Sentiment effects on Dutch stock market indices: ‘

Sentiment effects on Dutch stock market indices:

‘Interactions of consumer confidence, producer confidence and industrial

production with the AEX and Midcap.’

University of Groningen

International Business Management – Specialization International Financial Management Uppsala University

Business and Economics – Specialization International Financial Management

Tim Sebastiaan Gerritzen, s1384309

2

Abstract

This research examines sentiment effects on two Dutch stock market indices, the AEX and Midcap. Consumer confidence and Producer confidence are used as a sentiment measure whereas Industrial production growth is employed as a fundamental sentiment measure. Ordinary least squares regression, 1 lag Vector Auto Regression (VAR) and impulse response functions out of the VAR are employed to determine contemporaneous and lagged relationships, with monthly observations spanning January 1994 to December 2009. Results for producer confidence support previous research, where periods of high sentiment are followed by lower returns the next period for both the AEX and Midcap. Consumer confidence shows a positive effect for the AEX, both for contemporaneous and lagged values, indicating a longer lasting positive effect. Surprisingly, results indicate that sentiment effects are stronger for the AEX, possibly attributable to the high number of individual investors.

3

Table of Contents

Abstract ... 2

Introduction ... 4

I. Literature Review ... 6

A. The history of stock market inefficiencies ... 6

B. The birth of Behavioral Finance and investor sentiment ... 8

C. Sentiment measures ... 9

D. Sentiment and stock market interactions ... 11

E. Small vs. Big stocks and sentiment ... 13

II. Motivation ... 14

A. Consumer confidence index ... 14

B. Producer confidence index ... 16

C. Industrial Production growth ... 17

D. Market sizes ... 17

Table 1 ... 18

III. Data ... 18

A. Sentiment measures ... 18

B. Stock market returns ... 20

Table 2 ... 20

IV. Methodology and results ... 21

A. Correlations ... 21

Table 3 ... 21

B. Ordinary least squares regressions ... 21

Table 4 ... 25

C. Long term sentiment effects ... 26

D. Vector Auto Regression ... 26

Table 5 ... 28

E. Impulse response functions ... 29

Table 6 ... 30

Table 7 ... 32

Table 8 ... 33

V. Conclusion ... 35

A. Conclusion ... 35

B. Limitations and suggestions for further research ... 37

VI. Appendix ... 39

Table 9 ... 39

4

Introduction

The past decade, stock markets around the world have seen turbulent times. With markets rising high in 2000 and 2007, but plunging into deep depths the subsequent years. The past decade has seen two major crashes which involved widespread panic and major uncertainty. Both the market highs and market lows have once more raised questions about stock market rationality. Either questioning the highs, illustrated in the ‘irrational exuberance’ speech by Alan Greenspan in 1996 or the extreme lows of otherwise healthy companies. A considerable amount of research has been focusing on driving forces behind stock market movements. General sense implies that there are rational actors investing for economic profit. However, past research has shown that there are more forces than rational economic deliberation influencing stock markets around the world.

In this paper I intend to look at a small selection of those forces. The field of behavioral finance has produced evidence on how reigning sentiment in the stock market can influence prices. Research has shown that there are different effects on prices induced by different types of sentiment. Additionally, several academics have produced evidence that these sentiments work as a system with stock market returns. In this paper I will focus on three types of sentiment and their interaction with different sizes of stock markets in The Netherlands. The first two sentiment measures I employ are the monthly producer and consumer confidence indices of The Netherlands. My third sentiment measure is a fundamental sentiment measure, specified by the monthly industrial production growth index. The difference between the first two and the third is that the first two are based on perceptions and opinions of producers and consumers. Thus making the confidence indices of producers and consumers ‘true’ sentiment measures. On the other hand, the fundamental sentiment measure is based on a real fundamental figure, without opinions or perception. There has been extensive research on the interaction of the stock market with the consumer confidence index and some research on the interaction of industrial production growth and the stock market. However, producer confidence and its interaction with stock markets has not been looked at despite the fact that the producer confidence index has interesting features which differentiate it from the consumer confidence index.

5 the AEX and for medium sized stocks the Midcap. My research will cover the past 15 years of the Dutch stock markets, namely January 1994 to December 2009. Furthermore, sentiment has shown to have both contemporaneous and lagged effects on stock market indices. Usually following the general view that high sentiment is followed by lower returns in the next period(s). Therefore, in this study I will answer the following question: ‘What are the contemporaneous and lagged influences of consumer confidence, producer confidence and industrial production growth on the AEX and Midcap?’ In addition, I will give an insight into the underlying relations that these variables have amongst themselves, to illustrate the complex system that sentiment and stock markets represent.

This research will add to the existing research into the interactions of stock markets and sentiments on two aspects. First and foremost, the addition of producer confidence has not yet been researched. Producer confidence has unique features which differentiate it to consumer confidence. It is known to be a better and faster predictor of economic activity in The Netherlands. This is mainly ascribed to the fact that the Dutch economy is very open and susceptible to foreign influences. Accordingly, if economic activity is changing internationally, the producer confidence index will show these effects before they reach consumers and consequently the consumer confidence index. This is due to the fact that the Dutch producers are very dependent on international economic activity for success. This feature of producer confidence makes it interesting to research and adds to a better understanding of variables influencing stock markets. This is interesting for both practitioners and academics. Secondly, most research into sentiment effects and stock markets so far has focused on U.S. stock markets and a Dutch perspective is interesting. Yet again, mainly due to the fact that the Dutch economy is very open and depends highly on international activity. Additionally, research has shown that sentiments can have different effects depending on the country specifics as in culture and differing market institutions. Bearing in mind that the Dutch culture is quite different to US culture and that regulations and institutions are somewhat different too. The Dutch perspective on sentiments and stock market workings is interesting for academics in particular. However, a practitioner with interest in sentiment and stock markets interactions might find use in this research as well.

6 static OLS regressions to determine the lagged effects of the variables on the stock market indices. In the second OLS I add lagged values of the AEX and Midcap to see to what extant returns are driven by lags of itself. In the third I will employ lagged values of the sentiment measures to determine the lagged effects, to show the stock markets correcting features. Since previous research suggests that sentiments act as a system with each other, I will employ a 1 lag Vector Auto Regression (VAR) analysis. By using a VAR analysis I can determine the interaction these measures have with stock returns and each other. Finally, I will use impulse response functions generated from the VAR model to see the effect of onetime shocks to the system.

This paper is structured as follows. In section I will provide an extensive literature review and elaborate on past and current research. Section II will focus on my expected relations between my variables and motivate my choice for the variables. Section III will include my data. In section IV my methodology and results will be presented followed by my conclusion in section V. Suggestions for further research and limitations will be provided in the conclusion as well. My appendix is in section VI, followed by the references in section VII.

