Bad News 3

Journal of Economic Psychology 39 (2013) 426–438
Contents lists available at ScienceDirect
Journal of Economic Psychology

journal homepage: www.elsevier.com/locate/joep
Good news, bad news, consumer sentiment and consumption

behavior
Viet Hoang Nguyen a,⇑, Edda Claus a,b
a
The University of Melbourne, Australia
b
CAMA, Australia National University, Australia
a r t i c l e i n f o a b s t r a c t
Article history: We explore the reaction of heterogenous consumers to a range of financial and economic
Received 23 May 2013 news and find asymmetry in the response to news where consumers react to bad news but
Received in revised form 4 September 2013 not to good news. This asymmetry holds uniformly across heterogeneous consumer groups
Accepted 21 October 2013
and is consistent with the popular negativity bias in psychology. We find asymmetric map-
Available online 29 October 2013
ping from news to consumer sentiment and from consumer sentiment to aggregate con-
sumption that supports the notion of asymmetric consumption behavior.
JEL classification:
Ó 2013 Elsevier B.V. All rights reserved.
D12
D83
D84
E21
PsycINFO classification:
2229
2240
Keywords:
News
Consumer sentiment
Consumption
Negativity bias
1. Introduction
We explore the reaction of heterogenous consumers to news. Few empirical studies in macroeconomics allow for hetero-
geneity between consumers. This is in large part a reflection of the lack of time series data. Detailed panel datasets exist for
many developed economies but their survey frequency is typically low and it is generally not straightforward to match panel
data to national accounts data. In this paper, we proxy consumption behavior with consumer sentiment data. We use the West-
pac–Melbourne Institute Consumer Sentiment Index constructed from a monthly survey of 1200 Australian households.
Respondents are questioned on the short and medium term outlook for the economy, current and future household financial
conditions and are questioned on whether it is a good time to buy major household items. The aggregate Consumer Sentiment
Index (CSI) is the average of the responses to these five questions. Although the relationship between consumer sentiment and
consumption is not one for one, sentiment does map into consumption.1 So the reaction of consumers to news in the sentiment
⇑ Corresponding author. Address: Melbourne Institute of Applied Economic and Social Research, The University of Melbourne, Level 5, FBE Building, 111
Barry Street, VIC 3010, Australia. Tel.: +61 3 903 53621.
E-mail addresses: vietn@unimelb.edu.au (V.H. Nguyen), eclaus@unimelb.edu.au (E. Claus).
1
Among the many studies that show the tight relationship between consumer sentiment and consumption are: Carroll, Fuhrer, and Wilcox (1994), Acemoglu
and Scott (1994), Utaka (2003), Souleles (2004), Bryant and Macri (2005), Chua and Tsiaplias (2009) and Barsky and Sims (2012).
0167-4870/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.joep.2013.10.001
V.H. Nguyen, E. Claus / Journal of Economic Psychology 39 (2013) 426–438 427
survey should be indicative of how consumers adjust their consumption behavior in the wake of news. We explore the reaction of
heterogeneous consumers (disaggregated by age, gender, household income, home ownership and voting intentions) to a range of
financial and economic news. We investigate whether good and bad news have asymmetric effects on consumer sentiment and
whether increases and decreases in consumer sentiment have asymmetric effects on aggregate consumption. Empirical studies
typically assume a symmetric mapping from news to consumer sentiment and consumption.2 But if consumers react asymmet-
rically to good and bad news, assuming symmetry in empirical applications is likely to bias the results.
Consumer sentiment surveys are almost ideal for exploring the behavior of consumers with different characteristics as
these types of survey are representative, monthly as opposed to large scale annual household panels, and span a relatively
long time period. Studies on consumer sentiment typically use consumer sentiment surveys to forecast macro variables or
try to identify drivers of sentiment. We do neither but rather utilize the heterogeneity of respondents to inform behavior of
heterogeneous consumers.
The paper has two main contributions. First, it demonstrates that households react asymmetrically to good and bad news,
supporting the presence of a negativity bias. Second, households with different characteristics react to news differently.
We analyze the effects of good and bad news on the aggregate CSI and the effects of positive and negative changes in the
aggregate CSI on Australian real aggregate consumption, or real household final consumption expenditure which is part of
the Australian national accounts. We then disaggregate consumer sentiment responses to news by age, home ownership,
voting intentions, gender and income. Ideally, we would have liked to link the sentiment responses by disaggregated con-
sumer groups to their consumption expenditures. However, detailed consumption data from the Australian national ac-
counts are only available for different expenditure items such as food, rent or purchases of vehicles. No national accounts
data are available for consumer groups disaggregated by household characteristics such as age, gender or income. Household
survey measures of consumption like the Household Expenditure Survey of the Australian Bureau of Statistics (ABS) (con-
ducted every five years) or the Household Income and Labour Dynamics in Australia (HILDA) survey (conducted every year)
are of low frequencies and do not map directly into national accounts measures of household consumption.
We examine reactions to news about the labor market, monetary policy and the Australian and US share markets and
match these with the consumer sentiment survey. Specifically, news are defined as the actual changes in (i) the Australian
national unemployment rate, (ii) the target cash rate (Australia’s central bank rate set by the Reserve Bank of Australia
(RBA)), (iii) the Australian Standard and Poor’s 200 stock index (S&P 200), and (iv) the US Standard and Poor’s 500 stock index
(S&P 500). For unemployment and target cash rate news, data releases and monetary policy announcements are matched
with survey periods. For example, the December 2011 CSI survey (conducted 5–10 December 2011) is matched with the
0.1% point rise in the unemployment rate for November 2011 reported by the ABS on 8 December 2011. For share price news,
we focus on daily changes during the survey period. Specifically, the daily changes in share prices from Monday 5 December
2011 to Friday 9 December 2011 are matched with the December 2011 CSI survey.
To re-iterate, we define news as the announced changes in the underlying economic and financial series (new public
information) and are abstracting from media news. We analyze consumer reactions to new public information, not to news
surprises. In this context, expectations are not relevant. If the unemployment rate falls or share prices rise, that is good news
regardless of what people expected. The unemployment rate may not have fallen by as much or share prices may not have
increased by as much as people expected but the decline/rise is still good news regardless. Expectations are crucial for sur-
prises (or shocks) but not for the analysis presented here.
Our empirical results provide overwhelming evidence of a negativity bias in the mapping from news to consumer senti-
ment and from sentiment to consumption. Respondents only react to bad news and the negativity bias holds across all spec-
ifications. Also, decreases in consumer sentiment have a negative effect on consumption while increases have no effect.
Gender, home ownership and age do not seem to be important drivers of the behavior examined here. An unexpected result
is the importance of political conviction in consumer sentiment. Respondents are substantially more optimistic if the polit-
ical party they support forms government and are substantially more pessimistic if the party they support is in opposition.
The paper is organized as follows. Section 2 gives a description of the data and the matching of consumer sentiment data
with news and consumption data. Section 3 outlines the empirical models. Section 4 discusses the empirical results, dem-
onstrates the importance of voting intentions and reflects on how the results fit with explanations of the negativity bias put
forward in the literature. Section 5 offers some concluding remarks.
2. Data
Monthly CSI survey data are available from 1992 but data disaggregated by household characteristics are consistently
available only from January 1996. We match the monthly CSI survey data with four different news items between January
1996 and December 2011, and match the monthly survey data with quarterly real consumption between 1992Q1 and
2011Q4. We only investigate the relationship between sentiment and consumption from 1992 onwards, though quarterly
2
There are two contrasting views of the role of consumer sentiment (confidence) in macroeconomics; for a detailed discussion, see Ludvigson (2004) and
Barsky and Sims (2012). The ‘‘animal spirits’’ view suggests that autonomous fluctuations in consumer sentiment have causal effects on economic activity
(Blanchard, 1993; Hall, 1993; Romer, 1990). Meanwhile, the ‘‘news’’ view of consumer sentiment suggests that the relationship between consumer sentiment
and subsequent economic activity arises because consumer sentiment conveys fundamental news (information) about the current and future states of the
economy (Barsky & Sims, 2012; Cochrane, 1994).
428 V.H. Nguyen, E. Claus / Journal of Economic Psychology 39 (2013) 426–438
(a) Aggregate CSI (b) Disaggregate CSI by Age Groups

