January 1 2000 until December 31, 2014. Due to data limitations the periods measured for Chile are January 1, 2002 until December 31, 2007, January 1, 2008 until December 31, 2014 and January 1 2002 until December 31, 2014. Daily returns are calculated as percentage changes based on the closing values. Since the returns are calculated on the closing values, no additional changes have to be made with the data.

The study uses a sample of scheduled macroeconomic news announcements from the U.S.. The macroeconomic announcements tested in this paper are based on the findings of Graham, Nikkinen and Sahlström (2003) and the important economic indicators according to the BLS. Therefore, the macroeconomic announcements tested in this paper are consumer price index (CPI), employment cost indexes (ECI), employment situation (ES), gross domestic Product (GDP), import and export price indices (IEPI), NAPM nonproduction (NONNAMP), NAPM production (NAPM), producer price index (PPI), productivity and cost preliminary (PACP), productivity and cost revised (PACR) and real earnings (RE). The selected macroeconomic announcements and more information are reported in table 1. Since the economic calendars of the U.S. and Brazil, Chile and Mexico do not move parallel, some release dates are omitted from this research. The rule used for this is that when there was no trading day during the release day the variable was omitted. The reasoning behind this rule is that the information arrival is generally integrated in the market within one minute after an announcement (Balduzzi, Elton & Green, 2001). Still adding the release dates one trading day later would therefore lead to biased results.

(11)

P. 11 Table 1

Macroeconomic news announcements investigated in the study

Macroeconomic announcement Symbol Issued

Consumer price index CPI Monthly

Employment cost indexes ECI Quarterly

Employment situation ES Monthly

Gross domestic product GDP Quarterly* Import and export price indices IEPI Monthly NAPM (nonproduction) NONNAPM Monthly

NAPM (production) NAPM Monthly

Producer Price Index PPI Monthly

Productivity and cost (preliminary) PACP Quarterly Productivity and cost (revised) PACR Quarterly

Real earnings RE Monthly

Source: Bureau of labour statistics

* Note: the Gross domestic product is revised monthly

4. Methodology

To begin, the analysis was started by determining whether the turn-of-the-month effect exists for Brazil, Chile and Mexico. To test the turn-of-the-month we use a regression model which controls for autocorrelation, heteroscedasticity and other calendar effects, namely day-of-the-week and turn-of-the-year effect (Nikkinen, Sahlströhm & Äijö, 2007b; Szakmary & Kiefer, 2004). Most of these anomalies disappeared, but some might be still there in a reduced form so we have to account for these possible anomalies (Marquering, Nisser & Valla, 2006). This leaves us with the following regression model:

𝑟𝑡 = 𝑎1𝑀𝑜𝑛𝑑𝑎𝑦 + 𝑎2𝑇𝑢𝑒𝑠𝑑𝑎𝑦 + 𝑎3𝑊𝑒𝑑𝑛𝑒𝑠𝑑𝑎𝑦 + 𝑎4𝑇ℎ𝑢𝑟𝑠𝑑𝑎𝑦 + 𝑎5𝐹𝑟𝑖𝑑𝑎𝑦 +

𝑎6 𝑇𝑌 + ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 (1a)

Where 𝑟_𝑡 is the return for the index tested on day 𝑡; Monday, Tuesday, Wednesday, Thursday and Friday are dummy variables representing the trading days of the week; TY is a dummy variable for the turn of the year which is 1 the first trading day after the turn of the year and zero otherwise; 𝑖 refers to trading day of the month (-8, -7, …, +7, +8), 𝐷_𝑖,𝑡 is a dummy variable having the value of 1 on day 𝑖 and zero otherwise, 𝑅𝑂𝑀𝑡 is a dummy variable that

takes the value 1 on rest-of-the-month days (namely, other than -8,-7, …, +7, +8). Intercept terms have been omitted to avoid dummy variable traps in the regression. Regression (1a) directly addresses possible interaction between the turn-of-the-month effect and other calendar effects but lead in some cases to multicollinearity. Since no significant influence on the day-of-the-week and turn-of-the-year was found during the regressions, the following regression model is a good model to determine the turn-of-the-month effect.

𝑟𝑡 = ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 (1b)

To examine the impact of macroeconomic news announcements on the IBOVESPA, IPSA and IPC indexes, these returns are regressed on a set of dummies representing the chosen macroeconomic news announcements. This is necessary to which macroeconomic news announcement affect the chosen indexes returns so we can determine which announcements causes the turn-of-the-month effect. Which leaves us with the following regression:

(12)

P. 12

𝑟𝑡 = 𝑐 + ∑11𝑚=1𝛼𝑚𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆𝑚,𝑡+ 𝜀𝑡 (2)

Where 𝑟_𝑡 is the return for the chosen indexes on day t, 𝑐 is the intercept term, 𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆_𝑚,𝑡 is defined as a dummy variable for the macroeconomic news announcement 𝑚 = (𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 𝑅𝑒𝑝𝑜𝑟𝑡1, 𝑁𝐴𝑃𝑀 (𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑖𝑛𝑔)2, … , 𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡

𝑐𝑜𝑠𝑡 𝑖𝑛𝑑𝑒𝑥6), that takes a value of 1, if news 𝑚 occurs on day 𝑡, otherwise zero. This will be

done for the domestic and the U.S. macroeconomic news announcements.

