Popular in Social Sciences

Political Science

V-Dem’s Conceptions of Democracy and Their Consequences

University of Notre Dame Michael Coppedge

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Social Identities and Divisions

A new explanation of antisemitism: Jew hatred as a civilisational phenomenon

Boston University Liah Greenfeld

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

American Politics

Financial Expectations and Household Consumption: Does Middle-Inflation Matter?

View more insights in Social Sciences - Economics - Microeconomics

View more figures from University of Sheffield

View the page of Sarah Brown

Read the Study

Next in Economics

Do renewable energy investments create local jobs?

Natalia Fabra | University Carlos III Madrid

An early indicator for anomalous stock market performance

Thomas Gries | Universität Paderborn

Sarah Brown explores the finding that households often expect their financial position to remain unchanged compared to other alternatives.

Image courtesy of interviewee. April 24, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video video discusses the use and implications of survey data, particularly in economics and social sciences, focusing on individuals' financial expectations. It highlights a common tendency for respondents to select neutral or middle options in surveys, a phenomenon referred to as "middle inflation." The authors present a study examining financial expectations by analyzing survey responses from the British Household Panel Survey from 1999 to 2008, focusing on how individuals anticipate their financial situation a year ahead.

The findings reveal that a significant portion (27%) of responses in financial expectation surveys can be attributed to middle inflation, suggesting that respondents often align with middle-ground sentiments when expressing their outlook. The study introduces a new statistical framework (generalized middle inflated ordered probit model) to better capture the role of these middle responses and to assess their impact on household consumption behaviors.

The research contrasts traditional methods, which treat financial expectations as fixed variables, by allowing these expectations to be influenced by various psychological and personal factors. This video video consideration leads to findings that existing models may overestimate the impact financial expectations have on consumption, particularly highlighting a 38% overestimate regarding expenditure on durable goods when ignoring middle inflation.

The authors emphasize the importance of accurately modeling financial expectations for better predictions of household consumption patterns, emphasizing implications for policymakers, especially in the context of economic changes and consumer behavior during crises like inflation. This video video concludes by suggesting future research directions to investigate how recent discussions on inflation might affect household expectations.

Language:

Export as PDF

Export as PDF
Here are some key quotes from the video you that highlight the main themes, findings, and implications of the research:

1. **Convideo of Survey Data**:
- "So survey data is widely used in economics and right across the social sciences more generally."

2. **Neutral Responses**:
- "There is a tendency for 'em to choose a middle category, a kind of neutral category."
- "People sort of just go in the middle."

3. **Focus on Financial Expectations**:
- "We focus on financial expectations... looking ahead, how do you think you'll be financially a year from now?"

4. **Middle Inflation Concept**:
- "That's what was referred to as middle inflation, whereby we get this grouping, this concentration in the middle category."

5. **Statistical Modeling Contribution**:
- "We introduce a new statistical modeling framework whereby we can quantify how much of a survey response is likely to be related to overinflation rather than...a true belief."

6. **Implications on Consumption**:
- "If I think things are going to get better financially... I might go on the spending spree now... however, if I think things are gonna get worse financially, I may start cutting back now."

7. **Estimation of Middle Inflation Impact**:
- "We find that 27% of responses that we observe in our expectations data can be attributed to the impact of middle inflation."

8. **Overestimating Financial Expectations**:
- "Specifications...which do not explicitly control for middle inflation, tend to overestimate the effect of sentiment consumer expectations on expenditure."

9. **Research Importance**:
- "If we think about policymakers...we're trying to work out the relationship between expectations and things like household consumption."

10. **Future Research Directions**:
- "It will be interesting to see whether or not households’ financial expectations on more or less accurate going forward."

These quotes encapsulate the main findings of the research regarding the significance of middle inflation in survey responses related to financial expectations, how this affects consumption behavior, and the implications for policymakers.

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.085 --> 00:01.225
So just a little bit of context.

00:01.805 --> 00:05.105
So survey data is widely used in economics

00:05.245 --> 00:07.825
and right across the social sciences more generally.