Young (18-24) Old (45+)
140 140
120 120
100 100
80 80
60 60
96 98 00 02 04 06 08 10 96 98 00 02 04 06 08 10
(c) Disaggregate CSI by Home Ownership (d) Disaggregate CSI by Voting Intention
Mortgage Own Conservative Labor
140 140
120 120
100 100
80 80
60 60
96 98 00 02 04 06 08 10 96 98 00 02 04 06 08 10
(e) Disaggregate CSI by Gender (f) Disaggregate CSI by Income

Male Female Low (<20k) High (>60k)
140 140
120 120
100 100
80 80
60 60
96 98 00 02 04 06 08 10 96 98 00 02 04 06 08 10
⁄ ⁄
Fig. 1. Aggregate and disaggregate consumer sentiment indexes . The two vertical lines in the Sub-figure (d) for the disaggregate CSIs by voting intentions
represent two Australian Federal election dates during the sample period when there were changes in government (2 March 1996 and 24 November 2007).
data are available pre-1992, to be as consistent as possible with our analysis of the relationship between sentiment and
news. Shortening the sample past 1992, leads to too few quarterly observations.
2.1. Consumer sentiment
The CSI survey is stratified on a number of household/respondent characteristics and is based on the University of Mich-
igan Consumer Sentiment Survey. Specifically, the CSI is constructed from five core questions relating to short and medium
term economic conditions, to current and short term family finances and to ‘a good time to buy major household items’.3 The
answers to the survey questions are ordinal, such as ‘better’, ‘same’, ‘worse’ or ‘don’t know’. The responses are cumulated into
five main sub-indexes. Each sub-index is equal to the proportion of optimists (‘better’) minus the proportion of pessimists
(‘worse’) plus 100. This means that 100 can be thought of as the break-even point, when the number of optimists equals the
number of pessimists. And each sub-index is bounded between 0 and 200, with 0 indicating that everybody is pessimistic
and 200 that everybody is optimistic. The aggregate CSI is the average of the five sub-indexes. The CSIs of various house-
hold/consumer groups are constructed analogously.
We disaggregate consumer sentiment by age (young versus old), gender (male versus female), household income (low ver-
sus high), home ownership (own versus mortgage), and voting intentions (Conservative versus Labor). The young group is de-
fined as respondents aged between 18 and 24 years old and the old group as respondents aged 45 years old and above. The
low income group is defined as households earning less than AUD 20,000 per year and the high income group as households
3
For more details on the survey, see Leahy and Summers (2004).
earning AUD 60,000 per year and above. The own group consists of respondents owning their houses outright and the mort-
gage group consists of mortgagees.
Fig. 1 plots the aggregate and disaggregate CSI series. While sentiment indexes disaggregated by home ownership and
gender are similar in level and shape, differences are apparent for the disaggregate series by age, voting intentions and in-
come. Young respondents are always more optimistic than older respondents. The CSI for the young moves around 120 indi-
cating that optimists outweigh pessimists by about 20% over the sample period. It is only during the financial crisis of 2008–
2009, that the sentiment index of the young drops below 100. The old group does not seem to demonstrate a bias as the CSI
of the old moves around the neutral mark of 100.
Perhaps somewhat surprisingly, high income households tend to be more optimistic. Similar to the young, the high in-
come CSI moves around the level of 120 but with greater variation than the young. Low income households tend to be pes-
simistic, where pessimists generally outweigh optimists over the sample period, indicated by values below 100. Also, there is
a clear difference in the CSIs of Conservative and Labor voters, which is discussed in details in Section 4.3.
2.2. News
We focus on four popular economic and financial news items which are readily available at daily and monthly frequen-
cies. These are news about the Australian labor market proxied by changes in the unemployment rate, about monetary policy
identified by changes in the target cash rate, and news about the Australian and the US share markets proxied by changes in
the Australian Standard and Poor’s 200 index and the US Standard and Poor’s 500 index. The data on the Australian national
unemployment rate are from the ABS, the target cash rate from the RBA and the remaining two series are sourced from the
Thomson–Reuter Datastream database.
We assume that the data generating process (DGP) of all underlying series follows a random walk:
st ¼ st1 þ et ; ð2:1Þ
where st is the natural logarithm of the underlying series St ; et is an i:i:d. innovation with zero mean and variance r2e . News
about St between survey periods t 1 and t, is then defined as the change in st between t 1 and t i.e. et ¼ Dst ¼ st st1 .
Alternatively, the DGP could be assumed to follow an autoregressive process of order p (ARðpÞ process) where news about
St between t 1 and t is defined as et ¼ st /1 st1 . . . /p stp and /1 ; . . . ; /p are estimates of /1 ; . . . ; /p . Our empirical find-
ings are robust to either DGPs. For simplicity and without loss of generality, the empirical results presented below are based
on the random walk DGP; the other results are available on request.
2.3. Matching consumer sentiment data with news
Consumers are typically surveyed over a seven-day period at the beginning of each month. An important issue here is to
match correctly the survey period with available news.4 For the unemployment rate and the target cash rate news, we collect
news announced between the last day of the previous survey period ðt 1Þ and the last day of the current survey period (t):

