The Determinants of Consumer
Confidence in the Netherlands
Thesis presented by F. Drexhage
S1703463
Supervisor: Prof. Dr. J. de Haan
Master of Economics
Abstract
This paper focuses on the determinants of Dutch consumer confidence. A research conducted by Stokman (2005) showed that 88% of the changes in consumer confidence is explained by factors such as unemployment, the housing price, GDP growth and the stock market. The current paper raises several questions with regard to the research question conducted by Stokman (2005). For instance, what happens to Stokman his model if more recent data is included, can additional determinants of consumer confidence be found and to what extend does an economic recession influence the level of consumer confidence in the Netherlands? The empirical analysis in this paper uses the research conducted by Stokman (2005) as a starting point. In an ordinary least squares model setting, determinants and additional determinants of consumer sentiment are included to determine whether the model of Stokman still holds when including more recent data and to determine whether additional determinants of consumer sentiment can be found. After concluding which variables determine the level of consumer sentiment, rolling regressions are conducted to determine whether the parameters of the model are stable over time. Ultimately, the direct and interaction effect of an economic recession on consumer sentiment are tested.
In this paper, first of all the history of the consumer confidence indices of both the United States and the Netherlands is reviewed. Secondly, an extensive overview of the literature is given. The predictive power of consumer sentiment with regard to consumer expenditure is discussed, followed by economic and financial determinants and non-‐economic and non–financial determinants of consumer confidence as set out in the literature. Thirdly, the model is described and lastly, the results are discussed.
Keywords
Consumer Confidence Netherlands
Content
Abstract ii
Keywords iii
List of Tables v
List of Figures vi
1. Introduction 1
2. Consumer Confidence 2
2.1 Differences between the Michigan’s Index of Consumer Sentiment 4
and the Conference Board Survey
2.2 Dutch Consumer Confidence 4
2.3 Differences CBS consumer confidence and EC consumer confidence 5
3. Literature Review 8
3.1 The predictive power of consumer confidence 8
3.2 Determinants of consumer confidence 10
3.2.1 Economic and financial variables 10
3.2.2 Non-‐economic and non-‐financial variables 12
3.2.3 Conclusion on variables 14
4. Model 14
5. Data 16
5.1 Additional variables 17
6. Results 20
6.1 Results rolling regression 26
7. Conclusions and Recommendations 29
List of Tables
Table 1. Michigan Survey; US consumer confidence 6
Table 2. Conference Board Survey; US consumer confidence 7
Table 3. Survey CBS; Dutch consumer confidence 7
Table 4. Survey EC; Dutch consumer confidence 7
Table 5. Summary of hypotheses 19
Table 6. Descriptive statistics 19
Table 7. Correlations 21
Table 8. Regression Results 24
Table 8, continued. Regression Results 25
List of Figures
Figure 1. Dutch consumer confidence; difference between EC and CBS indicator 6 Figure 2. Consumer confidence 1986-‐2012; difference in volatility between 23 Monthly and annual data
Figure 3A. Results Rolling Regression, Unemployment 27
Figure 3B. Results Rolling Regression, Stock Market 27
Figure 3C. Results Rolling Regression, Housing Price 27
Figure 3D. Results Rolling Regression, Yield Spread 28
Figure 3E. Results Rolling Regression, Temperature 28
1. Introduction
“Consumer confidence at a historical low” “Consumer confidence decreased again” “Consumers are again a lot more pessimistic”
(source: nu.nl, translated from Dutch)
Contemporary news headlines demonstrate that consumer confidence has become a popular topic of daily news. In particular during an economic downturn the confidence or sentiment of the consumer receives a lot of attention from and in the mass media. Any significant change or lack of consumer confidence is considered to be a very influential sign of a nearby turning point or prolongation of the trough (Vuchelen, 2004). But despite the popularity of the index, there is little consensus about its ability to collect information on consumer spending that is not already captured by economic fundamentals (Bram & Ludvigson, 1998).