To analyze the impact of macroeconomic news on the turn-of-the-month, the residuals estimated in regression (2) will be used. These residuals can be considered as the portion of the IBOVESPA, IPSA and IPC returns that are orthogonal to the risk premiums related to the macroeconomic news announcements. Which makes it possible to investigate the effect of a scheduled macroeconomic news announcement on the turn-of-the-month effect. Which leaves us with the following regression model:

𝑟𝑒𝑠𝑖𝑑_𝑡= ∑8_𝑖=−8𝛼_𝑖𝐷_𝑖,𝑡+ 𝛼₀𝑅𝑂𝑀_𝑡+ 𝜀_𝑡 (3)

Where 𝑟𝑒𝑠𝑖𝑑_𝑡, refers to the obtained residuals from regression (2) and the other variables are as earlier defined. If the macroeconomic news announcements explain the effects, the coefficients for the dummy variables in these models should not differ from zero.

The Durbin-Watson autocorrelation test is used to reveal possible autocorrelation in the residuals and have not been found in any period tested. The regressions are made with robust standard errors so heteroscedasticity is accounted for.

5. Results

5.1 Turn-of-the-month effect

Before the effect of macroeconomic news announcements on the turn-of-the-month effect can be tested, the presence of the turn-of-the-month effect in the daily indexes measured must be provided. The data from table 2 shows the results for Brazil and Mexico for the period 2000-2014 and for Chile for the period 2002-2000-2014. The results show that the turn-of-the-month effect exists for Brazil since the first and the last trading day of the month are significant higher at a 1% significance level. Furthermore, the data from table 2 shows that the turn-of-the-month effect exists from Chile since the last two trading days of the month are significant higher at a 1% and 5% significance level. As well as for Brazil and Chile, table 2 shows that the turn-of-the-month effect exists in Mexico since the last and the second trading day of the month are significantly higher at a 5% level and the first trading day of the month at a 1% level. Therefore, we can conclude that during 2000-2014 the turn-of-the-month effect exists in Brazil, Chile and Mexico. Consequently, measuring the effect of macroeconomic announcements from the U.S. in these countries during this period is relevant since there is a turn-of-the-month effect to explain.

(13)

P. 13 Table 2

Returns Brazil, Chile**** and Mexico in period 2000-2014

Return Brazil Chile Mexico

-1 0.003234* (0.001144) 0.003494* (0.000718) 0.001844** (0.000824) -2 0.001087 (0.00158) 0.002092** (0.000859) 0.000861 (0.001073) -3 0.002336*** (0.001358) 0.000984 (0.000795) 0.001912** (0.000965) -4 0.001432 (0.001450) -0.00068 (0.000643) 0.002033*** (0.001076) -5 0.001146 (0.001373) 0.000001 (0.000663) 0.000738 (0.000938) -6 -0.00156 (0.001300) 0.000612 (0.000804) 0.000415 (0.000833) -7 -0.001900 (0.001472) -0.000980 (0.000884) -0.000400 (0.001069) -8 -0.00119 (0.001647) 0.000394 (0.000791) -0.000680 (0.001114) ROM 0.000615 (0.000648) 0.000919** (0.000378) 0.000751 (0.000456) 1 2 3 4 5 6 7 8 0.00539* (0.001532) 0.001752 (0.001477) -0.00071 (0.001281) -0.00255*** (0.001463) 0.000706 (0.001208) -0.000870 (0.001444) -0.00089 (0.001393) 0.000181 (0.001317) 0.001671*** (0.000880) -0.000730 (0.000771) -0.000560 (0.000817) -0.000400 (0.000901) -0.000690 (0.000824) -0.000460 0.000940 -0.000390 0.000870 0.000078 (0.000846) 0.005247* (0.001148) 0.002043** (0.000988) -0.000560 (0.000978) -0.001150 (0.001003) -0.000340 (0.000827) -0.000230 (0.000891) -0.000710 (0.001037) -0.000740 (0.001016)

Note: regression formula has the following form: 𝑟𝑡= ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 Where 𝑟𝑡 is the return for the index tested on day 𝑡; 𝑖 refers to trading day of the month (-8, -7, …, +7, +8), 𝐷𝑖,𝑡 is a dummy variable having the value of 1 on day 𝑖 and zero otherwise, 𝑅𝑂𝑀𝑡 is a dummy variable that takes the value 1 on rest-of-the-month days (namely, other than -8,-7, …, +7, +8). Intercept terms have been omitted to avoid dummy variable traps in the regression. Estimates that are significant at a 5% level are in bold faces. * coefficient is significant at 1% level

** coefficient is significant at 5% level *** coefficient is significant at 10% level **** Period measured for Chile is 2002-2014

Because Bekaert and Harvey (1995) found that the country specific integration in relation to the world market is time-varying. It is relevant to look if the turn-of-the-month effect also can be found during the time periods 2000-2007 for Brazil and Mexico, 2002-2007 for Chile and 2008-2014 for Brazil, Chile and Mexico. If there is no turn-of-the-month effect found during these periods, the is no use to look for the effect of U.S. macroeconomic news release simply because there is no anomaly to explain.