St
et ¼ Dst ¼ ln ; ð2:2Þ
St1
where St is the unemployment or the target cash rate.
For share prices, we focus on daily news during the survey period, assuming that house hold/consumers are more likely to
be affected by recent news.5 Specifically, we collect daily news on the working days of each survey period and aggregate them:
X
Friday X
Friday
Si;t
et ¼ Dsi;t ¼ ln ; ð2:3Þ
i¼Monday i¼Monday
Si1;t
where St is the Australian S&P 200 or the US S&P 500; when day i is Monday, day i 1 is the previous Friday.6
2.4. Matching consumer sentiment data with consumption
Australian real household consumption expenditure is quarterly and sourced from the ABS. To match the quarterly con-
sumption data with the monthly CSI data, we sum monthly changes in the aggregate CSI over the three months of each quar-
ter. Specifically,
4
Casey and Owen (2012) fail to discover asymmetry in consumer reactions. A likely driver of this result is that news are not matched with survey dates.
Survey days and news release days vary so that both sets of days require careful matching.
5
Our empirical findings are similar if news are aggregated from daily news on all working days between the last day of the previous survey period and the
last day of the current survey period.
6
Time lags between Australia and the US do not affect matching the US stock market news with the CSI data because the CSI survey week includes Saturday
and Sunday.
X
March X
June X
September X
December
DcsiMarch ¼ Dcsit ; DcsiJune ¼ Dcsit ; DcsiSeptember ¼ Dcsit ; DcsiDecember ¼ Dcsit ; ð2:4Þ
t¼January t¼April t¼July t¼October
where csit is the monthly natural logarithm of the CSI.
3. Methodology
People have been found to respond asymmetrically to favorable and unfavorable events or news because of cognitive
biases.7 For example, people can have a biased view of the world because they tend to be mildly optimistic (Soroka, 2006) while
results in Fellner and Krügel (2012) suggest that people are overconfident about the precision of private information (and thus
overweighting it) and about the precision of results. This overweighting and miscalibration may be re-enforced by cognitive
weighting (Skowronski & Carlston, 1989; Soroka, 2006 op cit) which suggests that people give too much weight to outliers. In
addition, individuals may also have limited capacity for processing information (Moscarini, 2004; Reis, 2006; Sims, 2003) which
can put a limit on the size and the frequency of updates. The reference point theory is another explanation that has been pro-
posed (see Skowronski & Carlston, 1989). The idea of the reference point is that the interpretation of new events depends on the
current state. The example typically used to describe the theory is that the perception of water temperature depends on body
temperature. Cold water feels colder if experienced with a warm hand rather than a cold hand. All of these cognitive biases can
lead to a misinterpretation of signals and an asymmetric response to news.
To analyze the potentially asymmetric responses of consumers to good and bad news, we decompose et into positive and
negative components as follows:
eþt ¼ maxðet ; 0Þ and et ¼ minðet ; 0Þ; ð3:1Þ
where et is defined as in (2.2) for the unemployment and target cash rates, and as in (2.3) for the share markets. The inter-
pretation of eþ
t and et depends on the type of news. For example, et is bad news with regard to the unemployment rate but
þ
good news with regard to the share markets.

We then estimate the following distributed lag regression for each of the four news items to analyze the responses of
consumers to good and bad news:
X
5
csig;t ¼ c þ ðbþg;i eþti þ bg;i eti Þ þ g;t ; ð3:2Þ
i¼0
where csig;t is the natural logarithm of the CSI; t is an unobserved disturbance, and c denotes a constant term. The subscript
g identifies the group, g ¼ all; age; . . . ; income. We allow for some persistence and let news affect consumer sentiment for up
to six months, i ¼ 0; . . . ; 5.8 bþ=g;i captures the effect of news on the consumer sentiment of group g after i months. A positive
value of the estimate of bþ þ
g;i ðbg;i Þ indicates a positive relationship between eti ðeti Þ and csig;t , i.e. an increase (decrease) in st

increases (decreases) consumer sentiment. A negative value of the estimate of bg;i ðb þ
g;i Þ indicates a negative relationship be-
tween eþti ðeti Þ and csig;t , i.e. an increase (decrease) in st decreases (increases) consumer sentiment.

We estimate the following distributed lag regression to examine whether rises and falls in aggregate sentiment have
asymmetric effects on aggregate consumption:
X
4
þ
Dconsumptiont ¼ c þ ðkþj Dcsitj þ kj Dcsitj Þ þ ut ; ð3:3Þ
j¼0
þ
where Dcsit ¼ maxðDcsit ; 0Þ and Dcsit ¼ minðDcsit ; 0Þ denote rises and falls in the aggregate CSI; Dconsumptiont denotes the
quarterly change in real aggregate consumption; ut is an unobserved disturbance, and kþ= i captures the effect of rises/falls in
consumer sentiment on the aggregate consumption after i quarters. The estimates of kþ= i are expected to be positive, i.e.
rises (falls) in consumer sentiment raise (reduce) aggregate consumption. Assuming that households smooth their consump-
tions, we allow changes in sentiment to affect consumption for up to 5 quarters.9
(3.2) and (3.3) are estimated using OLS with Newey–West HAC standard errors and covariance.
4. Estimation results
Tables 1–8 provide the estimation results. The tables present parameter estimates and their p-values. Regression fits are
þ
proxied by the adjusted R2 , denoted R2 , and two sets of Wald tests. Wald v2ðkÞ denotes the Wald statistic testing the joint
þ
significance of estimates of bi for i ¼ 0; . . . ; k and Wald vðkÞ denotes the Wald statistic testing the joint significance of esti-
2
mates b þ
i . Similar tests are carried out for parameter estimates of ki and ki in (3.3). Significant estimates are in bold and
others are in gray.
7
For a detailed discussion see Soroka (2006) op cit.
8
AIC and SBC suggest between 3 and 5 lags for news.
9
AIC and SBC suggest between 3 and 4 lags.
All parameter estimates for unemployment news are expected to be negative – falls in unemployment (good unemploy-
ment news) are expected to put upward pressure on sentiment and vice versa. All parameter estimates for share market news
are expected be positive – rises in share markets (good share market news) are expected to put upward pressure on senti-
ment and vice versa. The parameter estimates of target cash rate news could be positive or negative depending on the inter-
pretation of consumers.
Results at the aggregate level are discussed in Section 4.1. Section 4.2 examines the results by consumer characteristics
and demonstrates that political affiliation is particularly important in consumer responses to news. The importance of polit-
ical affiliation is discussed in detail in Section 4.3. Of interest are the signs of the parameter estimates that give the direction
of the responses. The sizes of parameter estimates cannot be compared between news items, only within each news item as
the sizes of the underlying news differ.
To summarize, the empirical investigation suggests the following. (1) Asymmetry is present in the response of consumers
to all four news items with respondents only reacting to bad news. (2) This negativity bias holds across consumer groups. (3)
Falls in consumer sentiment have negative effects on consumption but rises have no effects. (4) Target cash rate declines are
seen by consumers as bad rather than good news. (5) Gender, home ownership and age do not seem to be important in con-
sumer reactions. (6) Voting intentions are unexpectedly important. Respondents are substantially more optimistic if the
political party they support forms government and are substantially more pessimistic if the party they support is in
opposition.
Table 1
Effects of different news on the aggregate sentiment.
Unemployment Cash rate S&P 200 S&P 500