The current paper is composed of three main research questions, the first question is whether the increased access to information influences the relationship between economic variables and consumer confidence, and whether the model of Stokman (2005) can be maintained if more recent data are included. The second question that is raised is whether there are any additional variables, apart from the variables that Stokman included, that influence the way consumers value the economic future and whether the coefficients change over time. And thirdly, do times of economic recession affect the level of consumer confidence in comparison to periods of economic prosperity?
Section 2 of the paper starts by giving an overview of how consumer confidence is measured and reported in both the United States and the Netherlands. In Section 3 previous literature is discussed, and Section 4 reviews the econometric model that will be used in this paper. The data used is given in Section 5, which is followed by the results in Section 6. To end, section 7 presents the conclusions and recommendations for further research.
2. Consumer Confidence
Consumer confidence can be defined as the public its confidence in the economy. Measuring consumer confidence can be helpful to explain and predict consumer spending. The first country that started to index consumer sentiment was the United States. In the late 1940s, Professor George Katona of the University of Michigan initiated the Michigan Index of Consumer Sentiment as an annual survey. This survey became in 1952 a quarterly survey, and later in 1978 a monthly survey. The index has several objectives:1
-‐ To assess consumer attitudes on the business temperature, personal finance, and spending -‐ To try to gain an understanding of, and to forecast changes in, the national economy -‐ To measure the economic expectations and possible future spending behavior of consumers -‐ To get an insight of the consumer’s level of optimism or pessimism
Each month there are at least 500 respondents selected through a random digit dialing telephone sampling. A commercially available list of potential phone numbers is used, stratified by location and degree of urbanization. According to Franses and Van Oest (2008), stratified sampling may result in a more representative set of respondents and less sampling noise than normal sampling. Out of the 500
respondents, 60% is drawn using the procedure described above, while 40% consists of returning respondents who were interviewed six months ago. The respondents are requested to determine for each of the five questions (see table 1 below) whether they have a negative, neutral or positive attitude. Subsequently, the five questions are averaged. The final index is a linear transformation of the difference between the percentage of positive respondents and the percentage of negative respondents . A hundred points are added to this difference, the index is scaled by its base year 1966, and a small correction factor is added to account for past sample design changes. The Michigan Index of Consumer Sentiment is given by
However, the Michigan index is not the only CC index in the United States; its counterpart is the index of the “Conference Board Survey” (Bram & Ludvigson, 1998).
The Conference Board, which is an independent economic research organization, started surveying consumer confidence in 1967. Its surveys were initially sent out on a bimonthly basis, but from 1977 on this was done on a monthly basis and nowadays 5,000 representative families receive a questionnaire every month (see table 2). These families are selected out of a pool of 120,000 preselected families. Out of the 5,000 families, about 70% returns the five-‐questions survey. Each question’s positive responses are divided by the sum of its positive and negative responses, also known as the “relative value”. The relative value for each question is then compared against each relative value in 1985. The year of 1985 is used as the benchmark year because 1985 was neither a peak nor a trough (Ludvigson, 2004). The index values for all five questions are then averaged to form the Conference Board Consumer Confidence Index. The average of index values for question one and three form the Present Situation Index. The average of index values for question two and four and five form Expectations Index. The percentage of positive respondents is divided by total percentage of positive and negative respondents + .
2.1 Differences between the Michigan’s Index of Consumer Sentiment and the Conference Board Survey
Some differences exist between the survey questions of the Conference Board and those of the Michigan Index. The Conference Board poses many, very specific labor market oriented questions. As a result, the present situation component of the Conference Board closely tracks labor market conditions, such as the nation’s unemployment rate and the growth in payroll employment. In contrast, the Michigan Index asks respondents to comment on the advisability of big-‐ticket household purchases and to assess changes in their own financial situation. Although this latter question is about the personal financial situation of the respondent, it does not pose direct questions about changes in the employment outlook. Therefore, Michigan’s present conditions component is less closely tied to labor market conditions and tends to reflect the recent changes in the economy instead of the level of economic activity.
Although the financial markets and the business community closely follow both indexes, virtually all published academic research focuses on the Michigan index; probably because of its longer history.