The results for period 2000-2007 for Brazil and Mexico and period 2002-2007 for Chile are in line with the findings for period 2000-2014 for Brazil and Mexico and for period 2002-2014 for Chile. However, table 3 shows that for period 2008-2014 no turn-of-the-month effect has been found for Brazil, Chile and Mexico. Since there was no significant effect

(14)

P. 14

found during the turn of the month period in Brazil and Mexico. Furthermore, the returns for Chile show only during the last trading day of the month returns that are significantly higher. The results found for all periods measured are in line with the earlier found ambiguous results in Latin-American countries (Agrawal & Tandon, 1994; Batten & Szilagyi, 2011; Giovanis, 2009; Kunkel, Compton & Beyer, 2003; McConnell & Wu, 2008). No other studies have measured the same period yet, so there results cannot be generalized or linked to other regions. The disappearance of the turn-of-the-month effect cannot be explained by segmentation of the Latin-American countries from the world market which is suggested before. The reason for this is that Latin-American countries are segmented from the world market, but become more correlated with the world market during the crisis in 2008 (El Hedi Arouri, Bellalah and Nguyen, 2010).

Table 3

Returns Brazil, Chile and Mexico in period 2008-2014

-1 0.001971 (0.001840) 0.003097* (0.001137) 0.001133 (0.001308) -2 -0.000230 (0.002213) 0.001895 (0.001186) 0.001040 (0.001500) -3 0.002715 (0.001918) 0.001043 90.001121) 0.002902 (0.001334) -4 0.001716 (0.002274) -0.001340 (0.000867) 0.002171 (0.001748) -5 0.000289 (0.001991) -0.000470 (0.000871) 0.000532 (0.001405) -6 -0.00318*** (0.001867) 0.000133 (0.001167) 0.000251 (0.001090) -7 -0.002150 (0.002022) -0.002520*** (0.001377) -0.001840 (0.001684) -8 -0.002000 (0.002465) 0.001433 (0.001058) 0.000449 (0.001681) ROM 0.000741 (0.000952) 0.000941*** (0.000569) 0.000092 (0.000685) 1 2 3 4 5 6 7 8 0.003931*** (0.002368) -0.000620 (0.002071) -0.000720 0.001794 -0.002550 (0.002164) 0.000835 (0.001616) -0.002090 (0.002143) 0.000005 (0.001928) -0.000740 (0.001822) 0.000957 (0.001380) -0.001830 (0.001162) -0.000460 (0.001225) -0.000920 (0.001428) -0.000970 (0.001181) -0.000710 (0.001417) 0.000731 (0.001332) -0.000460 (0.001266) 0.003087 (0.001801) 0.000687 (0.001427) -0.000770 (0.001406) -0.002130 (0.001410) -0.000980 (0.001095) 0.000449 (0.001418) 0.000908 (0.001481) -0.001940 (0.001253)

Note: regression formula has the following form: 𝑟𝑡= ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 Where 𝑟𝑡 is the return for the index tested on day 𝑡; 𝑖 refers to trading day of the month (-8, -7, …, +7, +8), 𝐷𝑖,𝑡 is a dummy variable having the value of 1 on day 𝑖 and zero otherwise, 𝑅𝑂𝑀𝑡 is a dummy variable that takes the value 1 on rest-of-the-month days (namely, other than -8,-7, …, +7, +8). Intercept terms have been omitted to avoid dummy variable traps in the regression. Estimates that are significant at a 5% level are in bold faces. * coefficient is significant at 1% level

** coefficient is significant at 5% level *** coefficient is significant at 10% level

(15)

P. 15 5.2 The effect of U.S. macroeconomic announcements on Latin-American countries.

The periods where the turn-of-the-month period exists are defined. The influence of U.S. macroeconomic announcements can now be tested. The influence of macroeconomic announcements on the returns can be found in table 4. Despite the findings of Verma and Ozuna (2005) that stock market movements in Latin America cannot be attributed to movements in cross-country macroeconomic variables, table 4 shows that some macroeconomic announcements have a significant effectt on the returns in the stock markets of Brazil, Chile and Mexico for the measured time period. Table 4 shows that the GDP announcement is significant at a 1% level for Brazil and Chile and at a 5% level for Mexico. Furthermore, table 4 shows that the NAPM announcements is significant at a 1% level for Brazil and Mexico and the PACR announcement is significant at a 5% level for Brazil. Consequently, the U.S. macroeconomic announcements have an influence at the stock markets of Brazil, Chile and Mexico.