c 4.71 4.68 4.71 4.72
bþt
1.65 0.00 0.64 0.21
bþt1
1.26 0.04 0.27 0.10
bþt2
1.13 0.00 0.10 0.37
bþt3
0.72 0.08 0.01 0.03
bþt4
0.09 0.04 0.23 0.04
bþt5
0.20 0.06 0.45 0.10
bt 0.22 0.12 1.62 1.60
bt1 0.12 0.09 1.48 1.52
bt2 0.26 0.04 1.43 1.33
bt3 0.58 0.04 0.76 0.70
bt4 0.53 0.05 0.60 0.78
bt5 0.43 0.02 0.55 0.23

R2 0.16 0.15 0.36 0.32
þ 28.49 2.26 2.36 1.91
Wald

Wald 3.09 82.43 74.45 68.11
The estimation uses Newey-West HAC standard errors and covariance. ‘Bold’ numbers (e.g. 4.71) denote significant estimates at the 10% significance level.
þ
‘Gray’ numbers (e.g. 0.72) denote insignificant estimates at the 10% significance level. Wald v2ð6Þ is the Wald statistic for testing the null hypothesis

bþ
0 ¼ b þ
1 ¼ b þ
2 ¼ b þ
3 ¼ b þ
4 ¼ bþ
5 ¼ 0. Wald v2
ð6Þ is the Wald statistic for testing the null hypothesis b
0 ¼ b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0.
Table 2
Effects of changes in the aggregate sentiment on aggregate consumption.
Coefficient Standard error t statistics p value

c 1.52 0.16 9.23 0.00
kþt
1.44 1.41 1.02 0.31
kþt1
0.84 1.00 0.83 0.41
kþt2
1.72 1.22 1.42 0.16
kþt3
0.97 1.37 0.71 0.48
kþt4
1.06 1.40 0.76 0.45
kt 1.21 1.45 0.83 0.41
kt1 5.45 1.33 4.10 0.00
kt2 5.22 1.55 3.36 0.00
kt3 4.26 1.25 3.41 0.00
kt4 3.38 1.05 3.22 0.00

R2 0.22
þ 7.80 0.13
Wald

Wald 25.20 0.00
þ

kþ þ þ þ þ 2
0 ¼ k1 ¼ k2 ¼ k3 ¼ k4 ¼ 0. Wald vð5Þ is the Wald statistic for testing the null hypothesis k0 ¼ k1 ¼ k2 ¼ k3 ¼ k4 ¼ 0.
Table 3
Effects of unemployment news on the sentiment of heterogenous consumer/household groups.
Young Old Mortgage Own Conservative Labor Male Female Low High
c 4.83 4.67 4.73 4.70 4.83 4.67 4.77 4.67 4.58 4.81
bþt
1.35 1.75 1.66 1.64 2.37 1.17 1.55 1.56 1.45 1.96
bþt1
1.16 1.29 1.40 1.20 2.28 0.94 1.24 1.35 0.54 1.55
bþt2
1.32 1.03 1.03 1.18 2.04 0.21 1.15 1.01 0.33 1.28
bþt3
0.73 0.78 0.89 0.76 1.44 0.27 0.98 0.54 0.25 0.82
bþt4
0.60 0.10 0.14 0.09 0.57 0.29 0.05 0.05 0.41 0.12
bþt5
0.29 0.37 0.15 0.48 0.61 0.09 0.31 0.21 0.15 0.34
bt 0.19 0.43 0.27 0.18 0.09 0.68 0.14 0.29 0.67 0.19
bt1 0.08 0.02 0.07 0.07 0.14 0.08 0.07 0.05 0.13 0.32
bt2 0.10 0.37 0.34 0.30 0.26 0.25 0.32 0.27 0.05 0.23
bt3 0.19 0.56 0.77 0.44 0.51 0.59 0.56 0.55 0.27 0.67
bt4 0.27 0.70 0.64 0.70 0.41 0.69 0.58 0.68 0.64 0.66
bt5 0.08 0.60 0.50 0.50 0.37 0.67 0.58 0.40 0.44 0.56

R2 0.23 0.12 0.12 0.15 0.17 0.01 0.15 0.13 0.03 0.18
þ 34.80 25.82 19.56 27.14 23.10 7.80 25.70 24.43 20.08 27.86
Wald

Wald 2.13 3.87 2.48 2.88 0.51 4.84 2.85 2.64 6.93 4.19
þ

bþ þ þ þ þ þ 2
0 ¼ b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0. Wald vð6Þ is the Wald statistic for testing the null hypothesis b0 ¼ b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0.
Table 4
Effects of target cash rate news on the sentiment of heterogenous consumer/household groups.
c 4.80 4.62 4.71 4.66 4.80 4.63 4.74 4.64 4.55 4.78
bþt
0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.03 0.02 0.04
bþt1
0.02 0.03 0.10 0.02 0.13 0.01 0.05 0.03 0.02 0.09
bþt2
0.04 0.06 0.01 0.06 0.04 0.10 0.03 0.04 0.02 0.01
bþt3
0.02 0.09 0.13 0.08 0.16 0.00 0.10 0.08 0.02 0.14
bþt4
0.08 0.03 0.08 0.05 0.15 0.04 0.07 0.04 0.02 0.08
bþt5
0.04 0.05 0.11 0.01 0.22 0.05 0.06 0.05 0.03 0.09
bt 0.10 0.13 0.13 0.14 0.19 0.07 0.13 0.11 0.06 0.13
bt1 0.09 0.10 0.08 0.10 0.13 0.06 0.09 0.08 0.07 0.10
bt2 0.04 0.03 0.02 0.03 0.04 0.05 0.03 0.03 0.03 0.04
bt3 0.02 0.05 0.04 0.03 0.04 0.04 0.05 0.03 0.05 0.04
bt4 0.09 0.05 0.04 0.07 0.09 0.03 0.07 0.04 0.01 0.05
bt5 0.03 0.01 0.02 0.02 0.09 0.04 0.01 0.03 0.02 0.03

R2 0.24 0.14 0.10 0.19 0.13 0.02 0.17 0.10 0.04 0.13
þ 2.70 3.44 6.06 4.04 4.62 1.88 4.43 2.93 1.48 5.82
Wald

