Is there a direct effect of consumer confidence on production/unemployment?

(1)

BACHEOLOR THESIS

Is there a direct effect of consumer confidence on

production/unemployment?

Douglas Konadu 10342745

Bsc Economics and Finance Supervisor: N. Ciurilla

(2)

1. Introduction

It is an accepted belief that consumer confidence index has the power to predict the level of consumption in the economy, especially in the short run. Howrey (2001), for example, finds that the consumer confidence index for a certain month can be used to predict household consumption for the subsequent quarter. Ludvigson (2004), Slacalek (2003) and Gelper et al (2007) all confirm the predictive power of the consumer confidence index for household consumption. The index can even be used as an indicator to signal an upcoming recession or recovery in the economy, as found by Howrey (2001). Economic agents like governments, businesses and central banks hence attach great value to consumer confidence indexes. It even goes to the extent that political leaders1_{urge consumers to show more confidence in} the economy, which will make everything better.

In this paper, the interest lies not in the predictive power of consumer confidence index for consumption, but on whether it has the capability of affecting production and therefore unemployment, so the supply side of the economy. Do producers react to consumer confidence surveys as they are published by adjusting production firing people when confidence is low? Or is does the index have no notable influence on these factors?

To be able measure if there is any effect of confidence on production and

unemployment VAR models will be estimated. VAR models are particularly handy in making forecasts. Granger causality tests will be conducted in order to measure the channel

causation concerning the confidence index and unemployment. The areas to be analysed are the Euro Area and the United States. These areas have been selected because of their relative comparability and data availability.

The rest of the paper is built up as follows, chapter 2 presents a literature review in which existing literature on the subject is discussed. Chapter 3 follows with descriptions of the data to be used. Chapter 4 contains the econometric methodology by which the main findings of this paper can be researched. The results are presented and discussed in chapter 5, and the conclusion of the main findings follows in chapter 6.

1_{Example is the Prime Minister of the Netherlands urging Dutch citizens to show more}

confidence in the economy, (http://www.volkskrant.nl/dossier-kabinet-rutte-ii/rutte-er-wordt-teveel-gemopperd-en-gesomberd~a3424297/)

(4)

2. Literature Review

There is a lot of literature on consumer confidence and its effect on consumption and the economy as whole. This chapter aims at presenting some of the popular views on the relation between consumer confidence and output. However, unlike this paper, most of the existing literature focuses on consumer confidence and economic activity and not on the relation between consumer confidence index and unemployment. However, since GDP fluctuations and unemployment are closely related, the main conclusions on these effects can be extended to unemployment as well.

The consumption decision is crucial for short-run analysis because of its important role in determining aggregate demand in the economy. It forms two-thirds of GDP, therefore fluctuations in consumption are an essential element in booms and recessions (Mankiw, 2010). Adam Smith fittingly said, “Consumption is the sole end and purpose of all production.” Considering this role of consumption in GDP determination, the confidence of consumers in the economy can be assumed to have a significant role to play in the level of economic activity. This is indeed what most research on this relation have found.

Ludvigson (2004), for example, finds that measures of consumer confidence on their own have statistically and economically significant predictive power for quarterly

consumption growth. Using both the Conference Board’s and Michigan’s consumer confidence indexes, he finds that the overall index of both surveys explain about 15 percent of the one-quarter ahead variation in total personal consumption expenditure growth of households in the U.S.. The expectations components have even greater

predictive power. The Conference Board’s expectations component explains more than 20 percent of the variation in next quarter’s consumption growth whiles this is about 16 percent for Michigan’s expectations component. He however finds that this forecasting power of consumer confidence index weakens when a standard set of baseline economic indicators are included in the model. This means that the information from the confidence index is already contained in other popular economic and financial indicators2_{. Forecasting} power of the confidence index remains strong though for expenditure growth on goods (excluding motor vehicles).

Leeper (1992) also finds that consumer confidence index has very little incremental forecasting power. Consumer confidence, as he argues, is only weakly correlated to

unemployment and industrial production when financial indicators are controlled for. The

2_{Ludvigson (2004) uses lagged values of the confidence index, labour income growth, the (log) first} difference of the real stock price and the first difference of the three-month Treasury bill rate as baseline economic and financial indicators.