2.2 Dutch Consumer Confidence
Also for the Netherlands there are two different consumer indices. One is published by Statistics Netherlands (hereafter, CBS) and the other one is published by the European Commission (hereafter, EC). The Dutch survey data, collected by the CBS, are available from 1972 onwards. In the early years of the Dutch consumer index, the surveys were sent out to some 1000 households that had to fill in the questionnaire by hand, but later on this was done over the phone. The survey consists of five questions, of which two are about to the future situation, two questions refer to the past and one question addresses the present (see table 3). For each question, the respondent has to indicate whether he has a negative, neutral or positive attitude. The corresponding percentages of negative, neutral and, positive respondents are averaged over the five questions. Subsequently, the percentage of negative respondents (denoted ,) is subtracted from the percentage of positive respondents (denoted ) to arrive at the Dutch CBS confidence index (Jansen & Nahuis, 2003);
Consumer confidence as determined by the CBS is different from the EC indicator due to the fact that there is a difference in the definition, calculations and data that are used by the CBS to determine consumer confidence.
The EC sends a monthly survey to some thousand persons. The surveys are harmonized which implies that the questionnaires are identical for all European countries. The EC survey of consumer sentiment only consists of four questions (see table 4). With respect to the calculation of consumer confidence, the EC survey respondents state that the economic situation will be a lot better ( ), a little better ( ), the same, a little worse ( ) or a lot worse ( ). The values of and are respectively 1 and -‐1 whereas the values of and are respectively ½ and -‐½. The consumer confidence is defined as the average of the five questions of the survey (Jansen and Nahuis, 2003);
2.3 Differences CBS consumer confidence and EC consumer confidence
A comparison between the levels of consumer confidence in the CBS index and those given in the EC index shows that there are significant differences among them (see figure 1). This is due to several dissimilarities. The first dissimilarity is that whereas the EC indicator of consumer confidence only contains questions related to future situations, the CBS indicator takes into account questions concerning the past, the present and the future. The EC indicator, however, covers a greater number of subjects than the CBS indicator does, such as unemployment and the ability to save. Secondly, the data show that there are differences in the level of consumer confidence due to differences in measurements. In particular in periods of economic downturns the EC indicator shows higher levels of consumer sentiment than the CBS indicator does (see figure 1). A possible explanation for this dissimilarity is that the EC survey contains a question about the ability to save, as even in times of economic downturns people will be able to save. This particular question might thus explain that the EC indicator shows higher levels of consumer confidence than the CBS indicator does, especially in times of economic downturns.
Figure 1: Dutch consumer confidence; difference between EC and CBS indicator
Table 1: Michigan Survey; US consumer confidence
Q1) Do you think now is a good or bad time for people to buy major household items? [good time to buy/uncertain, depends/bad time to buy]
Q2) Would you say that you (and your family living there) are better off or worse off financially than you were a year ago? [better/same/worse]
Q3) Now turning to business conditions in the country as a whole—do you think that during the next twelve months, we’ll have good times financially or bad times or what? [good times/uncertain/bad times]
Q4) Looking ahead, which would you say is more likely—that in the country as a whole we’ll have continuous good times during the next five years or so or that we’ll have periods of widespread unemployment or depression, or what? [good times/uncertain/ bad times]
Table 2: Conference Board Survey; US consumer confidence
Q1) How would you rate present general business conditions in your area? [good/normal/bad]
Q2) What would you say about available jobs in your area right now? [plentiful/not so many/hard to get] Q3) Six months from now, do you think business conditions in your area will be [better/same/worse]? Q4) Six months from now, do you think there will be [more/same/ fewer] jobs available in your area? Q5) How would you guess your total family income to be six months from now? [higher/same/lower]
Table 3: Survey CBS; Dutch consumer confidence
Q1) Do you think that the general economic situation in our county has improved, worsened or stayed the same?
Q2) What is your opinion about the next twelve months? Will the general economic situation in the Netherlands improve, worsen or stay the same?
Q3) Do you think now is a good or bad time for people to buy major household items?
Q4) Did the financial situation in your household improve, worsen, or stay the same during the last twelve months?