Wald 123.49 77.41 64.14 112.11 49.30 21.19 120.31 56.79 48.11 58.14
þ

bþ
0 ¼ b þ
1 ¼ bþ
2 ¼ bþ
3 ¼ bþ
4 ¼ b þ
5 ¼ 0. Wald v 2
0 ¼ b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0.
4.1. Aggregate responses
Table 1 gives the estimation results for equation (3.2) for g ¼ all – the effects of good and bad news on the aggregate CSI.
The results show asymmetry in the responses to all four news items. Consumers react to bad news as evidenced by param-
eter estimates that are significantly different from zero, but remain unaffected by good news as evidenced by parameter esti-
mates that are statistically insignificant. In particular, consumer sentiment declines following a rise in unemployment and a
fall in Australian and US share prices. But changes in sentiment following a fall in unemployment and a rise in Australian and
US share prices are not statistically significant.
A surprising and robust result is consumer attitudes toward changes in monetary policy implemented through changes
in the target cash rate. Decreases in the target cash rate are seen as bad news demonstrated by positive and significant
parameter estimates for decreases. Mortgage rates generally move in tandem with changes in the target cash rate. Given
that a decline in mortgage rates means a decline in mortgage payments, one would have expected that a decline in the
target cash rate would put upward pressure on confidence. This is not supported by the empirical results here. Declines in
Table 5
Effects of Australian S&P 200 news on the sentiment of heterogenous consumer/household groups.
c 4.83 4.65 4.74 4.69 4.94 4.56 4.76 4.67 4.57 4.82
bþt
0.43 1.01 0.14 0.96 1.41 2.33 0.78 0.53 0.49 0.12
bþt1
0.58 0.47 0.07 0.45 1.73 2.15 0.78 0.14 0.85 0.24
bþt2
0.43 0.15 0.24 0.45 1.96 1.94 0.05 0.24 0.22 0.09
bþt3
0.00 0.15 -0.03 0.02 2.05 1.94 0.00 0.19 0.65 0.17
bþt4
0.58 0.23 0.66 0.36 2.35 1.39 0.57 0.23 0.41 0.82
bþt5
0.13 0.67 0.75 0.66 2.61 0.75 0.39 0.90 0.51 1.18
bt 1.66 1.42 1.93 1.54 3.42 0.04 1.47 1.55 1.46 1.98
bt1 0.94 1.61 1.70 1.52 3.02 0.23 1.40 1.50 1.07 1.88
bt2 0.86 1.66 1.70 1.51 2.70 0.91 1.58 1.42 1.18 1.72
bt3 0.97 0.65 0.45 0.78 1.27 0.36 0.74 0.60 0.07 0.70
bt4 1.14 0.60 0.69 0.55 1.54 0.07 0.78 0.50 0.47 0.90
bt5 0.55 0.69 0.70 0.68 1.30 0.37 0.55 0.87 0.29 0.84

R2 0.40 0.30 0.32 0.34 0.55 0.19 0.34 0.32 0.19 0.39
þ 4.86 4.91 2.00 7.27 25.26 18.50 4.99 3.70 5.04 4.05
Wald

Wald 88.67 68.38 53.97 76.16 112.99 7.21 77.61 55.87 50.52 71.41
þ

Table 6
Effects of US S&P 500 news on the sentiment of heterogenous consumer/household groups.
c 4.85 4.68 4.73 4.72 4.92 4.60 4.78 4.68 4.58 4.83
bþt
0.51 0.18 0.19 0.43 0.83 1.03 0.24 0.23 0.09 0.02
bþt1
0.03 0.29 0.20 0.30 1.51 0.97 0.10 0.23 0.12 0.22
bþt2
0.07 0.68 0.30 0.39 1.77 0.99 0.35 0.07 0.30 0.35
bþt3
0.06 0.01 0.34 0.06 1.01 1.10 0.10 0.27 0.53 0.09
bþt4
0.74 0.14 0.03 0.24 1.19 0.86 0.10 0.15 0.11 0.28
bþt5
0.37 0.18 0.21 0.03 1.04 1.04 0.28 0.06 0.20 0.03
bt 0.97 1.74 1.74 1.71 2.45 0.65 1.69 1.47 1.40 1.95
bt1 1.37 1.68 1.74 1.82 2.62 0.74 1.76 1.46 1.26 1.89
bt2 1.05 1.33 1.42 1.32 2.30 1.04 1.38 1.16 0.76 1.68
bt3 1.19 0.60 0.45 0.74 1.25 0.35 0.64 0.61 0.27 0.74
bt4 1.19 0.93 0.73 0.86 1.32 0.48 0.93 0.73 0.74 0.96
bt5 0.45 0.25 0.16 0.23 0.34 0.30 0.26 0.21 0.11 0.18

R2 0.39 0.27 0.25 0.34 0.31 0.12 0.35 0.24 0.13 0.35
þ 6.49 4.01 1.53 4.26 9.21 4.83 2.40 1.41 2.81 1.54
Wald

Wald 138.66 55.48 43.44 84.85 35.92 9.08 83.36 46.22 36.11 73.20
þ

the target cash rate put downward pressure on sentiment. This finding is robust for all consumer/household characteris-
tics presented below. This means that this peculiar result cannot be driven by a wealth effect because a decline in the
target cash rate would translate into a decline in investment income which is likely to affect the sentiment of high income
households and home owners only, but not the sentiment of other groups. A possible explanation for the finding here is
that consumers may view loosenings in monetary policy as signals of weakness in economic activity ahead, leading to a
decline in sentiment. This empirical result has an important implication for monetary policy. It suggests that there may be
a ‘consumer sentiment channel’ in the transmission mechanism of monetary policy and this sentiment channel may
potentially be important.
The Wald tests further confirm the asymmetry in the response of consumers to good and bad news. The null hypothesis of
joint insignificance of parameter estimates of bad news is overwhelmingly rejected while the null hypothesis of joint insig-
nificance of parameter estimates of good news is not rejected in all four news items. The results also show that bad share
markets news explain movements in consumer sentiment more than bad unemployment and target cash rate news. This
difference may result from the different frequencies of news. Share markets news are daily and might convey more infor-
mation pertinent to consumer sentiment.
Table 7
Effects of news on the sentiment of conservative and labor voters.
Unemployment news Cash rate news S&P 200 news S&P 500 news
Conservative Labor Conservative Labor Conservative Labor Conservative Labor
c 4.86 4.65 4.82 4.62 4.84 4.64 4.85 4.65
d 0.26 0.15 0.26 0.15 0.23 0.18 0.25 0.15
bþt
1.32 1.78 0.07 0.07 0.76 0.61 0.47 0.23
bþt1
1.20 1.56 0.00 0.06 0.60 0.31 0.08 0.09
bþt2
0.90 0.87 0.07 0.04 0.36 0.10 0.25 0.05
bþt3
0.29 0.93 0.02 0.08 0.20 0.15 0.41 0.22
bþt4
0.09 0.09 0.02 0.05 0.13 0.57 0.50 0.18
bþt5
0.05 0.46 0.02 0.06 0.27 1.10 0.42 0.14
bt 0.03 0.72 0.10 0.12 1.46 1.51 1.29 1.37
bt1 0.16 0.08 0.06 0.10 1.10 1.75 1.42 1.48
bt2 0.36 0.20 0.01 0.06 1.44 1.91 1.32 1.64
bt3 0.73 0.47 0.02 0.05 0.63 0.86 0.36 0.90
bt4 0.74 0.49 0.02 0.07 0.63 0.66 0.48 1.01
bt5 0.64 0.51 0.02 0.01 0.64 0.89 0.13 0.43