(5)

information contained in the index, tend to be dominated by other easily obtainable

variables, even when changes in consumer confidence are large and unexpected. According to his analysis, changes in consumer confidence is not a reflection of past economic

conditions.

The predictive power of consumer confidence cannot be generalized to all periods as found by Batchelor and Dua (1998). They conclude in their paper that the index would have been useful in predicting the 1993 recession in the U.S., but this can however not be generalized to other periods as well, and hence tend to reflect the special nature of that particular recession.

Matsusaka and Sbordone (1995) on the other hand, find that consumer confidence goes beyond consumption forecasting and directly affects GNP even after controlling for economic fundamentals3_{. They make use of the assumptions in multiple equilibria models} to analyse whether output can indeed fluctuate as a result of the expectations of

economic agents. A feature of multiple equilibria models, contrary to that of single equilibrium models, is that output can fluctuate simply because everyone expects it to. Expectations in these models are self-fulfilling since agents must expect to be in a certain equilibrium before the economy can progress further. The expectations of agents thus cause the movement of the economy to the equilibrium. This implies that when consumer confidence is low, reflecting consumers’ expectations of less expenditure (due to perhaps higher unemployment and lower income), a fall in GDP can be expected. These models therefore provide a theoretical backing for the influence consumer confidence can have on output. After estimating their model, the Matsusaka and Sbordone find that between 13 and 26 percent of innovation variance in GNP is caused by consumer confidence, and hence finds a quantitative economic relevance for consumer confidence. They conclude on basis of their findings that consumer confidence is an important independent factor in economic fluctuations.

Also Ryan and Shinnick (2011) find that large changes in consumer confidence among other leading economic indicators, has significant economic relation with changes in GDP growth. The authors find that lagged values of the confidence index have even more significance and urge policy makers to focus more on this rather than monetary and other macro-economic indicators when making predictions on GDP growth.

And contrary to the results found by Batchelor and Dua (1998) mentioned above, Howrey (2001) attributes a stronger forecasting power to consumer confidence. After developing a recession signalling model based on the consumer confidence index, he finds that current-quarter monthly values of the confidence index, either alone or in

(6)

conjunction with other indicators, contains significant information on the probability of a recession and is hence a statistically significant predictor of the future rate of real GDP growth.

Gelper et al. use a different approach and also find evidence of a significant forecasting power of US consumer confidence index for future consumer spending. Making use of the vector error-correcting (VEC) model, the authors are able to decompose the short-run and long-run effects. They establish a very strong long-run Granger causality relation between the consumer confidence index and consumption. Furthermore, consumer confidence for remains a useful predictor of consumption for larger time horizons.

Most of the research discussed so far covers the US economy. Deés and Brianca (2011) explores the relation between consumer confidence and consumption for the Euro Area as well as the U.S. They find similar results as found in some of the literature already introduced in this chapter. For the Euro Area, the authors establish that confidence is a particularly useful predictor of consumption expenditure in certain periods, mainly in periods of high uncertainty (like during financial crises or geopolitical tensions). They also establish evidence of a one-way “confidence channel” between the U.S. and the Euro Area. This entails that past changes in U.S. consumer confidence would contain information about current fluctuations in Euro Area consumption. The opposite of this relation however, does not hold.

This paper aims at finding the effect of consumer confidence on unemployment and production. It is hence imperative to establish the effect of fluctuations in consumption on unemployment. Mankiw (2010) shows from 1970 to 2009 data (U.S.) that unemployment rises whiles consumption growth declines in each recession. Okun’s Law, illustrates the negative relationship between unemployment and GDP. Arthur Okun found that a

percentage change in GDP causes a three percent change in real GDP. Other labour-market measures, like job vacancies, decline during recessions as well. This clearly shows a fall in demand of labour during recessions leading to a rise in unemployment. Considering this fact and the relation established in previous research between consumer confidence and recessions, it can be argued that there is indeed a direct effect of consumer confidence on unemployment and production. This effect would work through the following channel; consumers share their level of confidence in the economy, producers anticipate a rise or fall in consumption, and based on the confidence index adjust production and hence labour demand appropriately. A result of this, depending on the confidence index, could be a fall in production and rise in the rate of unemployment.