Table 4

Returns Brazil, Chile**** and Mexico for period 2000-2014

CPI 0.011747 (0.008298) 0.004648 (0.007194) 0.008504 (0.007644) ECI -0.001920 (0.002616) -0.00176 (0.001458) 0.000098 (0.001638) ES 0.001045 (0.001337) -0.000370 (0.000861) -0.000320 (0.000923) GDP 0.004777* (0.001562) 0.002450* (0.000753) 0.002010** (0.000974) IEPI -0.000880 (0.001391) -0.000630 (0.000793) -0.001050 (0.000872) NAPM 0.005112* (0.001563) 0.001318 (0.000903) 0.004587* (0.001173) NONNAPM -0.001620 (0.001383) -0.001250 (0.000857) -0.000910 (0.001029) PACP -0.002870 (0.002656) 0.000252 (0.001737) -0.000440 (0.001654) PACR 0.006017** (0.002560) 0.002806*** (0.001454) 0.000962 (0.002165) PPI RE 0.000787 (0.001634) -0.012440 (0.008145) -0.000260 (0.000819) -0.00437 (0.007162) -0.000580 (0.000962) -0.008980 (0.007580)

Note: regression fomula has the following form: 𝑟𝑡= 𝑐 + ∑11𝑚=1𝛼𝑚𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆𝑚,𝑡+ 𝜀𝑡 Where 𝑟𝑡 is the return for the chosen indexes on day t, 𝑐 is the intercept term, 𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆𝑚,𝑡 is defined as a dummy variable for the macroeconomic news announcement 𝑚 =

(𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 𝑅𝑒𝑝𝑜𝑟𝑡1, 𝑁𝐴𝑃𝑀 (𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑖𝑛𝑔)2, … , 𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 𝑐𝑜𝑠𝑡 𝑖𝑛𝑑𝑒𝑥6), that takes a value of 1, if news 𝑚 occurs on day 𝑡, otherwise zero * coefficient is significant at 1% level

** coefficient is significant at 5% level *** coefficient is significant at 10% level **** period measured for Chile is 2002-2014

However, the significance of the effect of macroeconomic announcements tested is for period 2000-2014 for Brazil and Mexico and period 2002-2014 for Chile is larger than for period 2000-2007 for Brazil and Mexico and 2002-2007 for Chile. Table 5 shows that the GDP announcement is only significant at a 5% level for Chile while in table 4 the GDP announcements are significant at a 5% level for Mexico and a 1% level for Brazil and Chile.

(16)

P. 16

El Hedi Arouri, Bellalah and Nguyen (2010) found that Latin American countries became more correlated with the world market during the crisis in 2008. This indicates that the Latin American countries became more integrated after the crisis. Consequently, the increased integration can be an explanation for the higher explanatory value of the GDP during the period measured in table 4.

However, the NAPM announcement is for period 2000-2007 for Brazil and Mexico and for period 2002-2007 for Chile an announcement with evenly strong explanatory value than for period 2000-2014 for Brazil and Mexico and for period 2002-2014 for Chile. This indicates that some U.S. macroeconomic announcements are not sensitive for the level of integration of Latin American countries. Consequently, the effect of integration seems to differ among U.S. macroeconomic announcements.

Table 5

Returns Brazil, Chile**** and Mexico for period 2000-2007

CPI 0.010944 (0.013412) -0.000018 (0.001046) -0.000046 (0.001468) ECI -0.001290 (0.003565) -0.002450 (0.002187) 0.002801 (0.002530) ES 0.002008 (0.001889) -0.001470 (0.001159) -0.00104 (0.001280) GDP 0.003741*** (0.002144) 0.002149** (0.001028) 0.000967 (0.001362) IEPI -0.00153 (0.002254) -0.001350 (0.001091) -0.000790 (0.001280) NAPM 0.006141* (0.001983) 0.001919*** (0.001029) 0.006750* (0.001415) NONNAPM -0.00208 (0.001969) -0.001330 (0.001075 -0.001030 (0.001392) PACP -0.004130 (0.003064) 0.000392 (0.001769) 0.002074 (0.001935) PACR 0.006500*** (0.003771) 0.002116 (0.002012) 0.001072 (0.003737) PPI RE 0.002716 (0.002141) -0.013200 (0.013110) -0.000033 (0.001399) -0.003180** (0.001574) -0.000390 (0.001413) 0.001519 (0.001568)

Note: regression fomula has the following form: 𝑟𝑡= 𝑐 + ∑11𝑚=1𝛼𝑚𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆𝑚,𝑡+ 𝜀𝑡 Where 𝑟𝑡 is the return for the chosen indexes on day t, 𝑐 is the intercept term, 𝑀𝐴𝐶𝑅𝑂𝑁𝐸𝑊𝑆𝑚,𝑡 is defined as a dummy variable for the macroeconomic news announcement 𝑚 =

(𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 𝑅𝑒𝑝𝑜𝑟𝑡1, 𝑁𝐴𝑃𝑀 (𝑚𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑖𝑛𝑔)2, … , 𝐸𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡 𝑐𝑜𝑠𝑡 𝑖𝑛𝑑𝑒𝑥6), that takes a value of 1, if news 𝑚 occurs on day 𝑡, otherwise zero * coefficient is significant at 1% level

** coefficient is significant at 5% level *** coefficient is significant at 10% level **** period measured for Chile is 2002-2007

Nikkinen, Sahlström and Äijö (2007b) found in their paper that the turn-of-the-month effect completely disappeared after U.S. macroeconomic announcements have been accounted for. Table 6 shows that for Brazil and Mexico residuals of the turn-of-the-month effect still exist. The last trading day of the month is significant at a 5% level for Brazil and a 1% level for Chile. However, the turn-of-the-month effect of Mexico disappeared completely, this indicates that the U.S. macroeconomic announcements are a reason for the turn-of-the-month effect in Mexico. Consequently, it is not possible to reject the influence of U.S. macroeconomic announcements in Mexico for the tested period. Furthermore, the

(17)

P. 17

macroeconomic announcements have explained part of the turn-of-the-month effect in Brazil and Chile so rejecting the influence of U.S. macroeconomic announcements for these countries is not possible.