R2 0.69 0.29 0.68 0.31 0.76 0.41 0.74 0.40
þ 14.60 24.15 1.85 3.80 2.79 4.68 4.27 0.90
Wald

Wald 3.98 4.01 26.28 58.81 26.69 32.23 46.86 40.65
þ

bþ
0 ¼ b þ
1 ¼ bþ
2 ¼ bþ
3 ¼ bþ
4 ¼ b þ
5 ¼ 0. Wald v 2
0 ¼ b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0.
Table 8
Effects of news on the sentiment of high and low income households.
Unemployment news Cash rate news S&P 200 news S&P 500 news
Low High Low High Low High Low High
c 4.58 4.82 4.55 4.79 4.59 4.79 4.59 4.80
d 0.00 0.09 0.00 0.10 0.05 0.08 0.02 0.09
bþt
1.46 1.58 0.02 0.01 0.04 0.82 0.01 0.48
bþt1
0.54 1.17 0.02 0.04 0.37 0.52 0.03 0.29
bþt2
0.33 0.87 0.02 0.03 0.70 0.66 0.40 0.19
bþt3
0.25 0.41 0.02 0.09 0.19 0.57 0.44 0.60
bþt4
0.41 0.12 0.02 0.02 0.10 0.01 0.22 0.32
bþt5
0.15 0.11 0.03 0.02 0.99 0.42 0.29 0.55
bt 0.67 0.17 0.06 0.09 1.86 1.34 1.47 1.54
bt1 0.13 0.32 0.07 0.07 1.46 1.25 1.34 1.47
bt2 0.05 0.26 0.03 0.03 1.44 1.31 0.83 1.33
bt3 0.27 0.75 0.05 0.03 0.06 0.49 0.21 0.43
bt4 0.64 0.79 0.01 0.03 0.66 0.60 0.80 0.66
bt5 0.44 0.66 0.02 0.00 0.42 0.63 0.09 0.10

R2 0.02 0.32 0.04 0.27 0.21 0.43 0.13 0.46
þ 16.73 17.50 1.54 3.27 6.07 3.32 3.38 4.83
Wald

Wald 6.72 5.41 39.50 26.81 35.20 22.32 33.06 59.76
þ

Importantly, we find that asymmetry extends to the effects of changes in consumer sentiment on consumption.10 Table 2
provides the estimation results for Eq. (3.3) – the effects of rises and falls in consumer sentiment on aggregate consumption. The
results show that falls in consumer sentiment are associated with declines in real household consumption while no relationship
seems present between rises in the consumer sentiment and consumption. Moreover, falls in consumer sentiment do not affect
consumption contemporaneously but with lags. This may be a reflection of the consumption smoothing behavior of households.
The Wald test results further confirm the asymmetric mapping from consumer sentiment to consumption. Variations in senti-
10
Ideally, the analysis should also be conducted on disaggregate sentiment and disaggregate consumption data but no national accounts measures of
consumption by household characteristics are available. Household survey measures of consumption such as the Household Expenditure Survey of the Australian
Bureau of Statistics (conducted every five years) or Household Income and Labour Dynamics in Australia (HILDA) of the Melbourne Institutes (conducted every
year) do not map into national accounts measures; see Australian Bureau of Statistics Household Expenditure Survey 2009–2010 (Cat. No. 6530.0) Summary
Results Appendix 3.
ment, mainly declines in sentiment, explain about 22% of quarterly variations in real household consumption over the sample
period.
Note that we do not attempt to examine whether movements in sentiment explain fluctuations in consumption better
than other factors or whether news are the only driver of sentiment. Rather, our results suggest that the asymmetric effects
of movements in sentiment on consumption result from the asymmetric responses of consumers to good and bad news.
4.2. Responses by heterogenous consumer/household groups
We examine whether the asymmetric mapping from news to sentiment at the aggregate level is also present at the dis-
aggregate level. Tables 3–6 give the estimation results for Eq. (3.2) – the responses by heterogeneous consumer/household
groups (g = age, home ownership, voting intentions, gender and household income) to each of the four news items. The tables
present the same statistics as in Table 1 and the expected signs of parameter estimates are the same as those in Table 1.
4.2.1. Unemployment news

Table 3 presents the effects of unemployment news on the sentiment of different consumer groups.
The disaggregate results are in line with the aggregate results. All groups display asymmetric responses, reacting nega-
tively to rises in unemployment but not to falls in unemployment. The magnitude of the responses is similar for the young
and old, for mortgagees and owners, and for males and females. The biggest difference in responses is between groups with
different political affiliation, Conservative (a coalition between Australian Liberal and National Parties) voters and Labor
(Australian Labor Party) voters. It is not immediately clear what drives this result. It cannot be that voting behavior is driven
by particular respondent characteristics such as age or gender. If this was the case, then the difference in responses would
also be visible by these characteristics. But results in Table 3 do not show much difference in responses by age or gender.
There are some differences in responses of high and low income households. Estimation results discussed below suggest that
high income voters tend to be more affiliated with the Conservative rather than the Labor side of politics. But this cannot be
the only explanation, because the difference in the responses of low and high income households is substantially smaller
than the difference in the responses of Conservative and
Labor voters. Furthermore, the Wald test results confirm the asym-
metry in all groups except for Labor voters while the R2 of the equation for Labor voters is negative. We will demonstrate in
Section 4.3 below that the puzzling results surrounding voting intentions is driven by the reaction to whether one’s party has
formed government.
4.2.2. Target cash rate news

Table 4 presents the effects of target cash rate news on the sentiment of different groups.
All estimation and test results are consistent with the estimation and test results at the aggregate level. Households react
significantly to declines but not rises in the target cash rate and declines decrease consumer sentiment, evidenced by the
positive and statistically significant parameter estimates of b ti across all household/consumer groups. The consistent re-
sponses across household/consumer groups suggest that a wealth effect is unlikely to be the driver behind the negative re-
sponse of consumers to declines in the target cash rate. Perhaps, households interpret declines in the target cash rate as
signals of weak economic activity ahead, and thus view them as bad news. Similar to the responses to unemployment news,
considerable magnitude differences are apparent in the responses of Conservative and Labor voters, and in the responses of
low and high income households to target cash rate news.
4.2.3. Australian S&P 200 news