Malley and Moutos (1996), on other hand, find that the unemployment rate has a significant influence on consumption spending, and that exogenous changes in

(7)

unemployment cause changes in consumption rather than the other way around. This result means, at least in the Granger sense, that changes in consumption do not cause changes in the rate of unemployment.

After observing and evaluating existing literature on the subject to be investigated in this paper, there seems to be a consensus that consumer confidence causes fluctuations in consumption and GDP as a whole. Evidence for this relation between confidence and unemployment is barely existent and is hence the focus of this paper.

(8)

3. Data

This chapter aims at providing the descriptions and sources of the data used in this paper. Time series data used covers the period of the 1995 to 2014. The focus in this paper lies in the analysis of consumer confidence index. How the index is measured varies

between countries. It is hence imperative to clearly define the discrepancies in the measurement between the countries that will be analysed here. Data on consumption, production and other forms of data used will be discussed briefly. The two areas being analysed in this paper, U.S. and the Euro Area, are both relatively large areas with many similarities and hence comparable, but also have the added bonus of freely available databases where data can be collected easily.

Consumer

Confidence index in U.S.

Ludvigson (2004) gives a detailed description of how the consumer confidence index is measured in the United States. As the author writes the University of Michigan’s

Consumer Sentiment Index and the Conference Board’s Consumer Confidence Index are the most broadly followed measures of U.S. consumer confidence. The two indexes, essentially measure the same concept, but are based on different questions and sometimes gives conflicting signals, as Ludvigson explains. He also adds that though financial markets and business community closely follow both indexes, much published academic research make use of the Michigan index, most likely because of its longer time series. Both the University of Michigan and the Conference Board base their overall index on five questions4_{. Each of} the questions are given equal weight in the overall index. Both institutions also report two component indexes, namely a present situation component and an expectations

component. Two of the 5 questions asked in the surveys assesses the present economic conditions of the respondents whiles the other three questions concerns the respondent’s expectations of the economy. The Conference Board’s present conditions component take an approach that evaluates current business conditions as well as labour availability. Hence its index tracks labour market conditions like the unemployment rate and the growth in payroll employment.

Ludvigson argues further that the Conference Board’s index is particularly useful when the components are examined individually. The reason lies in the correlation of the present conditions component with the unemployment rate (negative correlation of 0.88)

4_{For a more detailed description of the differences between both indexes and how they are} calculated, consult Ludvigson (2004). Exhibit one of the Appendix presents the questions asked in both surveys

(9)

and that of the expectations component with the rate of economic growth (correlation with a four-quarter percent change in GDP of close to 0.60). Both components of

Michigan’s index are closely correlated though. Considering the availability of Michigan’s index as opposed to the Conference Board’s, and the fact that most academic research make use of the Michigan’s index, it is most appropriate to use this index in this paper as well. Following Ludvigson’s recommendation, both components of the index will be analysed individually by computing the average index of the questions that assess both components.

Consumer Confidence index the Euro Area

In the case of the Euro Area, the consumer confidence index as constructed by the Directorate General for Economic and Financial Affairs (DG ECFIN) of the European

Commission is to be used. This institution conducts regular harmonised surveys for different sectors of the economies in the Euro Area. There are two surveys conducted, a monthly survey and a quarterly survey. The monthly survey is to be used in this paper as it possesses over a more comprehensive list of questions (12 compared to the 3 of the

quarterly survey). The questions asked concern financial and general economic situation, consumer prices, unemployment, major purchases of durable goods and savings. The U.S. index and that of the Euro Area have comparable questions. Both also have the two components mentioned earlier. The Euro Area data is more extensive as it also covers savings and consumer prices. It is however, not to be expected that this will produce any significant difference in the results. As in the case of the U.S. index, the components of the Euro Area confidence index will be calculated individually by computing the averages of the indexes that concern that particular component.

Control Variables

Production data to be used cover industrial production for the U.S. and Euro Area. This data can be retrieved from the website of the OECD. Unemployment and interest rate data for both areas as well as inflation data is obtained from the database of the OECD. Monthly consumption data for the U.S. is retrieved from the website of St. Louis Federal Reserve of the United States.

