Week 1 – Assignment 1: Assess Political Decision Making and Week 2 - Assignment: Examine the Impact of Economic Policies on Society

On the dynamics of U.S. consumer sentiment and economic policy assessment Hamid Baghestania and Polly Palmerb

aSchool of Business Administration, American University of Sharjah, Sharjah, UAE; bCollege of Arts and Sciences, American University of Sharjah, Sharjah, UAE

ABSTRACT We utilize the Michigan Surveys of Consumers data to first investigate the dynamic relationship between consumers’ assessment of current and future economic conditions (Index of Current Economic Conditions (ICC) and Index of Consumer Expectations (ICE)), and then examine how these assessments are influenced by deterioration/improvement in consumers’ appraisal of economic policies of the government (GP). We further assess how deterioration/improvement in ICC and ICE influences GP. For 1978–2015, our findings first indicate that ICC and ICE are cointegrated and both respond to disequilibrium to restore the long-run equilibrium relationship. Second, deterioration in GP results in deterioration in both ICC and ICE. Improvement in GP, however, results in improvement in ICE with no impact on ICC. Third, deterioration in both ICC and ICE results in deterioration in GP. Improvement in ICE results in improvement in GP, but improvement in ICC has no impact on GP. The observed asymmetries are in line with ‘negativity bias’, whereby people tend to give more weight to negative events than to positive ones.

KEYWORDS Consumer behaviour; government policy; asymmetry; negativity bias

JEL CLASSIFICATION D12; E32; E58

I. Introduction

Every month, the Michigan Surveys of Consumers (MSC) selects a nationally representative random sample of at least 500 U.S. consumers to collect responses to approximately 50 core questions. The goal of the survey is to probe consumer sentiment on personal finances, buying conditions, and busi- ness conditions. The widely reported Index of Consumer Sentiment (ICS) utilizes the responses to two current-looking and three forward-looking questions. The responses to the current-looking questions are used to construct the Index of Current Economic Conditions (ICC), and the responses to the forward-looking questions are used to construct the Index of Consumer Expectations (ICE). Another important question on the survey asks consumers to appraise the economic policies of the government (GP). The MSC utilizes the responses to this question to construct the index, which we refer to as consumers’ appraisal of eco- nomic GP.

In this study, we first investigate the dynamic relationship between consumers’ assessment of cur- rent and future economic conditions (ICC and ICE). Maintaining that these assessments are influenced by

consumers’ appraisal of economic GP, we then eval- uate the impact of deterioration/improvement in GP on both ICC and ICE. Conversely, we maintain that GP is influenced by both ICC and ICE in order to further assess the impact of deterioration/improve- ment in ICC and ICE on GP.

Our findings indicate that ICC and ICE are coin- tegrated and thus possess a long-run equilibrium relationship. The vector error-correction model (VECM) estimates suggest that both ICC and ICE respond to correct short-run deviations from the long-run equilibrium relationship and are thus endogenous to each other. Further findings indicate that deterioration in GP results in deterioration in both ICC and ICE. Improvement in GP, however, results in improvement in ICE with no impact on ICC. Similarly, deterioration in both ICC and ICE results in deterioration in GP. In addition, improve- ment in ICE results in improvement in GP, but improvement in ICC has no impact on GP.

The observed asymmetries are in line with the phenomenon known as ‘negativity bias’, whereby people tend to give more weight to negative events than to positive ones. Rozin and Royzman (2001) argue that this concept seems to have been part of

CONTACT Hamid Baghestani [email protected] School of Business Administration, American University of Sharjah, P.O. Box 26666, Sharjah, UAE

APPLIED ECONOMICS, 2017 VOL. 49, NO. 3, 227–237 http://dx.doi.org/10.1080/00036846.2016.1194964

© 2016 Informa UK Limited, trading as Taylor & Francis Group

human consciousness throughout the development of civilization. Specifically, negativity is very potent. People generally tend to accept positive situations with little reaction. But negative situations cause people to try to unravel the ‘why’ underlying the situation, and because more mental time is spent on that activity, the negative situation seems to have more weight. The potential for something negative to happen has far more power over people than the potential for something positive to occur. In addition, negative events tend to seem more ‘con- tagious’ and thus have more negative implications (Rozin et al. 1997). When making decisions, in par- ticular, people tend to ‘attach greater weight to pos- sible losses than to comparable gains’ (Lewicka, Czapinski, and Peeters 1992, 425–426). In fact, loss aversion is at the core of the prospect theory, which asserts that individuals are more concerned about a loss than an equal gain (Kahneman and Tversky 1979; Tversky and Kahneman 1991). Consistent with this theory, Bloom and Price (1975), who exam- ine the American voters’ response to short-term economic conditions, conclude that economic downturns reduce the vote for the incumbent party while economic upturns have no impact. Further evidence is provided by Soroka (2006, 381), who shows that ‘Public responses to negative economic information are much greater than are public responses to positive economic information. The same trend is evident in mass media content, and this content serves to enhance the asymmetry in public responsiveness’.