00:08.925 --> 00:11.785
So we think about surveys when households

00:11.785 --> 00:14.065
and individuals could be asked their opinions

00:14.065 --> 00:16.105
and attitudes about all sorts of things,

00:16.215 --> 00:18.585
political preferences, immigration,

00:19.125 --> 00:20.425
as well as their finances.

00:20.485 --> 00:23.225
So you name it. Um, anything could be in the survey.

00:24.885 --> 00:27.745
So when individuals are faced with survey questions,

00:27.745 --> 00:30.825
there is a tendency for 'em to choose a middle category,

00:30.905 --> 00:32.385
a kind of neutral category.

00:32.965 --> 00:37.105
So what we tend to observe is a concentration of responses

00:37.245 --> 00:39.465
by individuals in the middle category.

00:40.085 --> 00:42.505
So if you present somebody with a statement,

00:42.855 --> 00:45.425
they very often say they neither agree

00:45.485 --> 00:47.665
or disagree with a particular statement.

00:49.115 --> 00:51.535
So in our paper, as you say, we met,

00:51.535 --> 00:53.775
we focus on financial expectations

00:54.515 --> 00:58.495
and specifically, individuals are asked a, a,

00:58.615 --> 01:01.735
a question looking ahead, how do you think you'll,

01:01.735 --> 01:03.895
you'll be financially a year from now?

01:04.235 --> 01:05.655
So it's a very general question,

01:06.675 --> 01:09.655
and when we look at the responses to this, people can say

01:09.655 --> 01:11.975
that they're, they think they're gonna be worse off about

01:11.975 --> 01:13.455
the same or better off.

01:14.595 --> 01:17.535
And what we find is about 10%

01:17.755 --> 01:20.295
of individuals think they're gonna be worse

01:20.395 --> 01:21.575
off in a year's time.

01:22.105 --> 01:25.375
About 61% say they think they're gonna be about the same,

01:25.715 --> 01:27.455
and the remainder about,

01:27.455 --> 01:29.775
28% think they're gonna be better off.

01:30.355 --> 01:33.295
So this clump in the middle whereby people say,

01:33.295 --> 01:34.975
I just think I'm gonna be about the same.

01:35.275 --> 01:38.135
That's what was referred to as middle inflation,

01:38.645 --> 01:40.095
whereby we get this grouping,

01:40.095 --> 01:42.455
this concentration in the middle category.

01:43.275 --> 01:45.815
And that's, something that happens with lots

01:45.815 --> 01:47.095
of different survey questions,

01:47.155 --> 01:49.095
not just ones about financial expectations,

01:49.395 --> 01:51.815
but quite generally, people sort of just go in the middle.

01:53.575 --> 01:55.635
So there's a very small literature

01:55.665 --> 01:58.275
that looks at this characteristic of survey data.

01:58.685 --> 02:00.075
There are loads and loads of papers

02:00.145 --> 02:03.395
that use survey data in economics and social sciences,

02:03.695 --> 02:06.115
but there's very few papers actually think about the

02:06.275 --> 02:07.755
modeling implications of it.

02:08.695 --> 02:12.475
So what we actually do in our paper is we introduce a new

02:12.705 --> 02:16.915
statistical modeling framework whereby we can quantify

02:17.415 --> 02:21.155
how much of a survey response is likely to be related

02:21.415 --> 02:25.435
to overinflation rather than, for example, a true belief

02:25.625 --> 02:28.675
that I really think my income will be unchanged.

02:30.015 --> 02:34.815
So the other literature which we relate to is,

02:35.045 --> 02:38.215
that literature on individual and household consumption.

02:38.915 --> 02:41.015
So the contribution is kind of twofold.

02:41.515 --> 02:43.695
The first thing is to actually think about

02:43.715 --> 02:46.495
how should we model this type of survey data

02:46.705 --> 02:49.535
where we get this kind of clump in the middle response.