Table 5 presents the impact of domestic share market news, proxied by daily news about the Australian S&P 200 index, on
the sentiment of different consumer groups.
The estimation and test results based on disaggregate CSI data are consistent with those from the aggregate CSI, except for
the results by voting intentions. In general, consumers react to bad but not good domestic share market news and bad news
put downward pressure on sentiment. The magnitudes of parameter estimates are similar across age, gender and home own-
ership groups. High income households seem more responsive to bad share market news than low income households, as
one might expect.
The estimation and test results for Conservative and Labor voters are puzzling. Conservative voters react negatively to
both good and bad domestic share market news as evidenced by statistically significant estimates of the effects of good
and bad news. The Wald test results confirm that both good and bad news decrease Conservative voters’ sentiment. Mean-
while, Labor voters only react to rises but not falls in domestic share prices, but increases in share prices put upward pressure
on sentiment unlike the reaction of Conservative voters.
4.2.4. US S&P 500 news

Table 6 presents the effects of international share market news, proxied by daily news about the US S&P 500 index, on the
sentiment of different consumer groups.
The estimation and test results by age, gender, home ownership and income are consistent with the aggregate results and
align with those for domestic share market news. Reactions are asymmetric, where households react to falls in the US share
prices but not rises. The magnitudes of parameter estimates are similar by age, gender and home ownership but different by
income – high income households react stronger to bad share market news. Again, the results by voting behavior are some-
what puzzling. US share market news do not have much impact on the sentiment of Labor voters while Conservatives react
negatively to both good and bad news.
4.3. Conservative versus Labor voters
Over the sample period, there are two Australian Federal Election dates where governments changed, they are 2 March
1996 and 24 November 2007. On both dates, the incumbent lost the election: the government switched from Labor to Con-
servative in 1996 and from Conservative to Labor in 2007. Fig. 1 plots the aggregate and disaggregate CSI series. By far the
most interesting graph is the CSI disaggregated by voting intentions in Fig. 1(d). The two vertical lines in the graph denote
the two Federal Election dates when government changes occurred. The graph shows a bias related to voting intentions. Con-
servative voters are consistently substantially more optimistic than Labor voters prior to 2007 and the opposite is true post
2007. In fact, this switch happened at the 24 November 2007 Australian Federal Election when the Conservative (a coalition
of the Liberal and National Parties – LNP) government led by Prime Minister John Howard was ousted by the Australian Labor
Party (ALP) led by Kevin Rudd. Up until the last survey prior to the 2007 election, conducted from 5 to 11 November 2007,
Conservatives were consistently more optimistic than Labor voters but Conservatives have been consistently more pessimis-
tic than Labor voters since the first post election survey conducted from 3 to 9 December 2007. A similar switch in the level
of confidence occurred at the beginning of the sample period, at the 2 March 1996 Federal Election when ALP Prime Minister
Paul Keating was ousted by Liberal Party (LP) leader John Howard. The graph indicates that respondents are more optimistic
if the party they support has formed government and are more pessimistic if it has not, ceteris paribus. This is a completely
unexpected result and is in stark contrast to what rational expectations would suggest.11
To formally investigate the influence of the change in government on consumer sentiment, we only focus on the Novem-
ber 2007 Federal Election, because the earlier change in government occurred close to the sample starting point. We re-esti-
mate all models with a dummy variable for a change in government in November 2007. In particular, the dummy variable
takes a zero value prior to November 2007 and one from November 2007 onwards.12 This dummy variable is expected to have
a negative impact on the sentiment of Conservative voters but a positive impact on the sentiment of Labor voters.
To conserve space, the estimation and test results with a dummy variable are presented in Table 7 for Conservative and
Labor voters, and in Table 8 for high and low income households. All the other results are available from the authors. Table 7
shows that with the inclusion of the 2007 election dummy, the puzzling results in the responses of Conservative and Labor
voters discussed above have disappeared. All the estimation and test results by voting intentions are now consistent with
those at the aggregate level. To summarize: (i) the election dummy is insignificant at the aggregate level across all types
of news; (ii) the election dummy is significant at the disaggregate level across all types of news for only 3 groups: Conser-
vative and Labor voters, and high income households. Specifically, the estimates of the dummy suggest that the 2007 change
in government has a negative impact on the sentiment of Conservative voters while it has a positive impact on the sentiment
of Labor voters. The results also suggest an association between