The outline of this study is as follows: Section II discusses both the MSC data and survey questions. Section III presents the methodology and empirical results. Section IV concludes by discussing the implications of our findings.

II. Data and MSC survey questions

As already stated, our study utilizes the MSC data on the ICC, the ICE, and the index measuring consu- mers’ appraisal of economic GP. The MSC con- structs the ICC series by utilizing the responses to two current-looking questions: First, ‘We are inter- ested in how people are getting along financially these days. Would you say that you (and your family living there) are better off or worse off financially than you were a year ago?’ Using the individual responses, the index values are calculated as X1 = better − worse + 100. Second, ‘About the big things people buy for their homes – such as furni- ture, a refrigerator, stove, television, and things like that. Generally speaking, do you think now is a good or bad time for people to buy major household items?’ Using the individual responses, the index values are calculated as X2 = good − bad + 100. Put together, the MSC computes the ICC using the X1 and X2 index values. Figure 1 plots ICC and the actual unemployment rate for January 1978 to December 2015 (1978.01–2015.12), with the shaded areas representing the recessions identified by the National Bureau of Economic Research (NBER). As can be seen, the inverse relationship between ICC

2

4

6

8

10

12

40

60

80

100

120

140

1980 1985 1990 1995 2000 2005 2010 2015

Unemployment

ICC

Figure 1. ICC versus unemployment rate: 1978.01–2015.12.

228 H. BAGHESTANI AND P. PALMER

and unemployment is much more pronounced dur- ing recessions. The steady decline in ICC at the outset corresponds to unusually high inflation experienced from 1978 through 1982.

In constructing the ICE series, the MSC utilizes the responses to three forward-looking questions: First, ‘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?’ Using the individual responses, the index values are calculated as X3 = better − worse + 100. Second, ‘Now turning to business conditions in the country as a whole – do you think that during the next 12 months we’ll have good times financially, or bad times, or what?’ Using the individual responses, the index values are calcu- lated as X4 = good − bad + 100. Third, ‘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 5 years or so, or that we will have periods of widespread unemployment or depression, or what?’ Using the individual responses, the index values are calculated as X5 = good − bad + 100. Put together, the MSC computes the ICE using the X3, X4, and X5 index values.

Figure 2 plots ICE and ICC for 1978.01–2015.12. As indicated, ICE has a mean value of 77.83 index points with a high (low) value of 108.6 (44.2), and ICC has a mean value of 96.98 index points with a high (low) value of 121.1 (57.5). An important observation is that ICE and ICC move and turn in a similar pattern, and the difference between them appears to be mean reverting. As we shall see, ICE and ICC possess a long-run equilibrium relationship,

40

60

80

100

120

40

60

80

100

120

140

1980 1985 1990 1995 2000 2005 2010 2015

ICC: Mean = 96.98 High = 121.1 Low = 57.5

ICE: Mean = 77.83 High = 108.6 Low = 44.2

–50

–40

–30

–20

–10

0

10

1980 1985 1990 1995 2000 2005 2010 2015

ICE–ICC: Mean = –19.15 High = 1.70 Low = –41.50

Figure 2. ICE versus ICC (in index points): 1978.01–2015.12.

APPLIED ECONOMICS 229

and the difference between the two adjusted for the mean value (ICE – ICC + 19.15) represents short-run deviations from the long-run equilibrium relation- ship. The reason for the non-zero mean in the long- run equilibrium relationship may be due to the fact that the current-looking questions ask about perso- nal finances and buying conditions while the three forward-looking questions ask about personal finances, buying conditions, and business conditions.

The question used in gathering consumers’ responses for constructing the GP index is: ‘As to the economic policy of the government – I mean steps taken to fight inflation or unemployment – would you say the government is doing a good job, only fair, or a poor job?’ Using the individual responses, the MSC computes the index values as GP = good − poor + 100. As indicated in Figure 3 for 1978.01–2015.12, GP varies significantly with a mean value of 90.89 index points with a high (low) value of 143.0 (48.0).

III. Methodology and empirical results

Our analysis focuses on answering three questions:

(1) Do ICC and ICE possess a long-run equili- brium relationship?

(2) Are ICC and ICE asymmetrically influenced by GP?

(3) Is GP asymmetrically influenced by ICC and ICE?