(10)

4. Methodology

As specified in the introduction the aim of this paper is to determine whether there is a direct effect of consumer confidence on unemployment. This chapter discusses the methodology that is to be used to establish and analyse this effect should it be present. Vector Autoregression Model and Granger Causality

Since consumer confidence is expected to have an effect on unemployment, the confidence index must then be a useful forecasting tool for the unemployment rate. Estimating vector autoregression (VARs) models, one can determine whether the

unemployment rate can be predicted using confidence indicators and the control variables as dependent variables. A VAR model uses lagged values of the independent variables as well as lagged values of the dependent variable as regressors.

To be able to measure the effect of the confidence index on unemployment and production several VAR models will be estimated. The VAR models are to serve as a framework for conducting Granger causality tests (Granger, 1969). This is to test whether the index or its components has useful predictive content for unemployment or production beyond the economic fundamentals considered to have a high degree of influence on unemployment and production. VAR models with and without the control variables for the index will be estimated. Lag selection is typically done using Akaike information criterion. Lag selection in this paper will however, not be motivated by such analyses and will hence be assigned randomly.

The channel through which consumer confidence is expected to affect unemployment is through production. Whiles a lower/higher level of production is

expected to cause a higher/lower level of unemployment, it can be equally true that levels of production in turn affect the level of unemployment. To test how these two variables cause (in the Granger sense) changes in each other, a VAR model with the monthly changes in the unemployment rate and production index is estimated. This way it can checked to see if changes in the unemployment rate causes changes in production. The model to be estimated is given by:

∆𝑈𝑡 = 𝛽10+ 𝛽11∆𝑈𝑡−1 + . .. + 𝛽1𝑝∆𝑈𝑡−𝑝+ 𝛾11∆𝑃𝐼𝑡−1 + . . . + 𝛾1𝑝∆𝑃𝐼𝑡−𝑝+ 𝑢1𝑡 (1)

(11)

where ∆U is the monthly change in the unemployment rate (defined as 𝑈_𝑡 - 𝑈_𝑡−1), ∆P is the monthly change in the production index (defined as 𝑃𝐼_𝑡 /𝑃𝐼_𝑡−1*100-100), β’s and 𝛾′𝑠 are the unknown coefficients, p is the number of lags and 𝑢_1𝑡and 𝑢_2𝑡 the error terms. The null hypothesis of the corresponding Granger causality test states that all the coefficients of one the variables in equations (1) and (2) is zero.

A more general specification for the VAR models to be estimated is; 𝑥_𝑡= 𝛼₀+ Σ_𝑖=1𝑛 _𝛽

𝑖𝑥𝑡−1+ 𝛾𝑖𝑍𝑡−𝑖+ 𝜀𝑡 . (3)

𝑥𝑡 = dependent variable (production/unemployment) at time t

𝑍𝑡−𝑖 = vector of lagged values of the control variables

Granger causality tests will be conducted on the regressions resulting from the VAR model specified in equation (3).

(12)

5. Results

In this chapter the main results of the regressions are presented. Firstly the VAR models and the corresponding Granger causality tests are discussed. Then, the economic

implications of these results will follow.

The first representation of how consumer confidence and unemployment are related is represented in Figure 1 in the Appendix. In both graphs (Euro Area and the U.S.) confidence seems to lead unemployment, where general drops in the index corresponds with notable rises in the rate of unemployment and higher confidence levels with a fall in the unemployment rate. The correlation between unemployment and the overall index, present conditions component and expectations component for the U.S. are -0.74, -0.81 and 0.67 respectively. For the Euro Area the correlations are not as high, namely 0.30, -0.57 and -0.07 respectively. It appears from these correlations that the present conditions component may contain more information on unemployment than the overall index.

Table 1 presents the results of the granger causality test in equations (1) and (2). In the table, the variable under equation is the dependent variable. The independent

variable is the excluded variable, which represents lags of the independent variable. The number of lags used is represented under df, whiles the Prob > F corresponds to the p-value of the F statistic.