02:50.475 --> 02:52.015
The second thing that we do is

02:52.015 --> 02:56.255
that we look at the implications of how we model survey data

02:56.995 --> 02:58.655
for a particular outcome

02:59.075 --> 03:01.005
Such as household consumption.

03:02.185 --> 03:04.965
So if we think about expectations in household consumption,

03:05.505 --> 03:08.245
you know, the, the ex the, the, the relationship is,

03:08.265 --> 03:09.925
is quite kind of quite intuitive.

03:10.945 --> 03:13.525
If I think things are going to get better financially

03:13.745 --> 03:15.005
for me in a year's time,

03:15.405 --> 03:17.525
I might go on the spending spree Now,

03:17.955 --> 03:20.605
however, if I think things are gonna get worse

03:21.105 --> 03:24.245
for me financially, I may start cutting back now.

03:25.225 --> 03:27.885
So what we do in our paper is in contrast

03:28.065 --> 03:31.325
to the existing literature, which looks at the relationship

03:31.325 --> 03:32.525
between expectations

03:32.745 --> 03:36.325
and how household financial behavior IE consumption,

03:37.075 --> 03:39.845
what we do is we deviate from the standard approach,

03:40.135 --> 03:43.565
which is that financial expectations are treated as

03:43.565 --> 03:46.605
what we call an exogenous or a fixed variable.

03:47.425 --> 03:49.685
So we actually model expectations.

03:49.905 --> 03:51.085
We allow expectations

03:51.085 --> 03:54.685
to be endogenous within our framework rather than being

03:54.955 --> 03:58.605
exogenous fixed when we look at their effect on consumption.

03:59.385 --> 04:01.765
So there's two types of, of, of contribution.

04:02.025 --> 04:04.805
The first is, so statistical modeling contribution.

04:05.145 --> 04:07.445
How should we model financial expectations?

04:07.945 --> 04:11.765
And the second con contribution is if we model financial

04:11.765 --> 04:16.285
expectations appropriately allowing for this, concept,

04:16.435 --> 04:18.325
this phenomenon of mid middle inflation,

04:18.955 --> 04:21.965
what implications does that have for the relationship

04:21.965 --> 04:25.285
between expectations and household consumption?

04:26.305 --> 04:28.365
So let me say something about the data first,

04:28.505 --> 04:29.525
and then I'll, I'll try

04:29.525 --> 04:31.365
to explain the modeling framework

04:31.705 --> 04:33.085
is a little bit complicated.

04:33.625 --> 04:35.845
Um, but, but hopefully I'll be able to,

04:35.875 --> 04:37.085
explain the intuition.

04:37.985 --> 04:41.685
So we use data from the British household Panel survey,

04:41.695 --> 04:46.165
which is a longitudinal survey which covers the period 1999

04:46.385 --> 04:47.445
to 2008.

04:48.465 --> 04:52.565
So we conduct our analysis on what's called a balanced

04:52.825 --> 04:55.085
sample of individuals.

04:55.585 --> 05:00.285
And by balanced sample, I mean individuals are in the,

05:00.285 --> 05:03.525
in the sample for the full 18 year period.

05:04.225 --> 05:08.725
So we look at, 1,417 individuals

05:09.315 --> 05:11.165
over the 18 year period.

05:12.305 --> 05:15.125
And the, the fact that the ba the, the panel is balanced,

05:15.625 --> 05:19.005
we look at everybody over 18, years is very important

05:19.475 --> 05:21.645
because if we think about expectations

05:21.795 --> 05:23.565
data, it could be the case.

05:23.745 --> 05:25.205
The longer you are in the panel,

05:25.785 --> 05:28.245
the more you learn about your expectations,

05:28.385 --> 05:30.645
you get better at making predictions.

05:31.025 --> 05:32.125
So we want to make sure

05:32.125 --> 05:35.125
that everybody in our analysis is there for the,

05:35.145 --> 05:36.765
the same 18 years.