I. Literature Review

A. The history of stock market inefficiencies

7 informational efficient market is still unreachable when the zero cost property specified by Grossman and Stiglitz is fulfilled. Beja’s genuine trading model suggests that zero cost information is not enough to have fully data-informative prices. He adds that price information in such processes is limited and that collecting costly information is economically viable even when prices reflect some information about value. This follows from the argument that superior information is never fully reflected in market prices. Beja’s conclusion is that when prices are believed to be formed by trader demands, traders should carefully consider any useful information they may have. Even if every other trader has access to that information and even if prices reflect some information.

8 no fundamental risk. Noise traders effectively create their own space in financial markets. Following this, noise traders can earn higher average returns than actual rational investors. If noise trader either overestimate returns or underestimate risk, they will invest more in risky assets than rational investors thus gaining higher returns. The noise trader creates risk which makes a risk averse arbitrageur less likely to drive prices down of an asset. Therefore, earning higher expected return ascribed to their own influence, instead of bearing actual fundamental risk. Lo and Mackinley (1999) add to the noise theory by arguing that inefficiency is necessary for a liquid market. They argue that the degree of inefficiency determines the amount of effort investors are willing to expend to gain profit opportunities. They see excess return as economic rent to people who are willing to engage in costly information gathering activities.

B. The birth of Behavioral Finance and investor sentiment

9 C. Sentiment measures

There are several theories on how investor sentiment is actually influencing the price formation of securities. However, to understand the dynamics one has to understand the different types of sentiments and actors that interact on the stock market. First of all, a comprehensive study by Brown and Cliff (2004) on the effects of sentiment on near term stock returns define investor sentiment as ‘…the expectations of market participants relative to a norm: a bullish (bearish) investor to be above (below) average, whatever that average may be’. In summary, investor sentiment is the expectation that the market is going in a certain direction, whether up (Bullish) or down (Bearish). Furthermore, there are two types of actors in the stock market. Usually they are referred to as private investors and institutional investors that both make different systematic misevaluations. They define institutional investors as professionals making a living in investing whereas private investors are people who invest next to their normal job (Brown & Cliff, 2004; Verma et al., 2008). In general, the academic literature refers to individual investors as the irrational investors that create the general noise on the stock market. Furthermore, institutional investors are referred to as ‘rational’, mostly unresponsive to pseudo information induced trading. These two groups react differently to signals which form their sentiments and their effect on stock prices. While both groups do get affected by sentiment, and both sentiments are highly correlated with stock market returns (Verma & Verma, 2008). Brown and Cliff (2005) examine long term effects of institutional sentiment and find a strong relationship with long term stock returns. Accordingly, the effects of the two types of sentiment on stock returns are generally presumed to be different. Private investors for instance, are presumed to have no means in affecting prices by themselves (Brown & Cliff, 2004; Thaler, 1999). Brown and Cliff (2004) also state that perhaps institutional sentiment has a greater influence on larger stocks because of a disproportionally greater ownership by that group in those larger stocks. The same goes for smaller stocks and private investors.

10 measure for institutional investor sentiment, whereas the AAII survey was used as a measure for individual sentiment. This difference in measurement is due to the people answering the questionnaire, individuals or institutional investors. Several studies also employ indirect measures of sentiment, summarized by Verma et al. (2008). Indirect measures are usually proxying for sentiment as in closed-ended fund’s discount (Lee et al., 1991; Chen et al., 1993; Swaminathan, 1996; Elton et al., 1998; Neal & Wheatley, 1998; Baker & Wurgler, 2006), market-performance measures (Brown & Cliff, 2004), trading-activity measures (Brown & Cliff, 2004; Baker & Wurgler, 2006), derivative variables (Brown & Cliff, 2004), dividend premiums (Baker & Wurgler, 2006) and IPO related measures (Brown & Cliff, 2004; Baker & Wurgler, 2006). However, there is not yet a single measure which is accepted as the correct measure of investor sentiment since they are usually strongly correlated (Verma et al., 2008). The controversy surrounding indirect sentiment measures is visible in papers by Lee et al. (1991), Swamminathan (1996) and Neal and Wheatley (1998) who argue that closed-end fund discount is a measure for investor sentiment. However, Chen et al, (1993) and Elton et al. (1998) bring forth opposing evidence.

11 stocks relative to low-book/market stocks, momentum factors and currency fluctuation. Verma et al (2008) focused on both rational and irrational sentiment and explore how it may affect market returns, the results I will elaborate on in the next subsection. Verma and Verma (2008) researched to what extant institutional and individual sentiments are formed by rational and irrational factors. Verma and Verma (2008) looked at what kind of factors influenced the formation of sentiment measures provided by the AAII survey and II bull/bear spread. The authors find that institutional sentiment is more rational than individual sentiment, where it is mostly based on rational risk factors. However, both individual and institutional sentiment is formed by both rational outlook as well as noise, which are expectations not fully justified by information.

D. Sentiment and stock market interactions

12 investors on stock market returns. Similar to Brown and Cliff (2004), they find that there are significant effects of past stock performance on irrational sentiments, however no effects of past stock performance on rational sentiments. Secondly, there are immediate positive responses of stock returns to irrational investor sentiment. However, these are corrected by means of negative responses in the next periods. This supports the argument that the rational sentiment is a good determinant of stock returns for it symbolizes economic fundamentals. They employ generalized impulses to determine the short and long term effects of sentiments. Schmeling (2009) find that sentiment negatively affects stock returns in the short, medium and long term. They ascribe this to the stock market correcting mispricing due to present sentiments, where high sentiment is followed by low returns. However, they find that these effects wash out over longer horizons of 12 to 24 months. This indicates that there are limits to arbitrage in the short and medium term however diminishes at longer terms. From research by Jansen and Nahuis (2003) it turned out that stock market returns were positively related to changes in consumer confidence in 9 European countries. Similar to the Brown and Cliff (2004) research, yet with a considerably different sentiment measure, Jansen and Nahuis find that stock market returns drive consumer confidence in the short run, however not vice versa. They also find that the stock market-confidence relationship is driven by expectations about economy-wide conditions rather than personal finances. However, Schmeling (2009) finds that in the long run, consumer confidence as a private investor sentiment measure is negatively related to stock returns when controlling for fundamental factors. This is similar to the findings of Verma et al. (2008) and Brown and Cliff (2005) who also show that the short run positive (negative) relation of bullish (bearish) sentiment on stock market returns reverts in the long term.