The results indicate that at the 1 percent level, variations in the production index significantly cause variations in the rate of unemployment for all lag lengths considered in the models for the U.S. as well as the Euro Area. In the U.S. case, variations in the

unemployment rate also significantly causes variations in production 2 months and 8 months later at the 1 percent level, whiles the result is strongly insignificant (40%) in the intermediate term of 4 months. In the case of the Euro area, variations in unemployment ceases to cause any significant variations in production after 2 months. This is represented by a p-value of 44 and 16 percent for lags 4 and 8 respectively. These results suggest that in the very short term (1 to 2 month), unemployment and production influence each other, but only production consistently causes significant variations in the rate of unemployment in the longer run (beyond 4 months). These results thus provides and empirical basis for the channel discussed in the introduction.

(13)

Table 1: Granger causality Wald tests A)

Equation Excluded F df df_r Prob > F

Model 1 fdusut dusprodindex 12.019 2 194 0.0000 dusprodindex fdusut 4.9546 2 194 0.0080 Model 2 fdusut dusprodindex 7.325 4 150 0.0000 dusprodindex fdusut 1.0102 4 150 0.4041 Model 3 fdusut dusprodindex 4.0172 8 62 0.0007 dusprodindex fdusut 2.3921 8 62 0.0257 B)

Model 1 fdeaut deaprodindex 8.1107 2 194 0.0004 deaprodindex fdeaut 4.4226 2 194 0.0132 Model 2 fdeaut deaprodindex 4.8771 4 150 0.0010 deaprodindex fdeaut 0.93659 4 150 0.4445 Model 3 fdeaut deaprodindex 4.4856 8 62 0.0002 deaprodindex fdeaut 1.5428 8 62 0.1610 Confidence and production

The first regressions are run using only the confidence index as explanatory variable. The results can be found in table 2. The number in brackets corresponds with the number of lags used. For the U.S., the confidence index has significant predictive power (granger causality) for production at lag levels 2 and 8, with 1 percent and 6 percent significance respectively. The confidence index is strongly insignificant when 4 lags are considered. The production index of the U.S. is only a significant predictor of the confidence index at lag level 2. In the case of the Euro Area, the production index remains a significant predictor of the confidence index at all lag levels (5% significance), whiles the confidence index is only a significant predictor of production at lag levels 2 and 4.

Table 2

Equation Excluded P-value (2) P-value (4) P-value (8)

usindex usprodindex 0.0926 0.2346 0.3964

usprodindex usindex 0.0058 0.7817 0.0508

eaindex eaprodindex 0.0146 0.0481 0.0269

(14)

The next step is to look at the components of the confidence index individually. The results for the present conditions component in table 3A are quite similar to that of the overall index. The only notable difference is that the U.S. production index granger causes the present conditions component at all lag levels. The expectations component of the indexes are completely insignificant at predicting production, except in the case of the U.S. at lag level 2. There is no reverse causation to be recorded either.

The above analysis show that the overall index is sufficient in predicting production than when the components are taken individually.

Table 3 A)

uspc usprodindex 0.0035 0.0017 0.0012

usprodindex uspc 0.0065 0.9008 0.0283

eapc eaprodindex 0.0111 0.0108 0.0127

eaprodindex eapc 0.0030 0.0557 0.1904 B)

usexp usprodindex 0.1284 0.5154 0.9209

usprodindex usexp 0.0131 0.7556 0.1449

eaexp eaprodindex 0.2286 0.2042 0.3599

eaprodindex eaexp 0.5432 0.8840 0.3185

After looking at the overall index and the components individually, control variables can now be included to see if the index keeps any of its predictive power for production.

Table 4 presents the results of granger tests conducted inclusive of the control variables. At the 1 percent significance level, confidence causes unemployment at all lag levels, except in the case of the U.S. at lag level 4. For the Euro Area, all variables taken together significantly causes unemployment in all periods as well. In most cases, the interest spread variable also significantly causes unemployment. Inflation is a significant predictor of unemployment only after 8 periods. The results for the variation in the

production index is not included in this paper as this is very similar to the results discussed in the regressions here.