Q5) What do you expect from the financial situation of your household? Will it improve, worsen or stay the same during to next twelve months?
Table 4: Survey EC; Dutch consumer confidence
Q1) How do you think the financial position of your household will change over the next 12 months? [a lot better/a little better/ equal/a little worse/a lot worse]
Q2) How do you think the general economic situation in this country will change over the next 12 months? [a lot better/a little better/ equal/a little worse/a lot worse]
Q3) How do you think the level of unemployment in the country will change over next 12 months? [a lot better/a little better/ equal/a little worse/a lot worse]
Q4) Over the next 12 months, how likely are you to be able to save any money? [a lot better/a little better/ equal/a little worse/a lot worse]
3. Literature Review
Several studies on consumer confidence have been published over the past decades. A clear distinction can be made between two types of studies: on the one hand there are studies that are interested in the relation between consumer confidence and consumer spending, and on the other hand there are studies that determine the variables that influence the level of consumer confidence. Both types of studies are discussed below.
3.1 The predictive power of consumer confidence
Most studies on consumer confidence are dedicated to the predictive power of consumer confidence, such as the relationship between consumer confidence and consumer expenditure (people spend more when consumer confidence is high but spend less when consumer sentiment is low). One of the reasons that scholars are interested in the predictive power of consumer confidence on consumer expenditure is the fact that consumer expenditure is by far the most important single item of aggregate demand. Although private consumption belongs to the more stable economic variables, unexpected shifts occur which lead to important forecast errors in economic growth (Vuchelen, 2004). As business cycle analysts look for improvements in the quality of their consumption forecasts, they are interested in consumer confidence because consumer confidence can have forecasting properties for private consumption. The question remains, however, whether consumer confidence captures information on consumer spending that is not already captured by economic fundamentals. Leeper (1992) argues that consumer sentiment may have predictive power for consumer spending because consumer surveys are made available on a timelier basis than other economic indicators, such as data on income and consumption. However, he also points out that financial market indicators are available on an almost continuous basis and may contain much of the same information captured by consumer sentiment. Accordingly, Leeper concludes that consumer attitudes are only weakly correlated with variables, such as unemployment and industrial production, once financial indicators are included (Ludvigson, 2004). Ludvigson states that consumer confidence only contains modest information about the future path of consumer expenditure in comparison to other fundamentals. The conclusion of Benjanuvatra (2009) is even more radical; she concludes that there is no relation at all: in her view, consumer confidence variables do not determine consumer expenditure. Several other scholars, including Acemoglu and Scott (1994) and Fan and Wong (1998), strongly agree with Ludvigson and Benjanuvatra.
the permanent income hypothesis. This theory states that consumption is determined by the income of an individual over his or her entire lifetime, and for that reason the expenditure of consumers is affected by transitory changes and not by consumer confidence, assuming that consumer confidence cannot capture these transitory changes.
Still, there are authors who argue that consumer sentiment does have predictive powers. Stokman (2005), for example, states that the correlation coefficient of consumer confidence and spending is 0.88. This high correlation suggests that consumer confidence does have some predictive powers. Besides Stokman, there are a number of other authors that have found a significant relationship between consumer confidence and consumer expenditure, such as for instance Katona (1975) and Bram and Ludvigson (1998). The aim of the research of Bram and Ludvigson (1998) was to compare the predictive power of the Conference Board CCI and the University of Michigan CCI. According to their research, these two indexes do have predictive powers. Desroches and Gosselin (2002) assessed the usefulness of consumer confidence indexes in predicting aggregated consumer spending in the United States. They concluded that consumer confidence has significant predictive powers especially in periods of high economic or political uncertainty, even after controlling for other determinants.