13 gross domestic product (GDP), and likely positively correlated to other parts of GDP. This would reflect higher expected output by rising stock prices. Additionally, a decrease in discount rates is associated with higher investment and therefore higher output. Lettau and Ludvigson (2001) show that there is an important linkage between the expected stock market return and investments. Their argument is as follows, lower expected stock market return implies lower future stock price and higher future capital cost. Accordingly, investment will fall in the long term. This would imply that economic growth is positively related to stock market returns with a certain lag. Shazbaz et al. (2008) find that in an emerging market setting there is one-way causality between stock market returns to economic growth in the short run. However, on the long run the Granger-Causality estimation shows that there is a bi-directional causality between stock return and economic growth.

E. Small vs. Big stocks and sentiment

Academics have theorized that the effects of sentiments are different depending on the type of stock. This is due to the fact that these stocks have differing elements which makes them hard to compare. One of these elements is the differing ownership structure. Large stocks are usually disproportionally held by institutional investors, whereas smaller stocks are usually held by individual investors. First of all, Brown and Cliff (2004) argue that this differing ownership structure might change sentiment effects. While individual investors cannot change prices by themselves, institutional investors usually are able to influence prices. The interesting thing that Brown and Cliff (2004) show is that sentiment is not only affecting individual investors. The strongest relations are between institutional sentiment measures and large stock returns. This indicates that institutional investors, which are generally assumed to be free from sentiment, are indeed influencing prices with their sentiments. This goes against conventional wisdom that mainly private investors are swayed by sentiment and that their effect would be seen in smaller stocks. As the evidence of Brown and Cliff shows that institutional sentiment has a positive relationship to large stocks.

14 relative to growth stocks. Value stocks are stocks that tend to trade at a lower price relative to their fundamentals whereas growth stocks are stocks which earnings are expected to grow at an above-average rate relative to the market. They show that changes in the buy–sell imbalance of these individual investors are positively correlated with returns of value stocks. This relation shows that when individual investors grow relatively bullish (bearish), the stocks in these portfolios enjoy higher (lower) excess returns. Finally, they show that this relation is stronger for stocks which are profoundly harder to arbitrage, where the effects of individual sentiment are the strongest for those stocks. This is an example of noise trader risk in the stock market. According to Schmeling (2009), Barber et al. (in press-a) investigated stock returns of heavily traded stocks by US individual investors. They show direct evidence that individuals are noise traders. Out of their research they show that individual investor trading is highly correlated similar to Kumar and Lee (2006). Additionally, it shows that stock which are sold by individual investors outperform stocks which are bought by individual investors by 13,5% the following year. This illustrates the inability of individual investors to pick stocks based on fundamental values. The likely cause of these return differentials is strong herd behavior among individual investors. These studies are all in line with the earlier studies of DeLong et al (1990), Black (1986) and Thaler (1999) who assume that noise traders trade on non value relevant information.

II. Motivation

A. Consumer confidence index

15 lower returns using generalized impulse analysis. Finally, Schmeling (2009) used predictive regressions to estimate the effects of sentiment. Schmeling (2009) finds that periods of high investor sentiment are followed by lower returns, thus showing a negative relation. Regarding these three researches, a comparison is hard because all of them use different methodologies in reaching their conclusions. Therefore, to thoroughly grasp the relations of my variables I intend to do four types of regressions. Specifically, Static and non static OLS, VAR and generalized impulse responses. To see the different effects of my variables on the AEX and Midcap, OLS will show contemporaneous and lagged influences. The addition of VAR analysis will introduce the relations as a system, and out of the VAR generalized impulses can be produced. These impulses create shocks in one part of the system to illustrate the variables shock effect on the other variables while keeping the other variable effects constant.

Brown and Cliff (2004) and Verma et al. (2008) both use VAR as a basis. Where Verma et al (2008) mainly base their conclusions on generalized impulse response functions, Brown and Cliff (2004) mainly focus on the interpretation of the VAR results. Finally, Schmeling (2009) uses predictive regression. Schmeling (2009) and Brown and Cliff (2004; 2005) show negative relations when it come to monthly data. This stems from following the line of reasoning that periods of high sentiment are generally followed by lower returns. This is also shown by Verma et al. (2008) on a similar basis, using impulses. Therefore, with regard to consumer confidence and when performing a static OLS, I expect a positive relation for both AEX and Midcap as dependent variables. However, this only applies to static OLS. The static OLS regression will illustrate contemporaneous relations, the initial rise in stock return when sentiment is high. When I add lagged values of consumer confidence in the third OLS and VAR regression I expect the coefficient to be negative in relation to the AEX and Midcap. This is due to the use of monthly data instead of weekly. I expect the preliminary positive reaction to be of short nature, and to be quickly corrected by the stock market. Therefore, when employing a minimum of 1 month lags, the market is probably already in a corrective movement and thus showing a negative coefficient.

16 argued that the effects he found were not immediately transferable across countries and depended on market integrity and cultural effects. Therefore, a Dutch perspective can be interesting for both practitioners and academics. The three types of analyses will show the different types of relations to an extent which has not been shown for the Netherlands.

B. Producer confidence index

In economic forecasts, consumer confidence is usually combined with the producer confidence to get an outlook on economic expectations of both producers and consumers. The combined measure is used as early estimators for other economic indicators. Producer confidence is a compounded sentiment indicator published by the Dutch CBS, the national bureau of statistics. The measure is compounded by combining the answer to three questions regarding producer expectations. The questions involved are the expected output in the next quarter, their opinion on their own current order portfolio and their assessment of the volume of their current products in stock. To show these aggregated opinions in one measure and to additionally show industry production the CBS adjusts the difference of positive and negative answers for seasonal influences. The resulting measure provides a good indication of the sentiment under producers/entrepreneurs. The measure has historically been between -10 and +10 (Centraal Bureau Statistiek, 2009). However, following the past crisis values have gone as low as -23,5 in the end of 2008, illustrating the severity of the current crisis. Producer confidence is a key figure next to the consumer confidence index in estimating economic confidence, also useable as an indirect measure for stock market sentiment.

17 powerful tool of predicting stock market returns. Producer confidence is one of the driving forces of the real economy and therefore very suited to use as an indirect sentiment measure to add to existing indirect measures of stock market sentiment. Due to the fact that producer confidence is another indirect sentiment measure indicating the future expectations of producers instead of consumers, I expect a similar relation as consumer confidence. When conducting the static OLS regression, I expect to see a positive relationship. When I look at my expected relations for lagged values for producer confidence, I expect a negative relation to both the AEX and Midcap. Again, this is in line with the effects for consumer confidence. On the other hand, when using generalized impulses, I expect a positive effect in the first period then reverting to negative in the following periods.