The period under examination is 1978.01–2015.12 for which the MSC data are available at https:// data.sca.isr.umich.edu/tables.php.

Do ICC and ICE possess a long-run equilibrium relationship?

We start with investigating the stochastic behaviour of the individual series. In so doing, we first employ the Kwiatkowski, Phillips, Schmidt, and Shin 1992 (KPSS) (Kwiatkowski et al. 1992) test using the Bartlett win- dow approach with the lag truncation parameter of 12. As opposed to the augmented Dickey–Fuller (ADF) test, the KPSS test examines the null hypothesis of stationarity against a unit root alternative; the literature suggests that the KPSS stationarity test is relatively more powerful than the ADF unit root test (Bahmani-Oskooee 1998). As reported in rows 1–3 (column 1) of Table 1, we reject the null hypothesis of stationarity in favour of a unit root alternative for ICC, ICE, and GP. Further results in rows 1–3 (column 2) confirm that each series has a unit root, since we cannot reject the null hypothesis of stationarity for ΔICC, ΔICE, and ΔGP. We reach the same conclusions when using the ADF unit root test, with the augmented lag length determined using the modified Akaike cri- terion. For instance, as reported in rows 4–6 (column 1) of Table 1, we cannot reject the unit root hypothesis for ICC, ICE, and GP. Further results in rows 4–6 (column 2) confirm that each series has a unit root, since we reject the unit root hypothesis in favour of stationarity for ΔICC, ΔICE, and ΔGP. Put together,

40

60

80

100

120

140

160

1980 1985 1990 1995 2000 2005 2010 2015

GP: Mean = 90.89 High = 143.0 Low = 48.0

Figure 3. Time plot of GP (in index points): 1978.01–2015.12.

230 H. BAGHESTANI AND P. PALMER

such findings indicate that the precondition for coin- tegration is satisfied for ICC, ICE, and GP.

As the next step, we employ the residual-based cointegration test proposed by Shin (1994). This test extends the KPSS methodology to examine the null hypothesis of cointegration for a group of variables, and it is designed to overcome the low power of the standard tests examining the null hypothesis of no cointegration. Given the observation in Figure 2 that ICE and ICC move and turn in a similar pattern, we start with calculating the Shin (KPSS) test statistic for the difference between the two series adjusted for the mean value (ICE − ICC + 19.15) and compare it with the critical values in Shin (1994, Table 1). As reported in row 1 of Table 2, the calculated Shin test statistic (0.083) is below the 10% critical value. This means that we cannot reject the null hypothesis that ICE and ICC are cointegrated and, thus, the series ICE and ICC possess a long-run equilibrium

relationship in the form of ICE = ICC − 19.15 for 1978.01–2015.12. We reach the same conclusion when applying the Shin test to the residual series (ICE − 0.870 ICC + 6.55), which is obtained by regressing ICE on a constant and ICC. As reported in row 2 of Table 2, the calculated Shin test statistic (0.114) is below the 10% critical value, indicating again that we cannot reject the null hypothesis that ICE and ICC are cointegrated. Intuitively, it would be peculiar if ICE and ICC were not cointegrated, as it would imply irrational behaviour that consumers went one way in their assessments with regard to the future but another way with regard to the present.

Rows 3 and 4 report the Shin test statistics (0.570 and 0.569) on the residual series (ICE − 0.503 GP − 32.12) and (ICC − 0.403 GP – 60.38), respec- tively. These test statistics are above the 1% critical value, leading to the conclusions that (i) ICE and GP are not cointegrated, and (ii) ICC and GP are not cointegrated. Finally, row 5 reports the Shin test statistic (0.434) on the residual series (ICE − 0.606 ICC − 0.259 GP + 4.45), which is obtained by regres- sing ICE on a constant, ICC, and GP. Again, this test statistic is above the 1% critical value, meaning that we reject the null hypothesis that ICE, ICC, and GP are cointegrated.

Are ICC and ICE asymmetrically influenced by GP?

For the cointegrated series ICCt and ICEt, we specify the following structural VECM,

ΔICCt ¼ α1 þ X13

i¼1 β1iΔICCt�i þ γ1ΔICEt � λ1u1t�1 þ δ11d1 ΔGPtj j þ δ12 1 � d1ð ÞΔGPt þ V1t;

(1)

ΔICEt ¼ α2 þ X13

i¼1 β2iΔICEt�i þ γ2ΔICCt � λ2u2t�1 þ δ21d1 ΔGPtj j þ δ22 1 � d1ð ÞΔGPt þ v2t;

(2)

where the error-correction terms u1t (=ICCt − ICEt − 19.15) and u2t (=ICEt − ICCt + 19.15) measure dis- equilibrium defined as short-run deviations from the long-run equilibrium relationship. Accordingly, λ1 (λ2) measures the responsiveness of ΔICCt (ΔICEt) to disequilibrium. The negative signs by λ1 and λ2 indicate that the model is error correcting. More specifically, Equation (1) specifies the short-run con- vergence process of ICCt to the equilibrium

Table 1. Unit root test results: 1978.01–2015.12.