It can be said that the literature is in general skeptical about the predictive powers of consumer confidence, even though there is some evidence that suggests that consumer confidence is able to explain future consumption expenditure. An explanation for the various dissimilar outcomes can be that the papers conducting research not only include different indicators as controls but also focus on diverse forecasted economic outcomes. For example, academic research that focuses exclusively on the predictive power of the Michigan Index, generally does not find a significant relationship between consumer attitudes and future real economic activity (Bram & Ludvigson, 1998). However, if the Michigan Index is replaced by the Conference Board index, consumer confidence appears to be able to predict consumer expenditure better. These and other differences in outcome are the result of dissimilarities among surveys. One of the differences between the Michigan Index and the Conference Board index is that The Conference Board survey contains more specific questions about job perspectives. It seems that consumers spend more when they feel good about future job prospects (when the Conference Board index is high) than when they think business conditions are favorable, which is discussed in the survey of the Michigan Index.
estimating which variables in fact influence consumer confidence, as they hope to be able to influence the economic situation of a country through consumer confidence.
3.2 Determinants of consumer confidence
According to numerous articles published on the determinants of consumer confidence, two types of determinants exist. On the one hand there are economic and financial determinants and on the other hand there are non-‐economic or non-‐financial determinants. One of the main determinants of consumer confidence mentioned in a great number of papers is the lagged level of consumer confidence, which indicates that the dependent variable is highly autoregressive. When the dependent variable is specified as the change in consumer sentiment, the explained part of total variation is certainly not impressive (Vuchelen, 2004).
3.2.1 Economic and financial variables
Ferreira et al. (2005) investigated the relation between the yield spread, the difference between the long-‐run and short-‐run interest rate, and the variability of consumer confidence in Europe. They found that the European yield spread explains 93.7% of the variability of the economic sentiment indicator.
Jansen and Nahuis (2002) studied the influence of stock prices on consumer confidence in 11 European countries. They concluded that the stock market has a significant effect on consumer confidence. Jansen and Nahuis (2002) argue that higher stock prices may boost confidence for two reasons. The first reason is that of direct effect; higher stock prices mean higher wealth and therefore greater optimism. The direct effect is related to the traditional wealth effect, although the traditional wealth effect is probably smaller in the Netherlands than it is in the US as fewer Dutch households invest in stocks. Furthermore, if Dutch households do invest in stocks, they do so with a smaller share of their wealth compared to US households. The second reason is an indirect effect; economic agents may interpret higher stock prices as a sign of favorable economic conditions in the future. This indirect effect provides a channel through which equity prices influence the behavior of all consumers, regardless of whether they have a direct stake in the stock market or not (Jansen and Nahuis, 2002). Or as Marshall pointed out:
Stock exchanges are not merely the chief theatres of large business transactions; they are also barometers that indicate the general conditions of the atmosphere of business. (Marshall, 1923, p. 89)
Otoo (1999) investigated this relationship for the US using monthly data in the period from October 1995 until the end of 1997, having access to data about the amounts of stocks owned by consumers. She concluded that, apart from the question whether consumers own stocks or not, the stock market influences consumer confidence and that the influence is therefore due to the indirect effect (Prast, Mosch, & van Raaij, 2005). Beltran Lopez and Durrez (2003) also found a significant explanatory effect of the stock market fluctuations in the evolution of consumer confidence in the US. However, they found that the parameters driving consumer confidence change over time implying that the fluctuations of the stock market became an important factor during the nineties.
economic conditions but, more broadly, the “subjective” state of mind of consumers. Consumer confidence would thus contain information that cannot be deducted from economic and financial variables (Vuchelen, 2004). Desroches and Gosselin (2002) agree that economic and financial variables do not fully explain consumer sentiment. They find in their research that 72% of the variation in consumer confidence in the US can be explained by macroeconomic variables, and that accordingly, the other 28% has to come from non-‐macroeconomic or financial variables.
3.2.2 Non-‐economic and non-‐financial variables
Prast, Mosh and Van Raaij (1995) argue that trust variables are determinants of consumer confidence in the Netherlands. Trust variables are the common denominator for the degree of confidence that Dutch citizens have in their parliament, the integrity of Dutch businesses and the integrity and expertise of financial directors. The authors wondered whether consumer confidence in the Netherlands is influenced by the trust Dutch citizens have in their institutions. The conclusion that can be drawn from their paper is that in particular the confidence citizens have in financial institutions is closely related to consumer sentiment. Prast et al. are not the only scholars arguing that trust variables influence consumer confidence. Fred Bakker for example commented in a Dutch financial newspaper (het Financieel Dagblad) that the decrease in consumer confidence is also influenced by the lack of trust of citizens in the institutions and the governing elites (Prast et al., 2005).