C. Industrial Production growth

In respect to economic growth as a sentiment measure, it is clear that it is different to the previous two measures. First of all, economic growth is considered a fundamentals induced measure, or in other words a rational sentiment measure, whereas producer and consumer confidence stem from opinions of consumers and producers. Following research of Shazbaz et al. (2008) one would expect that economic growth is positively driven by stock market returns on the short term. However, this was a result in an emerging market setting. In a more developed market I expect that industrial production growth is driving stock market returns positively. This is the same result as Verma et al. (2008), as rational sentiment has a more profound effect on stock returns, although the information gets incorporated at a slower pace. Due to the slow incorporation, I expect no effect on stock returns for both AEX and Midcap in static OLS regression. However, in the long term with lagged values, the inclusion of fundamental data might be visible. Therefore, when considering the lagged values of industrial production, the relation is expected to be positive due to the slower inclusion. D. Market sizes

18 acting on non value relevant information, the effect of sentiments on returns should be stronger for the Midcap in comparison to the AEX. To summarize my expected relations in the different analyses’, I summarized them in table 1. The only expected result which is lacking is my expectation that sentiment effects are stronger for the Midcap than the AEX. Table 1

Expected signs of independent variables with AEX and Midcap

Model characteristic Independent variables

Cons. Conf. Prod. Conf. Ind. Prod.

𝑆𝑡𝑎𝑡𝑖𝑐 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠(𝑡=1) + + X

𝑁𝑜𝑛 𝑠𝑡𝑎𝑡𝑖𝑐 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 (𝑡−𝑝) - - +

𝐼𝑚𝑝𝑢𝑙𝑠𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠_{(𝑡=1,2,3..)}

(𝑡 = 1) + + X

(𝑡 > 1) - - +

Expected signs of the relations, grouped per model characteristic. Static indicates contemporaneous relations with no lagged values in the equation, with expected sign. Non static indicates the addition of lagged values in the equation, with expected sign. P indicates the number of lags. Impulse responses indicate the responses at time t. X indicates no expected significant relation. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage

III. Data

A. Sentiment measures

19 years the average consumer has been slightly pessimistic. Furthermore, the maximum value is 27 with a minimum of -40. You can see the descriptive statistics for all the sentiment measures and stock market indices in table 2, at the end of this section.

My second sentiment measure, producer confidence is also compiled monthly by the CBS. A standard panel of companies gets surveyed every week before the start of a new month. The questions are their current status of their inventory, their current order portfolio and their expected order portfolio in the next 3 months. Questions regarding their current or expected status have responses varying from ‘decline’ to ‘neutral’ or ‘increase’. The calculation of the index is similar to the consumer confidence, deducting the negative responses from the positive ones. However, in the producer confidence index larger companies are weighed more heavily than smaller ones as opposed to equal weighing in the consumer confidence index. Furthermore, there are differences in weighing between sectors in the industries and there is a base correction. This is a correction for structural positive or negative responses by companies. Finally, seasonal influences are removed. The period is once more from January 1994 to December 2009. There are 191 observations with no missing values. The mean is 1.17 with a standard deviation of 6.07. The maximum is 9.4 with a minimum of -23.5 (see table 2).

20 opposed to the previous month. The standard deviation is 2.19% with a maximum of 7.79% and a minimum of -6,67% (see table 2).

B. Stock market returns

The two different sizes of stock markets are aggregate indices and gathered from DataStream. The first and largest index is the AEX containing the 25 most traded and larger

stocks. The data gathered is from January 1994 to December 2009, from dates around the 29th

and the 30th of the month. There are no missing values. The measure is ‘Total return’ which

includes dividend. Furthermore, it resembles return per month as opposed to the previous month. Table 2 shows that the AEX had a mean monthly total return of 0.72% with a maximum of 15.86% and a minimum of -20.23%. The standard deviation is 6.15%. The Midcap on the other hand consists of 25 medium size stocks, with the same date range as the AEX and no missing values. As visible in table 2, the mean monthly total return of the Midcap is 0.90%. Furthermore, maximum return was 21.01% with a minimum of -23.96%. A slightly higher standard deviation of 6,44% indicates a slightly higher variance in comparison to the AEX. This higher variance could be explained due to the higher difficulty of valuing and arbitraging Midcap stocks in comparison to the larger stocks in the AEX.

Table 2

Descriptive statistics

AEX Midcap Cons. Conf. Prod. Conf. Ind. Prod.

Mean 0.72 0.90 -4.31 1.17 0.13 Maximum 15.86 21.01 27.00 9.40 7.79 Minimum -20.23 -23.96 -40.00 -23.50 -6.67 Std. Dev. 6.15 6.44 18.40 6.07 2.19 Skewness -0.77 -0.81 -0.14 -1.53 -0.09 Kurtosis 4.46 4.64 1.84 6.19 4.19 N 191 191 191 191 191

21

IV. Methodology and results

A. Correlations

Due to the large number of regressions I utilize, going against common thesis structure, I will present my methodology and results in one section. I believe this increases readability of both sections. Lee et al. (2002), Brown and Cliff (2004) and Verma et al. (2008) suggest that stock returns and sentiments act as a system. To take a preliminary look at the relations, the correlations are provided in table 3. The results are what one could expect from previous research. Brown and Cliff (2004) already showed that sentiment measures are usually highly correlated as can be seen from the high correlation between producer confidence and consumer confidence. Similarly, it is no surprise Midcap and AEX are highly correlated. The correlation is however not extremely close to 1, illustrating that the markets have indeed somewhat differing features. The low correlation with the AEX and producer confidence, and even negative correlation of producer confidence and Midcap is somewhat surprising. This suggests that my theory of initial positive effects of producer confidence might be misplaced. Industrial production is generally lightly correlated to all my variables, which could be expected.

Table 3

Correlation table

AEX Midcap Cons. Conf. Prod. Conf

Midcap 0.84

Cons. Conf 0.12 0.05

Prod. Conf. 0.03 -0.03 0.66

Ind. Prod. 0.09 0.08 0.05 0.09

Pair wise correlations for selected variables used in the analysis. Correlations are based on 190 observations from January 1994 to December 2009. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage.