Row no. Variable

Levels First differences

(1) (2)

Panel 1. KPSS test results 1 ICC 0.369a 0.055 2 ICE 0.389a 0.035 3 GP 0.505a 0.044 Panel 2. ADF test results 4 ICC −1.98 (5) −26.9a (0) 5 ICE −2.47 (7) −21.8a (0) 6 GP −2.46 (4) −22.6a (0)

Numbers in Panel 1 are the calculated KPSS test statistics (obtained using the Bartlett window approach with a lag truncation parameter of 12). These statistics are compared with the critical values in Kwiatkowski et al. (1992), which are 0.347 (10%), 0.463 (5%), and 0.739 (1%). Numbers in Panel 2 are the calculated ADF test statistics, compared with the MacKinnon (1996) critical values of −2.57 (10%), −2.87 (5%), and −3.44 (1%). The augmented lag lengths (in parentheses) are determined using the modified Akaike criterion. a indicates significance at the 10% (or lower) level of significance.

Table 2. Cointegration test results: 1978.01–2015.12. Row no. Linear combination

Calculated Shin test statistic

1 ICE − ICC + 19.15 0.083 2 ICE − 0.870 ICC + 6.55 0.114 3 ICE − 0.503 GP − 32.12 0.570a

4 ICC − 0.403 GP − 60.38 0.569a

5 ICE − 0.606 ICC − 0.259 GP + 4.45 0.434a

The calculated Shin (cointegration) test statistics are the KPSS stationarity test statistics on the difference between ICE and ICC adjusted for the mean value in row 1 and on the residual series in rows 2–5. These test statistics are obtained using the Bartlett window approach with the lag truncation parameter of 12. The test statistics in rows 1–4 are compared with the critical values of 0.231 (10%), 0.314 (5%), and 0.533 (1%) from Shin (1994). The test statistic in row 5 is compared with the critical values of 0.163 (10%), 0.221 (5%), and 0.380 (1%). aindicates significance at the 10% (or lower) level of significance.

APPLIED ECONOMICS 231

relationship, with convergence being assured when 0 ≤ λ1 < 1. Similarly, Equation (2) specifies the short- run convergence process of ICEt to the equilibrium relationship, with convergence being assured when 0 ≤ λ2 < 1. With the series ICCt and ICEt cointegrated, λ1 and λ2 cannot both be equal to zero.

As reported in rows 3 and 6 of Table 1, GPt has a unit root and thus ΔGPt is stationary. In addition, our results in row 5 of Table 2 do not suggest the inclusion of GPt in the long-run equilibrium rela- tionship. Therefore, the VECM includes ΔGPt as part of the short-run dynamics. The model also includes the dummy variable d1 (=1 when ΔGPt < 0, and =0 otherwise). As such, d1|ΔGPt| represents deterioration in consumers’ appraisal of economic GP, and (1 − d1) ΔGPt represents improvement in consumers’ appraisal of economic GP. Accordingly, δ11 (δ12) in Equation (1) measures the impact of deterioration (improvement) in GPt on ICCt. Similarly, the parameter δ21 (δ22) in Equation (2) measures the impact of deterioration (improve- ment) in GPt on ICEt.

Column 1 of Table 3 (Table 4) reports the ordin- ary least squares (OLS) estimates of Equation (1) (Equation 2) with the ΔGPt variable excluded. The estimate of λ1 (0.184) and the estimate of λ2 (0.102) are both significantly between zero and one,

indicating that the VECM is dynamically stable and both ICCt and ICEt respond to correct disequili- brium. We reach the same conclusion when estimat- ing Equations (1) and (2) with the ΔGPt variable included. More specifically, column 2 of Table 3 (Table 4) reports the OLS estimates of Equation (1) (Equation 2), and column 3 of Table 3 (Table 4) reports the corresponding two-stage least squares (TSLS) estimates.1 The TSLS estimates, which purge the simultaneity bias, are very similar to the OLS estimates in column 2 and pass a series of diagnostic tests. For instance, the Ljung–Box test p-values point to the absence of autocorrelation and the White test p-values point to the absence of heteroscedasticity in the error terms v1t and v2t. In addition, as shown in Figures 4 and 5, the cusum of squares test results indicate that Equations (1) and (2) are both stable in terms of parameters.