(15)

Table 4 A)

usprodindex usindex 0.0154 0.8788 0.0075 dusoil 0.5960 0.2714 0.2922 duscpi 0.6616 0.3925 0.0163 fduscon 0.1833 0.8694 0.1647 usint 0.0002 0.0767 0.0001 ALL 0.0000 0.3046 0.0004 B)

eaprodindex eaindex 0.0000 0.0016 0.0001

deaoil 0.0165 0.0323 0.0013

deacpi 0.2199 0.6289 0.0026

eaint 0.0024 0.0304 0.0000

ALL 0.0000 0.0008 0.0000

Confidence and unemployment

Table 1A of the Appendix presents the Stata output of the VAR (lag level of 2) of the overall confidence index and unemployment. The corresponding Granger causality results are presented in Tables 5A and 5B below. For both the U.S. and the Euro Area, the confidence index significantly granger causes the unemployment rate at the 1 percent significance level at all lag levels, except in the case of the Euro Area at lag level 4

(though still significant at the 5 percent level). The reverse causation, from unemployment to the confidence index, is only significant for the U.S. at lag level 8 (p-value of 4.79 percent).

(16)

Table 5 A)

Model 1 usindex usut 0.8685 2 194 0.4212 usut usindex 12.57 2 194 0.0000 Model 2 usindex usut 0.12189 4 150 0.9745 usut usindex 3.8604 4 150 0.0051 Model 3 usindex usut 2.1111 8 62 0.0479 usut usindex 2.8603 8 62 0.0090 B)

Model 1 eaindex eaut 7.7229 2 194 0.0006 eaut eaindex 25.549 2 194 0.0000 Model 2 eaindex eaut 6.6164 4 150 0.0001 eaut eaindex 2.9908 4 150 0.0207 Model 3 eaindex eaut 2.5671 8 62 0.0174 eaut eaindex 2.9763 8 62 0.0069 Table 6 presents the results for the components of the confidence index. Table 6

A)

usut uspc 0.0000 0.0000 0.0010

uspc usut 0.9588 0.7184 0.2923

eaut eapc 0.0000 0.2032 0.4873

eapc eaut 0.0000 0.0000 0.0033 B)

usexp usut 0.1651 0.8578 0.0491

usut usexp 0.0034 0.1088 0.1130

(17)

eaut eaexp 0.0034 0.0519 0.0207 The results for the U.S. of the present conditions component is similar to that of the overall index. From Table 6 it becomes clear that just as in the case of the production index, the components do not possess over sharper predictive power for the rate

unemployment. In the next analyses therefore, where the control variables are added, only the overall index will be considered.

The granger tests including the control variables are presented in results of the Table 6. For the U.S., the confidence index significantly causes unemployment at lag level 2 (1 percent significance level). Consistent with results obtained in earlier results, this causation however disappears after 4 periods and becomes significant again after 8

periods. The consumption variable behaves in this manner as well. The other variables are only significant predictors of the unemployment rate after 8 periods, although lagged values of all the explanatory variables taken together remain a significant predictor of the unemployment rate in all periods modelled (5 percent significance level). In the case of the Euro Area, the confidence index is only an insignificant predictor of unemployment after 8 periods. However, the inability of the other variables to predict unemployment may indicate exclusion of relevant control variables in the model. All variables together remain significant predictors of unemployment, which in turn indicate the strength of the confidence index in predicting unemployment.

Table 6 A)

usindex 0.0026 0.3181 0.0013 usut fduscon 0.0156 0.0534 0.0036 duscpi 0.3547 0.7759 0.0034 usint 0.3540 0.9946 0.0412 ALL 0.0000 0.0423 0.0002 B)

eaut eaindex 0.0000 0.0024 0.1393

deacpi 0.6140 0.2879 0.8873

eaint 0.7876 0.1420 0.1285

(18)

6. Conclusion

The ability of consumer confidence to predict consumption is barely debatable. Many a research done in this area find significant results of the predictable power of consumer confidence index for consumption, and in some cases even for GDP. These research were mainly focused on the demand side of the economy. There is hardly any research done to find the effect of confidence for production and labour supply. In this paper, the supply side of the economy was analysed, to find if confidence has similar impact on the supply side as it has on the demand side.