Uncertainty and its influence is another variable that has been discussed in the literature. Vuchelen (2004) included, in addition to the economic variables, uncertainty surrounding future economic conditions as an indicator. Vuchelen defines uncertainty surrounding future economic conditions as experienced by consumers as “the standard deviation of the growth forecasts or as the maximum less minimum predicted rate”. He argues that this approximation makes sense because as mass media tend to highlight divergences between forecasts, consumers are probably aware of this deviation or difference. The underlying hypothesis is that economic uncertainty as predicted by forecasters is transmitted to consumers. By including this uncertainty parameter, the inexplicable part of consumer confidence fell by half.
between the number of economic news items and the level of consumer confidence in the Netherlands; the absence of economic news seemed to have a positive effect on expectations sometimes.2
Franses and Van Oest (2007) find that temperature has a significant influence on the level of Dutch consumer confidence. Besides looking at the determinants of consumer confidence, Franses and Van Oest (2007) came up with a completely different explanation for the inconsistency in consumer confidence. They wondered whether the changes in the monthly consumer confidence are actual changes in the overall confidence of the population or whether these changes can be largely explained by the fact that the panel changes each month. The research by Prast et al. (2005) shows that in particular those people who are female, have low incomes and who receive social security are most pessimistic about their current and future economic situation. In line with the results of Prast et al. it is not surprising that a panel comprised of greater number of low-‐income people results in a lower level of consumer confidence than a panel fewer low-‐income people. Franses and Van Oest conclude that claims about increased or decreased consumer sentiment should be made with care as illustrations show that monthly changes in consumer confidence are not often significantly different from zero.
3.3 Conclusion on variables
The literature discussed above presents a clear overview of the main economic and non-‐economic determinants of consumer confidence resulting from previous research. The most important economic and financial determinants are indicators of income, the stock market index, the interest rate (short, long or real), consumption expenditures, inflation, housing prices, unemployment and the oil price. The most important non-‐economic or non-‐financial determinants are mainly politics, trust variables, uncertainty, temperature and the influence of the media.
4. Model
In order to determine which economic variables influence the level of Dutch consumer confidence, Stokman (2005) uses a simple regression, as demonstrated in equation 1. The dependent variable is the level of consumer confidence at time t. The explanatory independent macroeconomic variables are the change in the unemployment rate, the change in GDP, the growth in the stock market and, the growth rate of the housing prices.
2 Media will unfortunately not be considered as an additional explanatory variable as the amount of work it takes to categorize all media from 1986 until 2012 goes beyond the scope of this paper.
(1)
Where is level of consumer confidence in the Netherlands at time t, is the change in the unemployment rate at time t, is the growth rate of GDP in at time t-‐1, is the growth rate of the Dutch stock market at time t, is the growth in housing prices at time t, α, β, γ, δ and θ are parameters and finally ε is the random error term, which is expected to have zero mean, constant variance σ2 and is uncorrelated over time. For each of the independent variables, their long-‐term
average growth is subtracted from the growth at time t in order to see the growth rates in perspective (i.e. a monthly growth in the housing price of 0.3% might seem large, however taken into account that the average growth rate of the housing price from 1986-‐2012 was 0.46% a month, a growth percentage of 0.3% is in fact lower than the average). Equation 1 is used to answer the first research question, which is whether more recent data cause the determinants of consumer confidence to vary from the determinants of Stokman (2005). Equation 2 is concerned with the second and the third part of the research question, which is where the original equation 1 includes additional determinants and a dummy variable.
(2)
regression the last month of the time window is dropped out of the time window, while at the same time one month at the beginning of the time frame is included. The number of observations thus stays the same but the time window is moved forward by on month at the time. The difference between a rolling and recursive regression is that a recursive regression has either a fixed starting point or a fixed end point, but no fixed number of observations. The time window thus increases with every regression, it adds one additional month to the existing time frame at the time. Based on the results of Beltran Lopez and Durrez, the rolling regression method is used with a time window of 10 years, a decade.