Table 6

Regression on residuals Brazil, Chile**** and Mexico during period 2000-2014

Residuals Brazil Chile Mexico

-1 0.002411** (0.001138) 0.002937* (0.000716) 0.000989 (0.000830) -2 -0.000280 (0.001557) 0.001114 (0.000850) -0.000150 (0.001060) -3 0.001548 (0.001350) 0.000361 (0.000793) 0.001265 (0.000965) -4 0.000997 (0.001441) -0.001190*** (0.000642) 0.001452 (0.001075) -5 0.000677 (0.001385) -0.000400 0.000663 0.000244 (0.000938) -6 -0.001780 (0.001294) 0.000219 (0.000800) 0.000022 (0.000825) -7 -0.002070 (0.001472) -0.001330 (0.000887) -0.000850 (0.001068) -8 -0.001280 (0.001646) 0.000060 (0.000793) -0.001090 (0.001115) ROM 0.000583 (0.000647) 0.000680*** (0.000378) 0.000493 (0.000456) 1 2 3 4 5 6 7 8 0.000012 (0.001519) 0.001313 (0.001467) 0.000285 (0.001280) -0.003090** (0.001455) 0.000192 (0.001210) -0.000910 (0.001444) -0.000930 (0.001399) 0.000174 (0.001309) 0.000000 (0.000879) -0.001220 0.000765 0.000000 (0.000810) -0.001020 (0.000903) -0.000960 (0.000827) -0.000830 (0.000942) -0.000700 (0.000868) -0.000280 (0.000844) 0.000221 (0.001146) 0.001496 (0.000986) -0.000230 (0.000977) -.001730*** (0.001000) -0.000850 (0.000826) -0.000770 (0.000896) -0.001300 (0.001047) -0.001170 (0.001013)

Note: regression formula has the following form: 𝑟𝑒𝑠𝑖𝑑𝑡= ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 Where 𝑟𝑒𝑠𝑖𝑑𝑡, refers to the obtained residuals from regression (2) 𝐷𝑖,𝑡 is a dummy variable having the value of 1 on day 𝑖 and zero otherwise, 𝑅𝑂𝑀𝑡 is a dummy variable that takes the value 1 on rest-of-the-month days (namely, other than -8,-7, …, +7, +8). Intercept terms have been omitted to avoid dummy variable traps in the regression.

* coefficient is significant at 1% level ** coefficient is significant at 5% level *** coefficient is significant at 10% level **** period measured Chile is 2002-2014

Table 7 shows the results for period 2000-2007 for Brazil and Mexico and for period 2002-2007 for Chile. El Hedi Arouri, Bellalah and Nguyen (2010) found that Latin American countries became more correlated with the world market during the crisis in 2008 so this period can be seen as less integrated with respect to the world market. The results from table 7 show that when the Latin American countries are more segmented, the explanatory value of the macroeconomic announcements decreases. The significance of the last trading day of the month after macroeconomic announcements have been accounted for, is for Brazil significant

(18)

P. 18

at a 1% level. However, for the period 2000-2014 the significance was 5%, which indicates that the explanatory value of macroeconomic announcements are weaker for Brazil when the Latin American countries are more segmented.

Table 7

Regression on residuals Brazil, Chile**** and Mexico during period 2000-2007

Residuals Brazil Chile Mexico

-1 0.003679* (0.001319) 0.003411* (0.000833) 0.001865*** (0.001004) -2 0.000932 (0.002226) 0.001272 (0.001241) -0.000290 (0.001525) -3 0.001278 (0.001925) 0.000447 (0.001130) 0.000458 (0.001401) -4 0.000856 (0.001808) -0.000340 (0.000957) 0.001456 (0.001260) -5 0.001506 (0.001943) 0.000150 (0.001012) 0.000530 (0.001249) -6 -0.000160 (0.001792) 0.000769 (0.001085) 0.000219 (0.001262) -7 -0.001830 (0.002147) 0.000478 (0.001014) 0.000684 (0.001303) -8 -0.000500 (0.002191) -0.001160 (0.001183) -0.002200 (0.001465) ROM 0.000427 (0.000875) 0.000650 (0.000486) 0.001061*** (0.000606) 1 2 3 4 5 6 7 8 0.001366 (0.001904) 0.003570*** (0.002064) 0.000593 (0.001821) -0.003150 (0.001977) 0.000195 (0.001817) 0.000303 (0.001933) -0.001950 (0.002023) 0.001027 (0.001895) 0.000822 (0.001007) -0.000045 (0.000950) -0.000068 (0.001033) -0.000400 (0.001046) -0.000710 (0.001145) -0.000610 (0.001203) -0.002050*** (0.001047) 0.000099 (0.001087) 0.002342*** (0.001378) 0.002799** (0.001353) -0.000031 (0.001365) -0.000740 (0.001420) -0.000290 (0.001235) -0.001540 (0.001095) -0.003040 (0.001451) -0.000140** (0.001594)