B. Ordinary least squares regressions

22 is at the end of this sub section. My first OLS regressions are static and structured by the following equation:

(1) 𝑌_𝑡 = 𝑐 + 𝛽₁𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡+ 𝛽₂𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓._𝑡+ 𝛽₃𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑._𝑡+ 𝜀_𝑡

Where 𝑌_𝑡 is the dependent variable measured as the aggregate total return for either

the AEX or Midcap at time t. Furthermore, 𝑐 is a constant and 𝛽₁₋₃ are the coefficients that

will be estimated for the variables at time t. Finally, 𝜀𝑡 is the random error term. The result of

my OLS regression with AEX as a dependent variable is presented in table 4, regression 1. As one can see in table 4, the only significant relation is the positive relation between consumer confidence and AEX (p < 0.1). Furthermore, producer confidence indicates a negative coefficient but insignificant. Finally, industrial production shows no significant relationship either which was expected. In table 4 regression 1, we can see the first OLS results for the Midcap. Again, results show no significant relations for industrial production or producer confidence. Consumer confidence fails to provide any significant relation as well. This is surprising because previous research suggests that consumer confidence should be positive. Furthermore, I expected producer confidence to be positive and significant to both the AEX and Midcap. Finally, the effects were assumed to be stronger for the Midcap than for the AEX which does not show.

To delve into the relations between the variables a slightly deeper I conduct a second OLS regression adding a lagged variable of the dependent variable, of both AEX and Midcap. This is to check whether my dependent variables are actually mainly driven by lags of itself. Variables are often positively and significantly related to lags of itself; however the AEX and Midcap in theory shouldn’t be. It is imaginable that stock markets are driven by their own lags, as the theory of bubble formation would imply. Good returns drive investors to become more bullish, thus forcing up prices further. This is however in contrast with efficient market theory, where in an efficient market previous returns shouldn’t predict future returns. The OLS regression follows the equation:

(2) 𝑌_𝑡 = 𝑐 + 𝛽₁𝑌_𝑡−1+𝛽₂𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡+ 𝛽₃𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓._𝑡+ 𝛽₄𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑._𝑡+ 𝜀_𝑡

Where 𝑌_𝑡 is the dependent variable measured as the aggregate total return for either

the AEX or Midcap at time t. Furthermore, 𝑐 is a constant and 𝛽1−4 are the coefficients that

are estimated for the variables at time t or t-1. Finally, 𝜀𝑡 is the random error term. Including a

23 indeed an efficient market, where the past values do not predict future values. Adding a lagged dependent variable as an independent variable shows that the previous regression results are almost similar (see the result in table 4, regression 2). Consumer confidence is still the only significant variable with a positive relation (p < 0.1). The 1 lag variable of AEX does not improve the results and is not significant. However, if I run the same regression for the Midcap, results are different. The addition of a 1 month lag of the Midcap return in the equation is highly significant and relatively strong (p < 0.01). This indicates that first of all, it seems that the Midcaps previous returns is a stronger predictor of itself than the variables I chose. Secondly, following the results, the Midcap is not efficient in the eyes of the efficient market hypothesis. Where the assumption that past returns do not imply future returns is violated. This result could be ascribed to the fact that there are more noise traders active on the Midcap market. Indicating that possibly investors do get increasingly bullish by higher returns, and thus drive prices up by uncontrolled purchases effectively creating a bubble. This does not explain however why my sentiment measures, especially producer and consumer confidence are not significant. Following academic research, at least consumer confidence should have an impact. In particular if my results imply that there is a level of inefficiency in the Midcap, usually ascribed to noise traders.

Several academics have suggested that the incorporation of information in stock returns may not always be contemporaneous. This is due to time delays in the processing of information concerning noise and rational factors. This is especially true for macroeconomic variables. Due to these delays, there may be a lag between the observations and the incorporation of that information to stock prices. Therefore, if all the variables are measured at a certain time t, it would possibly show an unrealistic contemporaneous relation. For that reason I will do a final OLS regression where all independent variables are one month lags. With this regression, I intend to show the corrective act of the market and expect negative signs to both consumer and producer confidence. With respect to industrial production growth, I expect to see a positive relation, due the longer time processing fundamental data. These OLS regressions are structured by the following equation:

(3) 𝑌𝑡 = 𝑐 + 𝛽1𝑌𝑡−1+𝛽2𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓.𝑡−1+ 𝛽3𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓.𝑡−1+ 𝛽4𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑.𝑡−1+ 𝜀𝑡

Where 𝑌𝑡 is the dependent variable measured as the aggregate total return for either

the AEX or Midcap at time t. In this equation 𝑌_𝑡−1 stands for the dependent added as an

24

that are estimated for the variables at time t-1. Finally, 𝜀_𝑡 is the random error term. You can

see my results for the AEX in table 4, regression 3. First, the difference in lagged and contemporaneous effects is immediately visible. Consumer confidence with a lag 1 has a slightly higher coefficient but is more significant (p < 0.05). Indicating that possibly the positive effect lasts longer than expected for consumer confidence. Producer confidence is slightly significant now as well (p < 0.1) and shows a negative coefficient as expected.

25 Table 4

Ordinary Least Squares Regression results

Dependent Variable: AEX Dependent Variable: Midcap

Regression 1

Independent Variable Coefficient Independent Variable Coefficient

Cons. Conf. 0.060* Cons. Conf. 0.047

(0.032) (0.034)

Prod. Conf. -0.097 Prod. Conf. -0.134

(0.098) (0.103)

Ind. Prod. 0.239 Ind. Prod. 0.250

(0.203) (0.214)

C 1.062** C 1.223**

(0.500) (0.526)

R-squared 0.026 R-squared 0.018

Adjusted R-squared 0.010 Adjusted R-squared 0.002

Regression 2

Independent Variable Coefficient Independent Variable Coefficient

𝐴𝐸𝑋_𝑡−1 0.070 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡−1 0.229***

(0.074) (0.071)

Cons. Conf. 0.054* Cons. Conf. 0.034

(0.033) (0.033)

Prod. Conf. -0.092 Prod. Conf. -0.118

(0.098) (0.101)

Ind. Prod. 0.238 Ind. Prod. 0.232

(0.203) (0.209)

C 0.983 C 0.947*

(0.507) (0.520)

R-squared 0.030 R-squared 0.069

Adjusted R-squared 0.010 Adjusted R-squared 0.049

Regression 3

Independent Variable Coefficient Independent Variable Coefficient

𝐴𝐸𝑋_𝑡−1 0.084 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡−1 0.227*** (0.074) (0.071) 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡−1 0.064** 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓.𝑡−1 0.038 (0.033) (0.033) 𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓.𝑡−1 -0.180* 𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓.𝑡−1 -0.189* (0.099) (0.102) 𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑._𝑡−1 -0.063 𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑._𝑡−1 0.014 (0.204) (0.210) C 1.194** C 1.086** (0.508) (0.523) R-squared 0.033 R-squared 0.074

Adjusted R-squared 0.012 Adjusted R-squared 0.054

This table shows the results of the three OLS regressions. AEX is the left column, Midcap the right column. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage. Lagged values are indicated by t-1. The standard error is shown in parentheses. The OLS regressions are based on 188 observations spanning January 1994 to December 2009.