05:38.165 --> 05:39.985
So we take this longitudinal data

05:40.685 --> 05:43.785
and rather than model expectations data

05:44.455 --> 05:46.225
with just taking individuals

05:46.645 --> 05:50.745
and putting them in one of three groups, so optimistic,

05:50.995 --> 05:53.145
which means that I think things are gonna get better

05:53.925 --> 05:56.545
in the middle, which means that they're gonna stay the same,

05:56.915 --> 05:58.725
same or pessimistic, which means

05:58.725 --> 06:01.685
that it things are gonna get worse for me financially.

06:02.355 --> 06:04.085
What our modeling framework does,

06:04.385 --> 06:07.165
the generalized middle inflated ordered probate model,

06:07.905 --> 06:11.045
it introduces what we call splitting equations.

06:11.825 --> 06:15.645
And the splitting equations provide a potential mechanism

06:16.115 --> 06:19.645
whereby individuals can be steered away from saying

06:19.645 --> 06:20.725
that they're optimistic

06:20.785 --> 06:24.285
or pessimistic, towards this middle category.

06:24.435 --> 06:26.805
Whereby I think, expect my,

06:26.885 --> 06:29.285
I think my financial expectations are such

06:29.285 --> 06:32.365
that I think my income is gonna stay the same next year.

06:33.785 --> 06:35.685
So when we model expectations,

06:36.025 --> 06:38.605
we do things which are quite standard in the literature.

06:39.025 --> 06:40.925
We have a wide range of individual

06:41.025 --> 06:44.845
and household characteristics, age, gender, education,

06:45.145 --> 06:48.245
number of kids, marital status, home ownership,

06:48.415 --> 06:52.165
labor market status, household income, savings, and wealth.

06:53.265 --> 06:56.685
So when we do our standard model of expectations, we want

06:56.685 --> 06:59.605
to keep as close to the existing literature as possible.

06:59.905 --> 07:02.005
So there's nothing in there, I think,

07:02.215 --> 07:04.245
which is particularly different to

07:04.475 --> 07:06.685
what other researchers have used in the

07:06.685 --> 07:07.805
past, including myself.

07:09.385 --> 07:12.405
We then introduce these splitting equations,

07:13.025 --> 07:17.605
and in these equations, we, we use a wide range of,

07:17.875 --> 07:20.525
explanatory variables, things related

07:20.705 --> 07:24.045
to psychological factors, cognitive, factors,

07:24.385 --> 07:27.765
things which might contribute to generating an excess

07:27.785 --> 07:30.885
of observations within a particular category.

07:31.985 --> 07:34.725
So we still have variables such as age, gender,

07:34.785 --> 07:36.125
and education in there,

07:36.545 --> 07:40.005
but we also have information such as individual's,

07:40.005 --> 07:43.965
personality traits, the big five, agreeableness,

07:44.445 --> 07:47.485
extroversion, neuroticism, op, openness to experience

07:48.245 --> 07:49.485
and conscientiousness.

07:50.545 --> 07:53.965
We also, control for things such as the number

07:53.965 --> 07:58.485
of times within this 18 year period that individuals,

07:58.625 --> 08:01.005
are correct in their financial expectations.

08:01.765 --> 08:02.925
Remember, it's longitudinal data.

08:03.585 --> 08:06.285
So I can look at the first year of my panel,

08:07.005 --> 08:09.925
I can see whether the individuals think they're going

08:09.925 --> 08:11.205
to be better off financially,

08:11.795 --> 08:13.885
then I can use my longitudinal data

08:14.265 --> 08:15.965
to find out whether they're correct,

08:16.035 --> 08:17.525
they were correct in their prediction.

08:19.115 --> 08:22.015
The other things that we use in our spitting equation,

08:22.015 --> 08:25.815
which are relevant for today, is we use information related

08:25.915 --> 08:27.255
to interview conditions.

08:27.875 --> 08:31.295
So when the individual was making these expectations,

08:31.765 --> 08:34.095
were there problems related to language reading,

08:34.155 --> 08:36.895
interpretation, how long were the interviews?