Note: regression formula has the following form: 𝑟𝑒𝑠𝑖𝑑𝑡= ∑8𝑖=−8𝛼𝑖𝐷𝑖,𝑡+ 𝛼0𝑅𝑂𝑀𝑡+ 𝜀𝑡 Where 𝑟𝑒𝑠𝑖𝑑𝑡, refers to the obtained residuals from regression (2) 𝐷𝑖,𝑡 is a dummy variable having the value of 1 on day 𝑖 and zero otherwise, 𝑅𝑂𝑀𝑡 is a dummy variable that takes the value 1 on rest-of-the-month days (namely, other than -8,-7, …, +7, +8). Intercept terms have been omitted to avoid dummy variable traps in the regression.

* coefficient is significant at 1% level ** coefficient is significant at 5% level *** coefficient is significant at 10% level **** period measured Chile is 2002-2007

Furthermore, the turn-of-the-month effect for Mexico completely disappeared after U.S. macroeconomic announcements are accounted for during period 2000-2014. However, Table 7 shows that there are residuals left from the turn-of-the-month effect in Mexico during period 2000-2007. The second trading day of the month is significant at a 5% level and there are also residuals left from the first and last trading day of the month since these trading days are significant at a 10% level. Even though these results are partly marginally significant, this indicates that when the Latin American countries are more segmented, the explanatory value

(19)

P. 19

of U.S. macroeconomic announcements are weaker. In addition to the results for Brazil and Mexico in table 7, there are also residuals from the turn-of-the-month effect for Chile during period 2002-2007. The U.S. macroeconomic announcements do reduce the turn-of-the-month effect during the tested period, but does not completely explain the anomaly. Consequently, the U.S. macroeconomic announcements are not the reason for the turn-of-the-month effect when the market is segmented from the world market. However, the U.S. macroeconomic announcements do have some explanatory value since the turn-of-the-month effect is weaker after U.S. macroeconomic variables are accounted for.

Consequently, the findings of Nikkinen, Sahlström and Äijö (2007b) cannot be generalized to the Latin American countries because the turn-of-the-month effect have been found for these countries. Therefore, the theory that U.S. macroeconomic announcements are the only reason for the turn-of-the-month effect can be rejected. If the U.S. macroeconomic announcements would explain the turn-of-the-month effect, then the turn-of-the-month effect should not exist in the tested countries. However, the turn-of-the-month effect exists in the tested countries and the turn-of-the-month effect is not completely explained by the U.S. macroeconomic announcements. Consequently, there must be another explanation for the turn-of-the-month effect in Brazil, Chile and Mexico.

Earlier hypothesis like the ‘preferred habitat’ hypothesis had the same problem when trying to explain the turn-of-the-month effect. Finding a relation between the moment when accrued real wages are paid and the turn-of-the-month effect is achievable (Ogden, 1990). However, no hypothesis has proven to be strong enough to be seen as the only reason for the turn-of-the-month effect due to different characteristics in world markets and the presence of the turn-of-the-month effect in these different markets. The U.S. macroeconomic announcements hypothesis is no exception to this phenomenon.

5.3 Robustness of results

The results are tested with a Durbin-Watson autocorrelation test, and no autocorrelation has been found in the residuals. Furthermore, the regressions are made with the assumption of heteroscedastic standard errors is Stata, so homoscedasticity was not a problem. However, when testing the regression (1a), multicollinearity was a problem. This was solved by reducing the amount of independent variables resulting in regression (1b). Consequently, multicollinearity is not a problem in the results.

6. Conclusion

This paper focuses on the relation between U.S. macroeconomic announcements and the turn-of-the-month effect in Brazil, Chile and Mexico. Furthermore, the study provides a critical look at the findings of Nikkinen, Sahlström and Äijö (2007b) who found that U.S. macroeconomic announcements are the reason for the turn-of-the-month effect. All of the existing evidence supporting the theory of Nikkinen, Sahlström and Äijö (2007b) that U.S. macroeconomic news announcements are the reason for the turn-of-the-month effect is from markets which are highly integrated (Nikkinen et al., 2006). Nikkinen et al. (2006) found that Latin American markets are more segmented from the world market and are therefore less sensitive for news announcements from the world market. Therefore, the influence of U.S. macroeconomic announcements on the stock markets of Latin American countries can be challenged.

The results of the study show that the turn-of-the-month effect exists for the period 2000-2014 for Brazil and Mexico and for Chile in the period of 2002-2014. Furthermore, the turn-of-the-month effect is found for Brazil in Mexico in the period 2000-2007 and for Chile

(20)

P. 20

in period 2002-2007. However, during the tested 2008-2014 period, the turn-of-the-month effect has disappeared for Brazil, Chile and Mexico. Ambiguous results have been found for Latin American countries over time. Therefore these results are in line with earlier studies (Agrawal & Tandon, 1994; Batten & Szilagyi, 2011; Giovanis, 2009; Kunkel, Compton & Beyer, 2003; McConnell & Wu, 2008). After 2008 the Latin American countries became more integrated with the world market (El Hedi Arouri, Bellalah and Nguyen, 2010). Therefore, the disappearance of the turn-of-the-month effect cannot be explained by segmentation of the Latin American countries from the world market.