Employing the convenient analysis of Granger causality tests, this paper has been able to establish significant results for confidence index on production and unemployment. It has been shown that confidence when taken alone, produces significant results for causation of unemployment, at least in the granger sense. In all models analysed, lagged values of confidence produced significant causation results in the short term, in this case two months. This result was obtained for the U.S. as well as the Euro Area. Inclusion of control variables did not diminish the significance established when confidence is taken alone. In the case of the United States, confidence was found to usually be a significant predictor of unemployment and production after only in the short and long terms (2 and 8 months) and insignificant in the medium term (4 months). It is not well understood which economic channels causes this to happen. For the Euro Area confidence was found to be a significant predictor of production in all periods considered, and only insignificant after 8 periods for unemployment.

The results of this paper provide strong evidence of the effect confidence has on production and unemployment. A weakness of the paper, however, lies in the fact that some of the control variables used may not have the most relevant variables to control for the effect of confidence index. Perhaps a follow up research can make better use of more sophisticated econometrics models and better control variables to produce more confident results.

In conclusion, confidence index indeed has a significantly measurable effect on production and unemployment. It is however possible that this effect will vanish after the use of superior economic fundamentals as control variables.

(19)

References

Howrey, E.P. (2001). The Predictive power of the index of consumer sentiment, Brookings

Papers on Economic Activity, 1, 175-216.

Batchelor, R., and Dua, P. (1998). Improving macro-economic forecasts: The role of consumer confidence. International Journal of Forecasting, 14(1), 71-81. Bram, J. and Ludvigson, S. (1997). Does confidence forecast household expenditure? A

sentiment horse race. Research paper no. 9708, Federal Reserve Bank of New York. Leeper, M.E. (1992). Consumer attitudes: King for a day. Economic Review, Federal

Reserve Bank of Atlanta, 77 (4), 1-15.

Ludvigson, S.C. (2004). Consumer confidence and consumer spending. Journal of Economic

Perspectives, 18 (2), 29-50

Granger, C.J.W. (1969). Investigating causal relations by econometric models and crossspectral methods. Econometrica, 37, 424-438.

Dees, S., and Brinca, P.S. (2011). Consumer confidence as a predictor of consumption spending: evidence for the United States and the Euro area. Working Paper Series, No.1349, European Central Bank.

Gelper, S., Lemmens, A. and Croux, C. (2007). Consumer sentiment and consumer

spending: decomposing the Granger causal relationship in the time domain, Applied

Economics, 39 (1), 1-11.

Ryan, G. and Shinnick E. (2011). Real economic activity leading Indicators: should we have paid more attention?, Journal of Economic Policy Reform, 14 (2), 105-125.

Malley, J. and Moutos, T. (1996). Unemployment and consumption, Oxford Economic Papers, 48, 584-600.

Matsusaka, J.G. and Sbordone, A.M. (1995). Consumer Confidence and Economic

Fluctuations. Economic Inquiry, Western Economic Association International, 33

(2), 296-318.

Mankiw, N.G. (2009). Macroeconomics. Seventh Edition. Worth Palgrave Macmillan Romer, D. (2011). Advanced Macroeconomics. Fourth Edition, McGraw-Hill Irwin

(20)

Stock, J.H. and Watson, M.M. (2012). Introduction to econometrics. Third Edition, Pearson Education Limited.

(21)

Appendix

Exhibit 1

Michigan Survey Conference Board Survey

PRESENT CONDITIONS QUESTIONS PRESENT CONDITIONS QUESTIONS

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]

Q1) How would you rate present general business conditions in your area?

[good/normal/bad]

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]

Q2) What would you say about available jobs in your area right now? [plentiful/not so many/ hard to get]

EXPECTATIONS QUESTIONS EXPECTATIONS QUESTIONS

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]

Q3) Six months from now, do you think business conditions in your area will be

[better/same/worse]?

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]

Q4) Six months from now, do you think there will be [more/same/fewer] jobs available in your area?

Q5) Now looking ahead—do you think that a year from now, you (and your family living there) will be better off financially, or worse off, or just about the same as now?