08:37.205 --> 08:38.535
Because in a sense

08:38.535 --> 08:41.455
that there's a literature out there which states sat,

08:41.755 --> 08:43.055
the longer our interviews,

08:43.475 --> 08:46.535
the the more trust an individual has in the person

08:46.535 --> 08:47.655
that's interviewing them,

08:47.995 --> 08:51.415
and the more likely they are to give accurate responses.

08:52.355 --> 08:55.605
So the key part, part about our model is

08:55.605 --> 08:58.445
that we allow individuals to be steered away

08:59.115 --> 09:01.285
from their initial expectations.

09:01.905 --> 09:04.125
And the word allow is very important.

09:04.585 --> 09:07.645
We don't enforce it if the data doesn't support it.

09:07.945 --> 09:11.205
We just allow this mechanism whereby individuals may be

09:11.205 --> 09:14.005
steered away from a particular category.

09:15.225 --> 09:19.205
And once we have modeled our expectations appropriately

09:20.285 --> 09:21.765
allowing for this middle inflation,

09:22.345 --> 09:24.805
we can then look at the implications of this

09:25.145 --> 09:28.565
for the relationship between expectations and consumption.

09:29.945 --> 09:32.485
So what we find is that when we,

09:33.185 --> 09:35.445
set up our generalized ordered probate model

09:36.225 --> 09:39.085
and we do our analysis, we find

09:39.235 --> 09:42.365
that 27% of responses

09:42.515 --> 09:46.325
that we observe in our expectations data can be attributed

09:46.585 --> 09:48.565
to the impact of middle inflation.

09:49.425 --> 09:54.085
So just to repeat, 27% of responses can be attributed

09:54.345 --> 09:55.805
to the impact of middle inflation.

09:56.185 --> 09:58.445
So that's a high percentage in terms

09:58.465 --> 09:59.845
of the data that we are looking at.

10:00.585 --> 10:02.925
We find that the probability associated

10:02.955 --> 10:06.765
with being steered away from the pessimistic category,

10:06.785 --> 10:10.565
the worse off category, is greater than the probability

10:10.825 --> 10:12.405
of our splitting equations,

10:13.205 --> 10:15.765
steering people away from the better off the

10:15.765 --> 10:16.925
optimistic category.

10:18.025 --> 10:20.325
So given that middle inflation seems

10:20.345 --> 10:23.125
to be important in shaping individuals' financial

10:23.125 --> 10:26.685
expectations, the second stage of our analysis is

10:26.685 --> 10:29.925
to think about the implications of this when we're trying

10:29.925 --> 10:31.645
to look at the, patterns

10:31.665 --> 10:35.085
or model, the patterns of UK household consumption behavior.

10:36.425 --> 10:38.045
So we know that

10:38.595 --> 10:41.245
however we model financial expectations,

10:41.415 --> 10:43.285
there is a statistical relationship

10:43.315 --> 10:45.285
between financial expectations

10:45.385 --> 10:48.405
and consumption regardless of the modeling approach.

10:49.915 --> 10:53.895
But our results reveal that specifications models,

10:54.205 --> 10:57.135
statistical models, which do not explicitly control

10:57.135 --> 11:00.975
for middle inflation, tend to overestimate the effect

11:00.995 --> 11:05.655
of sentiment consumer expectations on expenditure.

11:06.235 --> 11:09.575
And this could be expenditure as measured by the probability

11:09.875 --> 11:12.335
of, for example, buying durable goods

11:12.555 --> 11:16.615
and the amount of the, the amount of, expenditure

11:16.615 --> 11:20.135
that households incur on, for example, a durable good.

11:20.755 --> 11:23.575
So our findings suggests that if you ignore the role

11:23.835 --> 11:27.935
of middle inflation, we get a situation whereby the impact

11:28.115 --> 11:31.735
of financial expectations on household consumption

11:32.355 --> 11:33.615
are overstated.