5. Data
The research of Stokman covers the time period of 1978 until 2005, and in his research, Stokman has used annual data. In order to answer the first research question, whether the results of Stokman (2005) still hold when using more recent data, the sample period will obviously change. The data that are currently available reach until mid-‐2012. When taking a closer look at the data, it turns out that for almost all indicators monthly data are available, except for GDP data. Monthly data is preferred as it provides more observations, and more observations lead to a sounder conclusion. However, as monthly data is not available for every variable until 1978, the period covered in this paper will be 1986 until 2012. GDP is only measured four times a year, so in order to include a variable for economic growth there has to be found an alternative for GDP. The International Monetary Fund (IMF) provides such a variable known as industrial production. Industrial production represents the growth of various sectors in the economy and is available on a monthly basis. Its data are seasonally corrected and the base year of industrial production is 2005. Data on almost all economic variables are taken from the CBS, or the Statistics Netherlands database, which can be found on http://statline.cbs.nl/statweb/. Dutch consumer confidence is also available on this website. Consumer confidence is biased by seasons, as during summer and spring people are on average more optimistic than they are during fall and winter. The CBS therefore provides a seasonal corrected value of consumer confidence. The seasonal corrected variable of consumer sentiment is the indicator for consumer confidence in this paper. The stock market is measured according to the AEX index, the largest stock trading market in the Netherlands, monthly data are conducted as the average of every month’s closing numbers.
5.1 Additional variables
four variables do not fully explain changes in consumer confidence. In order to determine consumer confidence better, several additional variables can be used. Section 3.2 gives a summary of the most important determinants of consumer confidence according to the literature. All the variables set out in section 3.2 will be included in equation (2).
Two different kinds of inflation measures are included. The first is the amount of actual inflation. The second is perceived inflation, which gives the percentage of Dutch citizens who argue to have witnessed a strong price increase in the past 12 months. Perceived inflation is included as an approximation to measure the confidence that citizens have in their currency. If people are not very confident about their currency, they will perceive inflation to be higher than it in fact is. The hypothesis is that high (perceived) inflation causes people to lose confidence in the economic strength of their country, which will result in a low level of consumer confidence.
The interest rate is presented by the yield spread, or the difference between long run and short run interest rates. When the yield spread is large, the long run interest rate is a lot higher than the short run interest rate. Short run interest rate is an important indicator for the business cycle. The yield spread is a rough indication for monetary policy. In example, assuming that the central bank is credible, the tightening of the monetary policy by the central bank will temporarily increase short rates, and market participants will expect future short rates to be lower than the current rate. As a result, long-‐ term interest rates will increase less than the short rate. Hence, short-‐term interest rates rise relatively more than long-‐term rates, a decrease in the yield spread, leading to a monetary contraction that may contribute to an economic slowdown and lower inflation (Ferreira et al., 2005). Hypothetically, a decrease in the yield spread has a negative influence on consumer sentiment, as a decrease in the yield spread may lead to an economic slowdown and people will for that reason rate their economic future more pessimistic. The long-‐term interest rate is the capital market interest rate, taken from the CBS database. The short-‐term interest rate is taken from the Eurostat date base and is denoted as the three-‐ month euro interest rate.
The oil price is measured as the crude oil price per barrel in Euros; this information is taken from the database of the European Central Bank (ECB). Hypothetically consumers respond negatively to increases in the oil price. This follows from the fact that prices of consumer goods rise and consumers thus have a relatively lower income, which will cause them to rate the economic future more pessimistic.
January 1986 until January 2012. The hypothesis is that if the temperature is above average, people are more optimistic about the future and consumer confidence will as a result increase.