The effect of integration of U.S. macroeconomic announcements on the stock markets of Brazil, Chile and Mexico seems to differ. While the NAPM announcement is hardly influenced by the difference in integration, the effect of the GDP announcement is reduced in the period where the Latin American countries are more segmented.

The results show that U.S. macroeconomic announcements do reduce the turn-of-the-month effect during the tested period, but do not completely explain the anomaly. During period 2000-2007 when the Latin American markets are believed to be more segmented (El Hedi Arouri, Bellalah and Nguyen, 2010) the U.S. macroeconomic announcements fail to explain the turn-of-the-month effect for Mexico and Brazil. The U.S. macroeconomic announcements also fail to explain the turn-of-the-month effect found for Chile during period 2002-2014. However, the turn-of-the-month effect is reduced for Brazil, Chile and Mexico after macroeconomic announcements are accounted for.

The U.S. macroeconomic announcements do explain the turn-of-the-month effect for Mexico during the period 2000-2014, but fails to explain the turn-of-the-month effect for Brazil in the same time period and for Chile in period 2002-2014. However, the turn-of-the-month effect is reduced for Brazil and Chile after U.S. macroeconomic announcements are accounted for. Consequently, the explanatory value of the macroeconomic announcements are stronger when the markets are more integrated.

Some recommendations for investigations in the future should be made. First, the turn-of-the-month effect has disappeared during period 2008-2014 for Brazil, Chile and Mexico. Further research could investigate if this a coincidence or a trend. Second, the results in this paper shows that the explanatory value of U.S. macroeconomic announcements on the turn-of-the-month effect is greater when the markets are more integrated. However, a causal relationship has not been accomplished. Further research could focus on accomplishing a causal relationship between the explanatory value of U.S. macroeconomic announcements on the turn-of-the-month effect and the segmentation of the market. At last, the hypothesis that macroeconomic announcements are the reason for the turn-of-the-month effect is rejected. Therefore, further research could focus on finding an explanation for the turn-of-the-month effect.

This paper has implications for academic researchers and investors. Academic

researchers have been looking for an explanation for the turn-of-the-month effect for decades and the paper of Nikkinen, Sahlström and Äijö (2007b) found a new, unchallenged

explanation. However, this explanation is rejected since the turn-of-the-month effect in Latin America is only partly explained by the U.S. macroeconomic announcements. Therefore, investors cannot count on this theory to make investment strategies. Furthermore, the disappearance of the turn-of-the-month effect has been found for Brazil, Chile and Mexico during period 2008-2014. This is also something to take in mind for investors when making an investment strategy.

(21)

P. 21 Literature list

Agrawal, A., & Tandon, K. (1994). Anomalies or illusions? Evidence from stock markets in eighteen countries. Journal of international Money and Finance, 13(1), 83-106.

Ariel, R. A. (1987). A monthly effect in stock returns. Journal of Financial Economics, 18(1), 161-174.

Balduzzi, P., Elton, E. J., & Green, T. C. (2001). Economic news and bond prices: Evidence from the US Treasury market. Journal of financial and Quantitative analysis, 36(04), 523-543.

Basher, S. A., & Sadorsky, P. (2006). Day-of-the-week effects in emerging stock markets.

Applied Economics Letters, 13(10), 621-628.

Batten, J., & Szilagyi, P. G. (Eds.). (2011). The impact of the global financial crisis on

emerging financial markets (Vol. 93). Emerald Group Publishing.

Bekaert, G., & Harvey, C. R. (1995). Time‐varying world market integration. The Journal of

Finance, 50(2), 403-444.

Becker, K. G., Finnerty, J. E., & Friedman, J. (1995). Economic news and equity market linkages between the US and UK. Journal of Banking & Finance, 19(7), 1191-1210. Bureau of Labour and Statistics. (2016) Retrieved May, 30, 2016 from

http://www.bls.gov/schedule/archives/all_nr.htm

Bodie, Z., Kane, A., & Marcus, A. J. (2011). Investment and portfolio management.

Bollerslev, T., Cai, J., & Song, F. M. (2000). Intraday periodicity, long memory volatility, and macroeconomic announcement effects in the US Treasury bond market. Journal of

empirical finance, 7(1), 37-55.

Booth, G. G., Kallunki, J. P., & Martikainen, T. (2001). Liquidity and the turn-of-the-month effect: evidence from Finland. Journal of International Financial Markets, Institutions

and Money, 11(2), 137-146.

Chordia, T., Roll, R., & Subrahmanyam, A. (2001). Market liquidity and trading activity. The

Journal of Finance, 56(2), 501-530.