The estimate of λ1 (0.180) in column 3 of Table 3, which is significantly between zero and one, indi- cates that 18% of the adjustment towards the equili- brium relation occurs through changes in ICCt within the immediate month. Also, the estimate of λ2 (0.081) in column 3 of Table 4, which is signifi- cantly between zero and one, indicates that 8.1% of the adjustment towards the equilibrium relation occurs through changes in ICEt within the

Table 3. Error-correction model: the ICC equation estimates.

EQ1 : ΔICCt ¼ α1 þ X13

i¼1 β1iΔICCt�i þ γ1ΔICEt � λ1u1t�1 þ δ11d1 ΔGPtj j þ δ12 1 � d1ð ÞΔGPt þ v1t

OLS TSLS

Column 1 Column 2 Column 3

α1 −0.026 (0.17) 0.403 (1.61) 0.387 (1.54) ∑ β1i −0.094 (0.47) −0.133 (0.67) −0.136 (0.68) γ1 0.429 (12.7) 0.368 (9.61) 0.358 (9.11) λ1 0.184 (8.40) 0.181 (8.30) 0.180 (8.27) δ11 −0.186 (3.25) −0.188 (3.27) δ12 0.001 (0.03) 0.011 (0.22)

Adjusted R2 0.396 0.410 0.410 Ljung–Box test p-value 0.276 0.600 0.620 White test p-value 0.277 0.285 0.277

See the text for the definition of the variables. Absolute t-ratios are in parentheses. The sample period, after adjusting for lags, is 1979.05–2015.12 with 440 observations. The Ljung–Box test examines the null hypothesis of no autocorrelation up to the 24th order. In order to ensure the absence of autocorrelation, we have initially estimated the equation by including 15 lags of the dependent variable. We have then kept the lagged dependent variables up to the one with a significant parameter estimate and dropped the remaining lags.

1We have utilized as instruments the predetermined variables in the VECM in addition to the ranks of ΔICEt, d1|ΔGPt|, and (1 − d1)ΔGPt for estimating Equation (1), and the ranks of ΔICCt, d1|ΔGPt|, and (1 − d1)ΔGPt for estimating Equation (2). The use of the rank is intended to increase the efficiency of the TSLS estimates (Baghestani 1991).

232 H. BAGHESTANI AND P. PALMER

Table 4. Error-correction model: the ICE equation estimates.

EQ2 : ΔICEt ¼ α2 þ X13

i¼1 β2iΔICEt�i þ γ2ΔICCt � λ2u2t�1 þ δ21d1 ΔGPtj j þ δ22 1 � d1ð ÞΔGPt þ v2t

OLS TSLS

Column 1 Column 2 Column 3

α2 0.085 (0.25) 0.652 (2.23) 0.618 (2.11) ∑ β2i −0.391 (1.59) −0.450 (2.12) −0.482 (2.10) γ2 0.591 (12.1) 0.447 (9.33) 0.450 (9.32) λ2 0.102 (3.02) 0.080 (2.22) 0.081 (2.54) δ21 −0.399 (6.12) −0.394 (6.05) δ22 0.154 (2.82) 0.163 (2.99)

Adjusted R2 0.277 0.378 0.378 Ljung–Box test p-value 0.984 0.955 0.954 White test p-value 0.315 0.311 0.301 Wald test p-value: H0: | δ21| = δ22

– 0.015 0.022

See the notes in Table 3.

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1985 1990 1995 2000 2005 2010 2015

___ Cusum of squares; ---- 5% significance

Figure 4. Parameter stability test results for the ICC equation.

–0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1985 1990 1995 2000 2005 2010 2015

____ Cusum of squares; ------ 5% significance

Figure 5. Parameter stability test results for the ICE equation.

APPLIED ECONOMICS 233

immediate month. Note that sentiment about the current economic conditions adjusts faster than sen- timent about future economic conditions. This makes sense, since consumers are more certain about the current economic conditions and, there- fore, there would be less hesitancy about returning to the equilibrium relationship. However, being uncertain about the future may create more hesi- tancy about returning to equilibrium.

The estimate of δ11 (−0.188) in Equation (1) is nega- tive and significant but the estimate of δ12 (0.011) is insignificant. This means that deterioration in GPt results in deterioration in ICCt, but improvement in GPt has no impact on ICCt. This asymmetry is consis- tent with the psychological negativity bias that people tend to give more weight to negative events than to positive ones. The estimate of δ21 (−0.394) in Equation (2) is negative and significant and the estimate of δ22 (0.163) is positive and significant. This means that dete- rioration (improvement) in GPt results in deterioration (improvement) in ICEt. However, the estimate of |δ21| is more than twice larger than the estimate of δ21. With the Wald test p-value of 0.022 (<0.10), we reject the null hypothesis that |δ21| = δ22 in favour of the alternative that |δ21| > δ22. Again, this asymmetry is consistent with psychological negativity bias.