11:34.815 --> 11:37.395
So if we just take one example, there's, there's,

11:37.395 --> 11:39.395
there's lots of different types of expenditure,

11:39.395 --> 11:40.795
which are looked at in the paper,

11:41.375 --> 11:42.995
but if we just look at the amount

11:42.995 --> 11:44.955
of expenditure on durable goods,

11:45.415 --> 11:49.955
the overestimate from a standard modeling approach compared

11:49.955 --> 11:54.005
to our new statistical framework is approx is approximately

11:54.005 --> 11:55.125
38%.

11:55.865 --> 11:57.485
So it is a large effect.

11:57.945 --> 12:00.445
Um, when I talk about the fact that the relationship

12:00.445 --> 12:02.405
between, expectations

12:02.465 --> 12:05.885
and consumption is, is overstated, it's overstated

12:05.985 --> 12:07.245
by about 38%.

12:08.685 --> 12:09.865
So why is this important?

12:10.095 --> 12:12.305
Well, if we think about policymakers,

12:12.525 --> 12:15.385
we can think about the, the recent budget,

12:15.535 --> 12:18.625
when we are thinking about the effect of, tax changes,

12:19.205 --> 12:20.425
tax cuts and so on.

12:20.765 --> 12:24.025
In order to model the effects of such policy changes,

12:24.495 --> 12:27.985
what we're interested in is thinking about how individuals

12:27.985 --> 12:29.305
and households respond.

12:30.045 --> 12:32.505
So for example, if the budget takes place,

12:32.685 --> 12:35.825
and I think I'm gonna be better off in a year's time, then

12:36.005 --> 12:38.705
as I said before, I may go on a spending spree.

12:39.205 --> 12:41.745
If on the other hand, I think that's not gonna happen,

12:42.265 --> 12:44.465
I may cut down in terms of my expenditure.

12:45.045 --> 12:48.785
So we're trying to work out the relationship between,

12:49.265 --> 12:52.945
expectations and things like household consumption.

12:53.415 --> 12:55.625
What we're trying to do is to work out the,

12:55.805 --> 12:58.225
the true relationship between these two things.

12:58.965 --> 13:00.385
And our paper shows.

13:00.605 --> 13:02.985
Our research shows that you need

13:03.045 --> 13:07.745
to appropriately model financial expectations if you want

13:07.885 --> 13:11.025
to get an accurate prediction of the extent

13:11.025 --> 13:14.425
to which expectations may affect things like consumption,

13:15.085 --> 13:17.705
and things like changes in monetary policy,

13:17.845 --> 13:21.145
fiscal policy, all of these sorts of things,

13:21.365 --> 13:25.185
are introduced in order to affect, for example, households,

13:25.505 --> 13:28.465
behaviors, their financial, decisions and so on.

13:29.125 --> 13:31.625
And if we think about the current living standards crisis,

13:32.105 --> 13:33.865
I think, you know, going forward in terms

13:33.865 --> 13:37.185
of future research, one thing that's very interesting to me

13:37.185 --> 13:39.945
as an economist is over the last, year

13:39.945 --> 13:41.265
or so, there's been a lot

13:41.265 --> 13:43.985
of discussion about inflation in the news.

13:44.605 --> 13:46.425
So in a sense, households,

13:46.425 --> 13:49.785
because terms like inflation, are, are

13:50.645 --> 13:51.945
now common terms, which,

13:51.945 --> 13:53.585
which households are faced in the news

13:53.845 --> 13:57.185
and media on a regular basis going, going forward.

13:57.425 --> 13:58.945
I think it will be interesting to see whether

13:58.965 --> 14:02.985
or not households, financial expectations on more

14:03.165 --> 14:05.305
or less accurate going forward.

14:05.845 --> 14:07.625
Um, but that's just something to kind

14:07.625 --> 14:10.425
of highlight possible areas for future research.