Economic Indicators (n.d.) Retrieved May, 23, 2016 from

http://www.ine.cl/canales/menu/calendario_de_indicadores_eng.php?lang=eng El Hedi Arouri, M., Bellalah, M., & Nguyen, D. K. (2010). The co-movements in

international stock markets: new evidence from Latin American emerging countries.Applied Economics Letters, 17(13), 1323-1328.

Estadícas Económicas (2016). Retrieved June, 3, 2016 from http://www.inegi.org.mx/saladeprensa/doc/cal_2016.pdf Eun, C., Resnick, B., 1984. Estimating the correlation structure of international share prices. Journal of Finance 39, 1311–1324.

Fleming, M. J., & Remolona, E. M. (1999). Price formation and liquidity in the US Treasury market: The response to public information. The journal of Finance, 54(5), 1901-1915.

Gay Jr, R. D. (2011). Effect of macroeconomic variables on stock market returns for four emerging economies: Brazil, Russia, India, and China. International Business &

Economics Research Journal (IBER), 7(3).

Gerlach, J. R. (2007). Macroeconomic news and stock market calendar and weather anomalies. Journal of Financial Research, 30(2), 283-300.

Giovanis, E. (2009). Calendar effects in fifty-five stock market indices. Global Journal of

Finance and Management, 1(2).

Graham, M., Kiviaho, J., & Nikkinen, J. (2012). Integration of 22 emerging stock markets: A three-dimensional analysis. Global Finance Journal, 23(1), 34-47.

(22)

P. 22

Graham, M., Nikkinen, J., & Sahlström, P. (2003). Relative importance of scheduled macroeconomic news for stock market investors. Journal of Economics and Finance,

27(2), 153-165.

Hensel, C. R., & Ziemba, W. T. (1996). Investment results from exploiting turn-of-the-month effects. The Journal of Portfolio Management, 22(3), 17-23.

Jones, C. M., Lamont, O., & Lumsdaine, R. L. (1998). Macroeconomic news and bond market volatility. Journal of Financial Economics, 47(3), 315-337.

Karpoff, J. M. (1987). The relation between price changes and trading volume: A survey.

Journal of Financial and quantitative Analysis, 22(01), 109-126.

Kayacetin, V., & Lekpek, S. (2016). Turn-of-the-month effect: New evidence from an emerging stock market. Finance Research Letters.

Kunkel, R. A., & Compton, W. S. (1998). A tax-free exploitation of the turn-of-the-month effect: CREF. Financial Services Review, 7(1), 11.

Kunkel, R. A., Compton, W. S., & Beyer, S. (2003). The turn-of-the-month effect still lives: the international evidence. International Review of Financial Analysis, 12(2), 207-221. Lakonishok, J., Shleifer, A., Thaler, R., & Vishny, R. (1991). Window dressing by pension

fund managers (No. w3617). National Bureau of Economic Research.

Lakonishok, J., & Smidt, S. (1988). Are seasonal anomalies real? A ninety-year perspective.

Review of Financial Studies, 1(4), 403-425.

Maher, D., & Parikh, A. (2013). The turn of the month effect in India: A case of large institutional trading pattern as a source of higher liquidity. International Review of

Financial Analysis, 28, 57-69.

Marquering, W., Nisser, J., & Valla, T. (2006). Disappearing anomalies: a dynamic analysis of the persistence of anomalies. Applied Financial Economics, 16(4), 291-302.

Martikainen, T., Perttunen, J., & Ziemba, W. T. (2012). The turn-of-the-month effect in the world’s stock markets, January 1988-January 1990. In Calendar Anomalies And

Arbitrage (pp. 355-363).

McConnell, J. J., & Xu, W. (2008). Equity Returns at the Turn of the Month. Financial

Analysts Journal, 64(2), 49-64.

McNichols, M. (1988). A comparison of the skewness of stock return distributions at earnings and non-earnings announcement dates. Journal of Accounting and Economics, 10(3), 239-273.

Muradoglu, G., Taskin, F., & Bigan, I. (2000). Causality between stock returns and macroeconomic variables in emerging markets. Russian & East European Finance

and Trade, 36(6), 33-53.

Nikkinen, J., Omran, M., Sahlström, P., & Äijö, J. (2006). Global stock market reactions to scheduled US macroeconomic news announcements. Global Finance Journal, 17(1), 92-104.

Nikkinen, J., Sahlström, P., & Äijö, J. (2007a). Do the US macroeconomic news announcements explain turn-of-the-month and intramonth anomalies on European stock markets. Journal of Applied Business and Economics, 7(3).

Nikkinen, J., Sahlström, P., & Äijö, J. (2007b). Turn‐of‐the‐month and intramonth effects: Explanation from the important macroeconomic news announcements. Journal of

Futures markets, 27(2), 105-126.

Nikkinen, J., Sahlström, P., Takko, K., & Äijö, J. (2009). Turn-of-the-month and Intramonth Anomalies and US Macroeconomic News Announcements on the Thinly Traded Finnish Stock Market. International Journal of Economics and Finance.

Nofsinger, J. R., & Prucyk, B. (2003). Option volume and volatility response to scheduled economic news releases. Journal of Futures Markets, 23(4), 315-345.

A study of the relation between U.S. macroeconomic announcements and the turn-of-the-month effect in Latin American countries