Is GP asymmetrically influenced by ICC and ICE?

In answering, we formulate the following structural augmented autoregressive model,

ΔGPt ¼ α3 þ X14

i¼1 β3iΔGPt�i þ η1d2 ΔICCtj j þ η2 1 � d2ð ÞΔICCt þ ω1d3 ΔICEtj j þ ω2 1 � d3ð ÞΔICEt þ v3t;

(3)

where given our unit root test results in Table 1, the series are all specified in the first-difference form. The dummy variable d2 (=1 when ΔICCt < 0, and =0 otherwise), meaning that d2|ΔICCt| represents dete- rioration in consumers’ assessment of current eco- nomic conditions, and (1 − d2)ΔICCt represents improvements in consumers’ assessment of current economic conditions. Accordingly, the parameter η1 (η2) measures the impact of deterioration (improve- ment) in ICCt on GPt. Similarly, the dummy variable d3 (=1 when ΔICEt < 0, and =0 otherwise), meaning that d3|ΔICEt| represents deterioration in consumers’ assessment of future economic conditions, and (1 − d3)ΔICEt represents improvement in consu- mers’ assessment of future economic conditions. Accordingly, the parameter ω1 (ω2) measures the impact of deterioration (improvement) in ICEt on GPt.

Column 1 of Table 5 reports the OLS estimates of Equation (3), and column 2 reports the correspond- ing TSLS estimates.2 The TSLS estimates, which purge the simultaneity bias, are very similar to the OLS estimates in column 1 and pass a series of diagnostic tests. For instance, the Ljung–Box and White test p-values point to the absence of autocor-

Table 5. Augmented autoregressive model: the GP equation estimates.

EQ3 : ΔGPt ¼ α3 þ X14

i¼1 β3iΔGPt�i þ η1d2 ΔICCtj j þ η2 1 � d2ð ÞΔICCt þ ω1d3 ΔICEtj j þ ω2 1 � d3ð ÞΔICEt þ v3t

OLS TSLS

Column 1 Column 2

α3 0.601 (1.19) 0.565 (1.21) ∑ β3i −0.548 (2.71) −0.531 (2.62) η1 −0.459 (3.53) −0.448 (3.27) η2 −0.029 (0.25) −0.066 (0.56) ω1 −0.412 (3.62) −0.438 (3.66) ω2 0.501 (4.79) 0.569 (5.20)

Adjusted R2 0.230 0.230 Ljung–Box test p-value 0.973 0.966 White test p-value 0.242 0.187 Wald test p-value:

H0: | ω 1| = ω2 0.622 0.493

See the notes in Table 3.

2For the TSLS estimation, we have utilized as instruments the predetermined variables in Equation (3) in addition to the ranks of d2|ΔICCt|, (1 − d2)ΔICCt, d3| ΔICEt|, and (1–d3)ΔICEt.

234 H. BAGHESTANI AND P. PALMER

relation and heteroscedasticity in the error terms v3t and, as shown in Figure 6, the cusum of squares test results suggest that Equation (3) is stable in terms of parameters.

With the estimate of η1 (−0.448) negative and significant and the estimate of η2 (−0.066) insignif- icant, we conclude that deterioration in ICCt results in deterioration in GPt, but improvement in ICCt has no impact on GPt. This asymmetry is consistent with psychological negativity bias. The estimate of ω1 (−0.438) is negative and significant and the estimate of ω2 (0.569) is positive and significant. This means that deterioration (improvement) in ICEt results in deterioration (improvement) in GPt. With the Wald test p-value of 0.493 (>0.10), we cannot reject the null hypothesis that |ω 1| = ω2. Therefore, in this case only, no asymmetry is detected.

IV. Conclusions

Numerous studies ask whether utilizing consumer sentiment data can help predict such economic indi- cators as growth in consumer spending and output. Earlier studies by Garner (1991), Carroll, Fuhrer, and Wilcox (1994), Bram and Ludvigson (1998), and Croushore (2005) offer mixed conclusions. More recent studies including Dees and Brinca (2013) and Christiansen, Eriksen, and Møller (2014), however, find that consumer sentiment has significant predictive power for economic indicators. The study by Christiansen, Eriksen, and Møller (2014) is noteworthy, as it shows that sentiment

indexes help to significantly improve the prediction of U.S. recessions.