[better/same/worse]

Q5) How would you guess your total family income to be six months from now? [higher/ same/lower]

Source: Consumer Confidence and Consumer Spending, Sydney C. Ludvigson (2004)

Exhibit 2

B. Index of Leading Indicators (1988 Version)

1. Average weekly hours of production or nonsupervisory workers in manufacturing 2. Average weekly initial claims for unemployment insurance in state programs 3. Manufacturers’ new orders in consumer goods and materials industries 4. Contacts and orders for plant and equipment

5. Index of new private housing units authorized by local building permits 6. Index of stock prices of 500 common stocks

(22)

8. Percent of companies receiving slower deliveries from vendors 9. Change in sensitive materials prices

10. Change in business and consumer credit outstanding

11. Change in manufacturing and trade inventories on hand and on order Bron: Matsusaka and Sbordone (1995)

Figure 1 A) B 0 2 4 6 8 10 12 0 20 40 60 80 100 120

US relation between consumer confidence and

unemployment

usindex usut 1995 2014 0 2 4 6 8 10 12 14 -40 -35 -30 -25 -20 -15 -10 -5 0 5 un emp lo ymen t co nf ide nce i nde x

Euro Area relation between confidence index and

unemployment

eaindex eaut

(23)

Table 1 A) _cons .7781351 .1567894 4.96 0.000 .4689044 1.087366 L2. -.0209961 .0711175 -0.30 0.768 -.1612589 .1192667 L1. .9809053 .0730107 13.44 0.000 .8369086 1.124902 usut L2. -.0021915 .0029013 -0.76 0.451 -.0079138 .0035307 L1. -.0039661 .0027929 -1.42 0.157 -.0094745 .0015423 usindex usut _cons 9.421252 4.096816 2.30 0.023 1.341234 17.50127 L2. .443361 1.85826 0.24 0.812 -3.221624 4.108346 L1. -.7731033 1.907727 -0.41 0.686 -4.535652 2.989445 usut L2. -.0246768 .0758103 -0.33 0.745 -.1741949 .1248413 L1. .9381648 .0729775 12.86 0.000 .7942336 1.082096 usindex usindex Coef. Std. Err. t P>|t| [95% Conf. Interval]

usut 5 .151905 0.9924 6476.296 0.0000 usindex 5 3.96918 0.9167 547.1653 0.0000 Equation Parms RMSE R-sq F P > F

Det(Sigma_ml) = .3359687 SBIC = 4.851012 FPE = .3714973 HQIC = 4.752498 Log likelihood = -456.2092 AIC = 4.685519 Sample: 3025m07 - 3604m07, but with gaps No. of obs = 199 Vector autoregression

(24)

B) _cons .0902128 .0667283 1.35 0.178 -.0413933 .221819 L2. .0765745 .0740761 1.03 0.303 -.0695234 .2226724 L1. -.0800265 .0739495 -1.08 0.281 -.2258747 .0658218 deaprodindex L2. -.5855508 .8369833 -0.70 0.485 -2.236306 1.065204 L1. -2.147544 .8611124 -2.49 0.013 -3.845888 -.4492001 fdeaut deaprodindex _cons .0057663 .0047953 1.20 0.231 -.0036913 .0152238 L2. -.0119505 .0053233 -2.24 0.026 -.0224495 -.0014516 L1. -.0195052 .0053142 -3.67 0.000 -.0299863 -.0090242 deaprodindex L2. .4148649 .0601477 6.90 0.000 .2962375 .5334923 L1. .1343579 .0618817 2.17 0.031 .0123106 .2564052 fdeaut fdeaut Coef. Std. Err. t P>|t| [95% Conf. Interval]

deaprodindex 5 .942705 0.0640 3.4016 0.0103 fdeaut 5 .067745 0.3888 31.65098 0.0000 Equation Parms RMSE R-sq F P > F

Det(Sigma_ml) = .0037525 SBIC = .3564062 FPE = .0041493 HQIC = .2578926 Log likelihood = -8.99589 AIC = .1909135 Sample: 3025m07 - 3604m07, but with gaps No. of obs = 199 Vector autoregression

(25)

Is there a direct effect of consumer confidence on production/unemployment?

BACHEOLOR THESIS