Politics can be measured in several ways, but in this paper there will be a focus on political stability. It is fair to say that one can speak of political instability when extreme political parties receive a lot of votes. These extreme political parties are also known as protest parties, political parties such as such as the Socialist Party (SP) or the Partij Voor de Vrijheid (PVV) receive a great number of votes from people who disagree with the current government and its policy. Political instability is measured based on the monthly average of weekly polls carried out by Maurice de Hond, from September 2002 onwards. Data from all his polls can be found on Maurice de Hond’s website www.peil.nl. Before 2002 there were only sporadic political polls, especially during election time. The weekly polls were averaged as to obtain monthly figures. The hypothesis is that if there is an increase in the amount of votes that protest parties receive, there is a growing instability in politics causing people to rate the economic future negatively. Examples of extreme political parties from 2002 until now are Lijst Pim Fortuyn (LPF), the Socialistische Partij (SP), Trots op Nederland (ToN) and the Partij Voor de Vrijheid (PVV).
We include a dummy variable for periods of economic recession. One can speak of a recession if there is a decrease in the growth of the GDP for two quarters in a row. Between 1986 and 2012 there were 2 official recessions, from the first of October 2008 until the thirty-‐first of June 2009 and from the first of July 2011 until the thirty-‐first of March 2012. The dummy variable will take a value 1 in times of an economic recession and 0 otherwise. Hypothetically, if a country’s economy is in an economic recession, people will lose confidence in their economic future and will as a consequence rate consumer confidence lower opposed to a situation where there is no economic recession.
As equation 2 shows, the long run average growth rate of all the independent variables, except for the rate of unemployment, have to be subtracted from their growth rate at time t. The growth rate of each variable is measured as the percentage change in level from month t-‐1 to month t. The yield spread and perceived inflation however, are variables that are already given in percentages so the growth will be measured in percentage points. The long run average is determined as average of the entire monthly growth rate or percentage points from 1986-‐2012, this in order to put the variables in better perspective.
Table 5: Summary of hypotheses
Table 6: Descriptive statistics
CC: level of consumer confidence; HP-‐LR: growth in housing price – long-‐run average growth; SM-‐LR: growth in stock market – long-‐run average growth; UNEMP: increase in unemployment; YS-‐LR; growth in yield spread – long-‐run average growth; INFL-‐ LR: growth in inflation – long-‐run average growth; POL-‐LR: growth in political instability – long run average growth; IP: growth in industrial production – long-‐run average growth; OILP-‐LR: growth in oil price – long-‐run average growth; TEMP-‐LR: increase in temperature – average temperature; PERCI-‐LR: growth in perceived inflation – long-‐run average growth.
Hypothesis Expected Sign
1. An increase in unemployment rate leads to a lower level of consumer confidence -‐
2. A rise in housing prices leads to a higher level of consumer confidence +
3. A rise in industrial production growth leads to a higher level of consumer
confidence +
4. A rise in the AEX index leads to a higher level of consumer confidence +
5. An increase in perceived inflation leads to a lower level of consumer confidence -‐
6. An increase in the yield spread leads to a higher level of consumer confidence +
7. An increase in the temperature leads to a higher level of consumer confidence +
8. An economic recession leads to a lower level of consumer confidence -‐
9. An increase in support for extreme political parties leads to a lower level of
consumer confidence -‐
10. An increase in the oil price leads to a lower level of consumer confidence -‐
11. An increase in the actual inflation leads to a lower level of consumer confidence -‐
Variable Mean Std. Deviation Min Max
6. Results
Regression analysis is used in order to find the determinants of Dutch consumer confidence. In this section the estimation results are discussed. Table 6 provides a summary of all the variables, in table 8 the outcomes are outlined. Section 6.1 discusses the results of the rolling regressions. The first research question is whether the results of Stokman (2005) hold if more recent data is used. The determinants that were found by Stokman are GDP growth, the change in the stock market the change in housing prices and, the change in the amount of unemployment. As already explained in section 5, this paper uses monthly data from January 1986 until June 2012. Instead of GDP growth, industrial production is used as an indicator. The monthly long run averages of industrial production growth, housing price growth and stock market growth are respectively, 0.17%, 0.46% and 0.46%. Table 7 shows the test for correlations; if variables are highly correlated, the coefficient estimates may change erratically in response to small changes in the model or data and bias the standard errors (Lawson, 2006). As can be noted from table 7, the correlations do not give rise to any worry.