Our study is more in line with Dua and Smyth (1993), Blood and Phillips (1995), and Garz (2014), among others, who utilize survey data to analyse public opinion on various aspects of the economy. In particular, a number of studies in the literature have provided empirical evidence in support of the phenomenon known as ‘negativity bias’, whereby people tend to give more weight to negative events than to positive ones. Our study provides additional evidence using consumer sentiment data.

Specifically, we first investigate the dynamic rela- tionship between consumers’ assessment of current and future economic conditions (ICC and ICE), and then examine how these assessments are influenced by deterioration/improvement in consumers’ apprai- sal of economic GP. We further assess how dete- rioration/improvement in ICC and ICE influences GP. We find a number of important results. First, ICC and ICE are cointegrated and thus possess a long-run equilibrium relationship. Second, in the short run within a structural VECM, both ICC and ICE respond to disequilibrium to restore the long- run equilibrium relationship. This means that ICC and ICE are endogenous to each other and, thus, there is feedback.

Third, deterioration in GP results in deterioration in ICC, but improvement in GP has no impact on ICC. Fourth, deterioration in ICC results in deterioration in GP, but improvement in ICC has no influence on GP. In line with psychological negativity bias, consumers

–0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1985 1990 1995 2000 2005 2010 2015

___ Cusum of squares; ---- 5% significance

Figure 6. Parameter stability test results for the GP equation.

APPLIED ECONOMICS 235

appear to pay more attention to negative events than to positive ones. The study by Ito et al. (1998, 887) con- cludes that ‘negative information tends to influence evaluations more strongly than comparably extreme positive information’. Rozin et al. (1997) note that negative events tend to seem more ‘contagious’ and thus have more negative implications. Intuitively, negative situations cause people to try to unravel the ‘why’ underlying the situation, and because more men- tal time is spent on that activity, the negative situation seems to have more weight.

Fourth, deterioration (improvement) in GP results in deterioration (improvement) in ICE. However, consistent with psychological negativity bias, we find that the impact of deterioration in GP on ICE is more than twice larger than the impact of improvement in GP on ICE. The only exception in our findings not consistent with ‘negativity bias’ relates to the impact of ICE on GP. That is, dete- rioration (improvement) in ICE results in deteriora- tion (improvement) in GP, with the impact of deterioration in ICE on GP significantly equal to the impact of improvement in ICE on GP.

Many believe monetary policy, in particular, should anchor market expectations through increased trans- parency and communication (Blinder 1998; Woodford 2001). Lyziak et al. (2007, 67–68) point out that increased transparency ‘. . . involves not only the issue of releasing adequate, in terms of its quantity and quality, information by the central bank, but also cor- rect interpretation of the released information by the public’. In addition, the success of policy depends on whether the public views policymakers as credible. Blinder (2000, 1422) notes that ‘A central bank is credible if people believe it will do what it says’. Our findings that deterioration in GP results in deteriora- tion in both ICC and ICE, and deterioration in both ICC and ICE results in deterioration in GP indicate that, during economic downturns, policymakers have a difficult time rallying public support for turning the economy around. However, our finding that improve- ment in GP (ICE) results in improvement in ICE (GP) is encouraging, since, through increased transparency and communication, policymakers may be able to convince the public that the economic policies put in place will be effective in improving the economy. Such attempts, if successful, would then boost both public trust in economic policies and public expectations about the future of the economy.

Acknowledgement

The authors would like to thank two anonymous referees for helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.

References

Baghestani, H. 1991. “Cointegration Analysis of the Advertising-Sales Relationship.” The Journal of Industrial Economics 39: 671–682. doi:10.2307/2098669.

Bahmani-Oskooee, M. 1998. “Do Exchange Rates Follow a Random Walk Process in Middle Eastern Countries?” Economics Letters 58: 339–344. doi:10.1016/S0165-1765 (98)00013-5.

Blinder, A. S. 1998. Central Banking in Theory and Practice. Cambridge, MA: MIT Press.

Blinder, A. S. 2000. “Central-Bank Credibility: Why Do We Care? How Do We Build It?” American Economic Review 90: 1421–1431. doi:10.1257/aer.90.5.1421.

Blood, D. J., and P. C. B. Phillips. 1995. “Recession Headline News, Consumer Sentiment, the State of the Economy and Presidential Popularity: A Time Series Analysis 1989–1993.” International Journal of Public Opinion Research 7: 2–22. doi:10.1093/ijpor/7.1.2.

Bloom, H. S., and H. D. Price. 1975. “Voter Response to Short-Run Economic Conditions: The Asymmetric Effect of Prosperity and Recession.” The American Political Science Review 69: 1240–1254. doi:10.2307/1955284.