Conservative and high income voters in Australia. In addi-
tion, the inclusion of the 2007 election dummy increases the R2 of equations for Conservative and Labor voters and high in-
come households considerably across all types of news.
4.4. Potential drivers of the negativity bias
The empirical evidence presented here naturally leads to the question of the drivers of the bias. Our results suggest that
consumers react only to bad news but not good news. This negativity bias has been extensively documented. For example,
studies in psychology find that unfavorable information has a stronger impact on impressions than favorable information
(e.g. Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Rozin & Royzman, 2001; Skowronski & Carlston, 1989; Vonk,
1996). In politics, Bloom and Price (1975) and Lau (1985) provide evidence that negative information has a greater influence
on voting behavior than positive information. In economics, Kahneman, Knetsch, and Thaler (1991) stress that the response
of individuals to unfavorable changes is more intense than it is to favorable changes.
A number of explanations for the drivers of the bias have been put forward. The negativity bias may be a product of evo-
lution as attention to negative events makes survival more likely and increases the probability to passing on genes. ‘‘Survival
requires urgent attention to possible bad outcomes, but it is less urgent with regards to good outcomes.’’ (Baumeister et al.,
2001, p. 325). The negativity bias may also result from loss aversion, which refers to the notion that people care more about a
loss in utility than a gain of equal magnitude (e.g. Kahneman & Tversky, 1979).
Our finding that consumers react more strongly to bad news rather than to goods news is in line with the loss aversion
literature although loss aversion does not explain the absence of reaction to good news found here. The reaction to good
11
These results contradict those presented in Barsky and Sims (2012) who find that confidence innovations are well characterized as noisy measures of
changes in expected productivity growth. However, the empirical application seems to include variables at different order of integration evidenced in the
impulse response analysis presented in the paper which casts doubt on the results.
12
We also re-estimate all models with a dummy variable that takes into account both election dates. The estimates on this dummy are insignificant in almost
all specifications, as one might expect due to the opposite impacts of the two election dates on the sentiment of Conservative and Labor voters. Moreover, the
estimation based on two dummy variables for two election dates yields quantitatively similar results as those presented in Tables 7 and 8.
news may be muted by the presence of costs associated with acquiring, absorbing and processing information (see Reis,
2006). This may put a threshold on the size and frequency of good news that elicit consumers reaction. We also completely
abstract from the news media. If people’s expectation formation is influenced by media (see Carroll et al., 1994; Casey &
Owen, 2012; Dom & Morin, 2004), a negativity bias in the news media (Harrington, 1989) may be a driver of the consumers’
apparent negativity bias.
A draw-back of the analysis presented here is that we cannot quantify the news which makes it impossible to compare
consumer reactions across news items. An interesting and useful extension would be to compare consumer reactions to
unemployment versus interest rate news. Another extension of the paper is to examine the presence of a threshold where
consumers accumulate evidence about the state of the economy before they adjust their beliefs and hence responses. Esti-
mation within a Bayesian framework may reveal a tight prior distribution for favorable events and news but flat priors for
unfavorable events and news. Both extensions are left for future work.
5. Concluding remarks
This paper analyzes the effects of news on the sentiment of heterogeneous consumers and the effects of changes in con-
sumer sentiment on consumption. In line with the psychology literature, we find asymmetry in response to various eco-
nomic and financial news where consumers react to bad but not to good news. Similarly, we find that falls in consumer
sentiment decrease consumption while rises have no effects.
In standard macroeconomic models, households react symmetrically to shocks and adjust their consumption symmetri-
cally. Our empirical results suggest that this may not be the case in practice: positive and negative shocks of the same mag-
nitude seem to have different impacts on household consumption behavior. This potential presence of a negativity bias is
particularly important if policy makers use fiscal stimuli to offset declines in consumption following negative shocks as
was the course of action of many developed economies affected by the 2007–2008 financial crisis.
Furthermore, our results indicate that sentiment differs among income groups and most surprisingly among consumers
with different political views. In particular, consumers are consistently more optimistic if the political party they support has
formed government and are consistently more pessimistic if the party they support is in opposition.
Acknowledgements
We are grateful for comments and suggestions from Iris Claus and two anonymous referees. Claus acknowledges funding
from The University of Melbourne Faculty of Business and Economics.
References
Acemoglu, D., & Scott, A. (1994). Consumer confidence and rational expectations: Are agents’ beliefs consistent with the theory? Economic Journal, 104(422),
1–19.
Barsky, R. B., & Sims, E. R. (2012). Information, animal spirits, and the meaning of innovations in consumer confidence. American Economic Review, 102(4),
1343–1377.
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370.
Blanchard, O. J. (1993). Consumption and the recession of 1990–1991. American Economic Review, 83(2), 270–274.
Bloom, H. S., & Price, H. D. (1975). Voter response to short-run economic conditions: The asymmetric effect of prosperity and recession. The American
Political Science Review, 69(4), 1240–1254.
Bryant, W. A., & Macri, J. (2005). Does sentiment explain consumption? Journal of Economics and Finance, 29(1), 97–111.
Carroll, C. D., Fuhrer, J. C., & Wilcox, D. W. (1994). Does consumer sentiment forecast household spending? If so, why? American Economic Review, 84(5),
1397–1408.
Casey, G. P., & Owen, A. L. (2012). Good news, bad news, and consumer confidence. Social Science Quarterly. http://dx.doi.org/10.1111/j.1540-
6237.2012.00900.x.
Chua, C. L., & Tsiaplias, S. (2009). Can consumer sentiment and its components forecast Australian GDP and consumption? Journal of Forecasting, 28(8),
698–711.
Cochrane, J. H. (1994). Shock. Carnegie-Rochester Conference Series on Public Policy, 41, 295–364.
Dom, M., & Morin, N. (2004). Consumer sentiment, the economy, and the news media. Federal Reserve Bank of San Francisco Working Paper 2004–09.
Fellner, G., & Krügel, S. (2012). Judgmental overconfidence: Three measures, one bias? Journal of Economic Psychology, 33(1), 142–154.
Hall, R. R. (1993). Macro theory and the recession of 1990–1991. American Economic Review, 83(2), 275–279.
Harrington, D. E. (1989). Economic news on television. Public Opinion Quarterly, 53(1), 17–40.
Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1991). The endownment effect, loss aversion, and status quo bias. Journal of Economic Perspectives, 5(1),
193–206.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–292.
Lau, R. R. (1985). Two explanations for negativity effects in political behavior. American Journal of Political Science, 29(1), 119–138.
Leahy, A., & Summers, P. (2004). Consumer surveys of the Melbourne Institute of Applied Economic and Social Research. Australian Economic Review, 37(4),
446–453.
Ludvigson, S. C. (2004). Consumer confidence and consumer spending. Journal of Economic Perspectives, 18(2), 29–50.
Moscarini, G. (2004). Limited information capacity as a source of inertia. Journal of Economic Dynamics & Control, 28(10), 2003–2035.
Reis, R. (2006). Inattentive consumer. Journal of Monetary Economics, 53(8), 1761–1800.
Romer, C. (1990). The great crash and the onset of the great depression. The Quarterly Journal of Economics, 105(3), 597–624.
Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5(4), 296–320.
Sims, C. A. (2003). Implications of rational inattention. Journal of Monetary Economics, 50(3), 665–690.
Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in impression formation: A review of explanations. Psychological Bulletin, 105(1),
131–142.
Soroka, S. N. (2006). Good news and bad news: Asymmetric responses to economic information. The Journal of Politics, 68(2), 372–385.
Souleles, N. (2004). Expectations, heterogenous forecast errors and consumption: micro-evidence from the Michigan consumer sentiment surveys. Journal
of Money, Credit and Banking, 36(1), 39–72.
Utaka, A. (2003). Confidence and the real economy: The Japanese case. Applied Economics, 35(3), 337–342.
Vonk, R. (1996). Negativity and potency effects in impression formation. European Journal of Social Psychology, 26(6), 851–865.

Bad News 3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bad News 3

Uploaded by

Copyright:

Available Formats

Journal of Economic Psychology 39 (2013) 426–438

Contents lists available at ScienceDirect

Journal of Economic Psychology

Good news, bad news, consumer sentiment and consumption

(a) Aggregate CSI (b) Disaggregate CSI by Age Groups

(e) Disaggregate CSI by Gender (f) Disaggregate CSI by Income

2.1. Consumer sentiment

2.3. Matching consumer sentiment data with news

2.4. Matching consumer sentiment data with consumption

where csit is the monthly natural logarithm of the CSI.

good news with regard to the share markets.

Unemployment Cash rate S&P 200 S&P 500

Coefﬁcient Standard error t statistics p value

4.1. Aggregate responses

4.2. Responses by heterogenous consumer/household groups

4.2.1. Unemployment news

4.2.2. Target cash rate news

4.2.3. Australian S&P 200 news

4.2.4. US S&P 500 news

4.3. Conservative versus Labor voters

4.4. Potential drivers of the negativity bias

You might also like