26 C. Long term sentiment effects

It is interesting to see the long term effects of consumer and producer confidence on the Midcap and AEX. For instance, although coefficients of consumer confidence might differ between AEX and Midcap, the long term effect might show that the lower coefficient has a larger influence than the higher coefficient. To examine this long term effect, I rewrite part of the equation of the third Midcap OLS regression previously used:

(4) 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡 = 0.038 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡−1+ 0.227 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡−1

To

(5) 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡 − 0.227 𝑀𝑖𝑑𝑐𝑎𝑝_𝑡−1 = 0.038 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡−1

To the long term effect on Midcap:

(6) (1 − 0.227) 𝑀𝑖𝑑𝑐𝑎𝑝𝐿𝑇 = 0.038 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓.𝑡−1

Ending with

(7) 𝑀𝑖𝑑𝑐𝑎𝑝𝐿𝑇 = _(1−0.227)0.038 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓.𝑡−1

(8) 𝑀𝑖𝑑𝑐𝑎𝑝_𝐿𝑇 = 0.05 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡−1

This shows the long term coefficient of consumer confidence on the Midcap is 0.05. If I repeat this for the AEX and consumer confidence, the long term effect shown by the coefficient is 0.07. This indicates that the long term effect of consumer confidence is stronger for the AEX than the Midcap. For producer confidence, the long term coefficients are -0.24 for Midcap and -0,20 for the AEX. These results indicate that the long term effects of consumer confidence are slightly stronger for AEX in comparison to Midcap. On the other hand, producer confidence has a slightly stronger negative influence on the Midcap in comparison to the AEX.

D. Vector Auto Regression

27 symmetrically and it includes for each variable an equation explaining its evolution based on its own lags and the lags of all the other variables in the model. The VAR is structured by the following general equation:

(9) 𝑌_𝑡 = 𝑐 + 𝑝_𝑖=1𝛷_𝑖 𝑌_𝑡−𝑝+𝜀_𝑡

Where 𝑌𝑡 = 𝐴𝐸𝑋𝑡 𝑀𝑖𝑑𝑘𝑎𝑝𝑡 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓.𝑡𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓.𝑡𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑.𝑡, the column

vector of variables under consideration and 𝛷_𝑖 is a matrix of coefficients. Also, 𝑐 is a

constant, 𝑝 is the number of lags and 𝜀𝑡 is the random error term. Information criterion

models are employed to determine the appropriate lag length time. According to the Akaike, Schwarz and Hannahn-Quinn information criterion, the appropriate number of lags for our sample is 1. Likelihood ratio tests indicate a longer lag time. However, for parsimony reasons I will set the number of lags equal to 1. AR roots tests shows that the VAR is stationary, with root values being lower than 1 and no roots outside the unit circle. Additionally, the Augmented Dickey Fuller test rejects the null hypothesis of a unit root. In table 5 on the next page one can see the results of my VAR analysis.

28 Table 5

Results VAR analysis

Independent variable

Dependent variable

AEX Midcap Cons Conf. Prod. Conf Ind. Prod

𝐴𝐸𝑋𝑡−1 -0.021 0.129 0.191*** -0.031 -0.020 (0.136) (0.140) (0.076) (0.040) (0.046) 𝑀𝑖𝑑𝑐𝑎𝑝𝑡−1 0.117 0.123 0.028 0.086*** 0.032 (0.129) (0.133) (0.071) (0.038) (0.044) 𝐶𝑜𝑛𝑠. 𝐶𝑜𝑛𝑓._𝑡−1 0.064** 0.035 0.968*** 0.018** 0.002 (0.033) (0.033) (0.018) (0.009) (0.011) 𝑃𝑟𝑜𝑑. 𝐶𝑜𝑛𝑓._𝑡−1 -0.173** -0.192*** 0.011 0.922*** 0.035 (0.099) (0.102) (0.055) (0.029) (0.034) 𝐼𝑛𝑑. 𝑃𝑟𝑜𝑑._𝑡−1 -0.063 0.009 0.095 -0.038 -0.315*** (0.204) (0.210) (0.113) (0.060) (0.070) C 1.194*** 1.08*** -0.283 0.112 0.111 (0.510) (0.524) (0.283) (0.150) (0.174) R-squared 0.033 0.078 0.967 0.912 0.106 Adj. R-squared 0.012 0.053 0.966 0.910 0.087

This table shows the result of the VAR. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage. The standard error is shown in parentheses. Lagged values are shown as t-1. The VAR is based on 188 observations spanning January 1994 to December 2009.

P-values ***:0.01% **:0.05% *:0.1%

Therefore, if the AEX is doing well, it indicates that those multinational companies are doing well. There is no real connection or effect to more nationally oriented Dutch producers (confidence). This line of reasoning is reinforced by the positive and significant relation between the Midcap and producer confidence (p < 0.01). The Midcap features companies which are medium sized, and those are generally more nationally oriented than AEX companies. Good returns on the Midcap make producers more positive towards the future, again reinforcing the bandwagon effect. Additionally, the missing positive relation between Midcap and consumer confidence can be explained by the same reason that AEX is

29 The positive and significant (p < 0.05) relation from consumer confidence to AEX, was already shown by my previous regression. This again indicates that the positive effects of consumer confidence last longer than expected. The lagged effect of consumer confidence on itself shows a highly positive and significant relation (p < 0.01) which strongly indicates that optimistic consumers tend to get more optimistic by past values of consumer confidence. This effect spills over to producer confidence where there is a slightly positive and reasonable significance (p < 0.05) effect of a lagged value of consumer confidence. This indicates that producers get more optimistic by higher levels of consumer confidence. The explanation can be found in the fact that the market producers are generally aiming at are consumers, and optimism on their side is a predictor for producer success. Producers are probably well aware that the consumer confidence index for instance contains questions regarding the feasibility of large purchases and the consumer’s current financial situation. The lack of a significant relation between consumer confidence and Midcap has already been discussed in my previous regression results. The result for producer confidence shows that 1 lag of producer confidence has a negative and reasonable significant (p < 0.05) relationship with AEX. The same relation is seen with the Midcap, however slightly stronger and more significant (p < 0.01).