Bram, J., and S. Ludvigson. 1998. “Does Consumer Confidence Forecast Household Expenditure? A Sentiment Index Horse Race.” Economic Policy Review 4: 59–78.

Carroll, C. D., J. C. Fuhrer, and D. W. Wilcox. 1994. “Does Consumer Sentiment Forecast Household Spending? If So, Why?” American Economic Review 84: 1397–1408.

Christiansen, C., J. N. Eriksen, and S. V. Møller. 2014. “Forecasting US Recessions: The Role of Sentiment.” Journal of Banking & Finance 49: 459–468. doi:10.1016/j. jbankfin.2014.06.017.

Croushore, D. 2005. “Do Consumer Confidence Indexes Help Forecast Consumer Spending in Real Time?” The North American Journal of Economics and Finance 16: 435–450. doi:10.1016/j.najef.2005.05.002.

Dees, S., and P. S. Brinca. 2013. “Consumer Confidence as a Predictor of Consumption Spending: Evidence for the United States and the Euro Area.” International Economics 134: 1–14. doi:10.1016/j.inteco.2013.05.001.

Dua, P., and D. J. Smyth. 1993. “Survey Evidence on Excessive Public Pessimism about the Future Behavior of Unemployment.” Public Opinion Quarterly 57: 566–574. doi:10.1086/269396.

236 H. BAGHESTANI AND P. PALMER

Garner, C. A. 1991. “Forecasting Consumer Spending: Should Economists Pay Attention to Consumer Confidence Surveys? Federal Reserve Bank of Kansas City.” Economic Review 76: 57–71.

Garz, M. 2014. “Good News and Bad News: Evidence of Media Bias in Unemployment Reports.” Public Choice 161: 499–515. doi:10.1007/s11127-014-0182-2.

Ito, T. A., J. T. Larsen, N. K. Smith, and J. T. Cacioppo. 1998. “Negative Information Weighs More Heavily on the Brain: The Negativity Bias in Evaluative Categorizations.” Journal of Personality and Social Psychology 75: 887–900. doi:10.1037/0022-3514.75.4.887.

Kahneman, D., and A. Tversky. 1979. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica 47: 263– 291. doi:10.2307/1914185.

Kwiatkowski, D., P. C. B. Phillips, P. Schmidt, and Y. Shin. 1992. “Testing the Null Hypothesis of Stationarity against the Alternative of a Unit Root.” Journal of Econometrics 54: 159–178. doi:10.1016/0304-4076(92)90104-Y.

Lewicka, M., J. Czapinski, and G. Peeters. 1992. “Positive- Negative Asymmetry or ‘When the Heart Needs a Reason’.” European Journal of Social Psychology 22: 425– 434. doi:10.1002/(ISSN)1099-0992.

Lyziak, T., J. Mackiewicz, and E. Stanislawska. 2007. “Central Bank Transparency and Credibility: The Case of Poland, 1998–2004.” European Journal of Political Economy 23: 67–87. doi:10.1016/j.ejpoleco.2006.05.004.

MacKinnon, J. G. 1996. “Numerical Distribution Functions for Unit Root and Cointegration Tests.” Journal of Applied Econometrics 11: 601–618. doi:10.1002/(ISSN)1099-1255.

Rozin, P., J. Haidt, C. R. McCauly, and S. Imada. 1997. “The Cultural Evolution of Disgust.” In Food Preferences and Taste: Continuity and Change, edited by H. M. Macbeth, 65–82. Berghahn: Oxford.

Rozin, P., and E. B. Royzman. 2001. “Negativity Bias, Negativity Dominance, and Contagion.” Personality and Social Psychology Review 5: 296–320. doi:10.1207/ pspr.2001.5.issue-4.

Shin, Y. 1994. “A Residual-Based Test of the Null of Cointegration against the Alternative of No Cointegration.” Econometric Theory 10: 91–115. doi:10.1017/S0266466600008240.

Soroka, S. 2006. “Good News and Bad News: Asymmetric Responses to Economic Information.” The Journal of Politics 68: 372–385. doi:10.1111/j.1468-2508.2006.00413.x.

Tversky, A., and D. Kahneman. 1991. “Loss Aversion in Riskless Choice: A Reference-Dependent Model.” The Quarterly Journal of Economics 106: 1039–1061. doi:10.2307/2937956.

Woodford, M. 2001. “Monetary Policy in the Information Economy.” In Economic Policy for the Information Economy, 297–370. Kansas City: Federal Reserve Bank of Kansas City. A symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming. August 30 - September 1, 2001.

APPLIED ECONOMICS 237

Copyright of Applied Economics is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.