The final variable, industrial production is surprisingly is not significantly related to any measure except itself. First of all, against my expectations of having positive influences on AEX or Midcap it has no significant effect. Apparently fundamental figures measured as industrial production have no significant effect on the stock markets. Maybe the inclusion of fundamental data is so gradual that no effect can be distinguished. Another possible explanation is that industrial production is not a true sentiment measure as in opinion or belief, and therefore no market misvaluation takes place due to optimism. Therefore, no correction is needed and thus no visible effect. Also puzzling is the strong negative relation of industrial production with itself. Apparently industrial production is negatively related to its own lag (p < 0.01). This might be explained by a higher production at time t would usually mean increasing stock, thus lowering production at t+1.

E. Impulse response functions

30 shock in one variable while keeping the effect of other variables initially constant. Since impulse responses provide complicated interactions of the variables, confidence bands are constructed around the mean response. Responses are considered statistically significant at the 95% confidence level when the upper and lower bands carry the same sign. A simple form of impulse responses is a shock of one standard deviation of the residual. Table 6 shows the residual correlation matrix of the VAR equations along with the standard deviations of the residuals.

Table 6

Residual correlation matrix with standard deviations

AEX Midcap Cons. Conf. Prod. Conf. Ind. Prod.

AEX (6.12)

Midcap 0.85 (6.29)

Cons. Conf. 0.16 0.20 (3.39)

Prod. Conf. 0.20 0.20 0.10 (1.80)

Ind. Prod. 0.09 0.09 -0.07 0.01 (2.09)

Table shows the residual correlation matrix with variable standard deviations in parentheses. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage.

If I use this residual standard deviation, the effect can be illustrated following the coefficients in the previous VAR in table 5. For instance, the shock to AEX produced by producer confidence is:

(10) 𝐴𝐸𝑋_𝑡=1 = − 0.173 𝜎_{𝑃𝑟𝑜𝑑 .𝐶𝑜𝑛𝑓 .}

When we use the standard deviation of the residual of producer confidence provided in table 6 (𝜎 = 1.80), the effect would be -0.31. The impulse response shows the effects of such types of shocks over time for my three sentiment measures. If the initial shock of 1.80 is given in period 1, the next period will show a decline of the AEX index of -0.31, as can be seen in table 9 of the appendix. The results follow the general conclusions previously made about the (lagged) negative impact of producer confidence and the (lagged) positive impact of consumer confidence. The graphs provide a good visual impression of my previous results.

31 (see table 6) does not provide zero correlation. In fact, the previous method fails to pick up the initial rise in variables due to correlation. The residual correlation matrix shows us that a one standard deviation shock of producer confidence produces an immediate effect to the AEX of:

(11) 𝐴𝐸𝑋 _𝑡=1 = 𝜌_{𝐴𝐸𝑋 ,𝑃𝑟𝑜𝑑 .𝐶𝑜𝑛𝑓 .}(𝜎_𝐴𝐸𝑋)

(12) 𝐴𝐸𝑋𝑡=1 = 0.2(6.12) = 1.224

32

Table 7

Generalized impulse response functions of AEX and Midcap*

Response: AEX Response: Midcap

One S.D. Cons. Conf. Innovation

One S.D. Prod. Conf. Innovation

One S.D. Ind. Prod. Innovation

This table shows the generalized impulse responses of AEX and Midcap to a one time shock of one standard deviation of consumer confidence, producer confidence and industrial production. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage.

Notes: The dashed lines on each graph represent the upper- and lower-95% confidence bands. When the upper- and lower bounds carry the same sign the response becomes statistically significant.

*On each graph, ‘percentage returns’ are on the vertical and ‘horizon’ is on the horizontal axis.

33

Table 8

Generalized cumulative impulse response functions of AEX and Midcap

Cumulative response: AEX Cumulative response: Midcap

One S.D. Cons. Conf. Innovation

One S.D. Prod. Conf. Innovation

One S.D. Ind. Prod. Innovation

This table shows the generalized accumulated impulse responses of AEX and Midcap to a one time shock of one standard deviation of consumer confidence, producer confidence and industrial production. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage.

Notes: The dashed lines on each graph represent the upper- and lower-95% confidence bands. When the upper- and lower bounds carry the same sign the response becomes statistically significant.

*On each graph, ‘percentage returns’ are on the vertical and ‘horizon’ is on the horizontal axis.

34 The results of my response functions are largely as expected, and show the initial rise in stock returns. First of all, consumer confidence has a positive and significant influence on both the AEX and the Midcap the first two periods. This confirms my previous regressions which also showed increases even with a one month delay. My assumption that the return the following period(s) is negative is however faulty. Following these responses, both the AEX and Midcap do not correct the initial positive return with negative return the next period. At least, my current model does not show a correction. Furthermore, as can be seen in table 8, the net effect is positive and significant for 10 periods of AEX and 9 for Midcap. The cumulative graph indicates that the net effect continues to rise.

Producer confidence provides a different picture, showing significant and positive effects in period 1, and then reverting to negative in the second and following periods for both the AEX and Midcap. This was expected, however the significance is lacking after the first period. The cumulative table shows that producer confidence eventually goes into a net negative effect around period 6, where the first two periods are significant. The interesting point is that producer confidence shows very similar results to both the AEX and Midcap were both the generalized impulse and cumulative responses are almost exact matches. Finally, when one looks at industrial production the effect of a fundamental variable is visible, showing a onetime rise and stabilization henceforth, however not significant.

35 partially correct. There is an initial positive effect, where the aggregate market indices show positive results following producer and consumer confidence. Additionally, the correcting motion of the stock market is visible in my regressions, the VAR and impulse response functions for producer confidence. However, my model failed to pick up the correcting motion for consumer confidence, indicating a longer positive effect, against expectations.

V. Conclusion

A. Conclusion

In this study I investigate the contemporaneous and lagged influences of consumer confidence, producer confidence and industrial production growth on the AEX and Midcap. In addition, I give insight into the underlying relations that these variables have amongst themselves. My first OLS regression indicates that the contemporaneous relation between sentiments and stock returns is almost negligible, because only consumer confidence shows a positive relation to the AEX. My other sentiment measures fail to produce significant results to the AEX and the Midcap in contrast to my expectations. To test the relationships further, the second OLS regression added a lagged value of the dependent variables, the AEX and Midcap, into the equation. It has no visible effects on the contemporaneous results for AEX. The Midcap on the other hand, shows that the addition of its own lag is significant and relatively strong.

My third OLS regression was employed to test for the correcting feature of the market, where periods of high sentiment are followed by lower returns. I add lagged values of my sentiment measures. In the case of AEX as a dependent variable, it shows that consumer confidence and producer confidence become more significant. Apparently, the lagged effect of producer confidence becomes visible after one period with OLS regression. The lagged effect shows a negative relation, implying a correcting motion by the market. On the other hand, consumer confidence shows a positive relation for its lagged value. This might suggest that the positive sentiment holds for longer periods than expected. Producer confidence also becomes significant and negative for the Midcap as a dependent variable, indicating a similar effect as for the AEX. Again, the Midcap returns as a dependent variable are significantly driven by their own lag.

36 due to the fact that the AEX is under high media coverage. Good returns in the AEX make the news, in turn raising consumer confidence. Surprisingly, good returns on the AEX do not spill over to producer confidence increases. This is likely due to the fact that the AEX mostly contains companies competing internationally. In comparison, the Midcap is mostly comprised of more nationally oriented companies. Therefore, if the AEX is exhibiting high returns, those internationally companies are presumed to be doing well. This has however no direct influence on producers and their confidence in the Netherlands. The Midcap on the other hand does have a positive relation to producer confidence, indicating that producers do get more optimistic if the Midcap (with more local firms) exhibits high returns. This backs up the assumption that the distinction of ‘international index’ and ‘national index’ is likely.

The VAR supports the positive effect that consumer confidence has on the AEX. This unexpected relation again is consistent with previous results. Since it has successfully been used as an individual investor sentiment measure before, it might indeed be a true measure of investor optimism. Or differently, individual investors possibly use it as an investing ‘advice’. The implication is that the effects do not wear off as quickly as anticipated. In addition, it has a strong influence on itself, indicating that consumers become more positive by their own previous mindset. Finally, consumer confidence has a positive effect on producer confidence, indicating that producers get more positive due to increasing consumer confidence. This may also be explained by the fact that consumers are the main customers of producers. A raise in their optimism predicts higher spending thus higher revenues for producers. This is however, not in line with the idea that producer confidence is generally assumed to signal changes in the economy faster than consumer confidence.

37 Finally, the generalized impulse response function generated out of the 1 lag VAR model provides support for my final conclusion. Firstly, consumer confidence has a positive effect for a longer period than producer confidence. Furthermore, it seems this effect is not corrected by the market, at least not showing in my model. Consumer confidence has a positive effect on AEX in the first three periods. With regards to the Midcap, the positive effect decreases and is significant for the first two periods. The cumulative effect shows a net positive and significant effect for 10 periods for the AEX. The net positive effect is significant for 9 periods for the Midcap. The fact that the effect is longer and stronger for the AEX is surprising. Possibly attributable to a higher number of individual investors, that trade on data like consumer confidence. Secondly, the generalized impulse responses for producer confidence does show a significant initial rise in stock return, a feature which my other analyses fail to show. The effect of producer confidence is positive for the first period and then reverts to negative for both the AEX and Midcap. Apparently the market does have a corrective feature regarding producer confidence. The cumulative table shows that the effect goes into negative after period six for both the AEX and Midcap. However, these effects loose significance after period 2. Finally, industrial production as a fundamental sentiment measure fails to provide a significant result in the impulse response functions.

B. Limitations and suggestions for further research

First of all, the use of monthly sentiment data for something as volatile and fast adapting as a stock market is less ideal. When I look at the short lived positive sentiment effects produced by producer confidence, the question is for how long these effects last. My first regression fails to pick up the positive effect. And only through the use of generalized impulses the positive effect is visible. While using monthly data it was more or less expected to have difficulties showing the positive effect. Therefore, it may be interesting to investigate how long these effects last in more precise manner. Even if this would not add to existing theory, it might provide an interesting result for practitioners. My research has already shown that for producer confidence the market corrects within one period. A suggestion for further research would be to measure the lagged effects of producer confidence from when it is made public to the next month with daily stock returns. This would provide an interesting view in the initial positive effects, left mostly uncovered in my analyses’.

38 consumer confidence might be interesting. Preliminary analysis not shown in this paper

suggests that up until period 5, the effects are rising and positive for the AEX. This

39

VI. Appendix

Table 9

Impulse response functions of AEX and Midcap*

Response: AEX Response: Midcap

Residual One S.D. Cons. Conf. Innovation

Residual One S.D. Prod. Conf. Innovation

Residual One S.D. Ind. Prod. Innovation

Response of AEX and Midcap to a one time shock of residual one standard deviation of consumer confidence, producer confidence and industrial production. AEX is the aggregate market index monthly total return. Midcap is the aggregate market index monthly total return. Cons. Conf. is the consumer confidence index, Prod. Conf. is the producer confidence index and Ind. Prod. Is the industrial production index’ monthly growth percentage.

Notes: The dashed lines on each graph represent the upper- and lower-95% confidence bands. When the upper- and lower bounds carry the same sign the response becomes statistically significant.

*On each graph, ‘percentage returns’ are on the vertical and ‘horizon’ is on the horizontal axis.

40

VII. References

Baker, Malcolm, and Jeffrey Wurgler, 2006, Investor sentiment and the cross-section of stock returns. Journal of Finance 61, 1645–1680.

Barro, Robert J., 1993 The Stock market and investment, Review of financial studies 3(1), 15-31.

Beja, Avraham, 1977, The Limits of Price Information in Market Processes, Working Paper 61, Research Program in Finance, University of California, Berkeley

Brown, Gregory W., and Michael T. Cliff, 2004, Investor sentiment and the near-term stock market, Journal of Empirical Finance 11, 1–27

Brown, Gregory W., and Michael T. Cliff, 2005, Investor sentiment and asset valuation, Journal of Business 78, 405-439..

Carhart, Michael M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52(1), 57-82.

Chan, N., Kan, R. and Miller, M., 1993, Are the discounts on closed-end funds a sentiment index?, Journal of Finance 48, 795–800.

DeLong, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, 1990, Noise trader risk in financial markets, Journal of Political Economy 98, 703–38. DeLong, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, 1991, The survival of noise traders in financial markets, Journal of Business 64, 1–19. Elton, Edwin J. , Martin J. Gruber, and Jeffrey A. Busse, 1998, Do investors care about

sentiments?, Journal of Business 71, 477–500.

Fama, Eugene F. and Kenneth R. French, 1989, Business conditions and expected returns on stocks and bonds, Journal of Financial Economics 25, 23–49.

Fama, Eugene F. (1970) Efficient capital markets: a review of theory and empirical work, Journal of Finance 25, 383–417.

Fama, Eugene F., 1981, Stock Returns, Real Activity, Inflation, and Money, American Economic Review 71, 545-565.

Fama, Eugene F., 1990, Term structure forecasts of interest rates, inflation, and real returns, Journal of Monetary Economics 25, 59–76.

Fisher, Kenneth L., and Meir Statman, 2003. Consumer confidence and stock returns, Journal of Portfolio Management 30, 115–128.