Understanding Middle Responses in Survey Data

00:00:00 - So just a little bit of context. So survey data is widely used in economics and right across the social sciences more generally. So we think about surveys when households and individuals could be asked their opinions and attitudes about all sorts of things, political preferences, immigration, as well as their finances. So you name it. anything could be in the survey. So when individuals are faced with survey questions, there is a tendency for 'em to choose a middle category,
00:00:32 - So what we tend to observe is a concentration of responses by individuals in the middle category. So if you present somebody with a statement, they very often say they neither agree or disagree with a particular statement. So in our paper, as you say, we met, we focus on financial expectations and specifically, individuals are asked a, a, a question looking ahead, how do you think you'll,
00:01:04 - So it's a very general question, and when we look at the responses to this, people can say that they're, they think they're gonna be worse off about the same or better off. And what we find is about 10% of individuals think they're gonna be worse off in a year's time. About 61% say they think they're gonna be about the same, and the remainder about, 28% think they're gonna be better off. So this clump in the middle whereby people say,
00:01:35 - That's what was referred to as middle inflation, whereby we get this grouping, this concentration in the middle category. And that's, something that happens with lots of different survey questions, not just ones about financial expectations, but quite generally, people sort of just go in the middle. So there's a very small literature that looks at this characteristic of survey data. There are loads and loads of papers that use survey data in economics and social sciences,
00:02:06 - modeling implications of it. So what we actually do in our paper is we introduce a new statistical modeling framework whereby we can quantify how much of a survey response is likely to be related to overinflation rather than, for example, a true belief that I really think my income will be unchanged. So the other literature which we relate to is,
00:02:38 - So the contribution is kind of twofold. The first thing is to actually think about how should we model this type of survey data where we get this kind of clump in the middle response. The second thing that we do is that we look at the implications of how we model survey data for a particular outcome Such as household consumption. So if we think about expectations in household consumption,

Modeling Financial Expectations and Consumer Behavior

00:03:08 - is quite kind of quite intuitive. If I think things are going to get better financially for me in a year's time, I might go on the spending spree Now, however, if I think things are gonna get worse for me financially, I may start cutting back now. So what we do in our paper is in contrast to the existing literature, which looks at the relationship between expectations and how household financial behavior IE consumption,
00:03:40 - which is that financial expectations are treated as what we call an exogenous or a fixed variable. So we actually model expectations. We allow expectations to be endogenous within our framework rather than being exogenous fixed when we look at their effect on consumption. So there's two types of, of, of contribution. The first is, so statistical modeling contribution. How should we model financial expectations?
00:04:11 - expectations appropriately allowing for this, concept, this phenomenon of mid middle inflation, what implications does that have for the relationship between expectations and household consumption? So let me say something about the data first, and then I'll, I'll try to explain the modeling framework is a little bit complicated. but, but hopefully I'll be able to, explain the intuition.
00:04:41 - which is a longitudinal survey which covers the period 1999 to 2008. So we conduct our analysis on what's called a balanced sample of individuals. And by balanced sample, I mean individuals are in the, in the sample for the full 18 year period. So we look at, 1,417 individuals
00:05:12 - And the, the fact that the ba the, the panel is balanced, we look at everybody over 18, years is very important because if we think about expectations data, it could be the case. The longer you are in the panel, the more you learn about your expectations, you get better at making predictions. So we want to make sure that everybody in our analysis is there for the, the same 18 years. So we take this longitudinal data
00:05:44 - with just taking individuals and putting them in one of three groups, so optimistic, which means that I think things are gonna get better in the middle, which means that they're gonna stay the same, same or pessimistic, which means that it things are gonna get worse for me financially. What our modeling framework does, the generalized middle inflated ordered probate model, it introduces what we call splitting equations.

Modeling Financial Expectations and Middle Inflation

00:06:16 - whereby individuals can be steered away from saying that they're optimistic or pessimistic, towards this middle category. Whereby I think, expect my, I think my financial expectations are such that I think my income is gonna stay the same next year. So when we model expectations, we do things which are quite standard in the literature. We have a wide range of individual and household characteristics, age, gender, education,
00:06:48 - labor market status, household income, savings, and wealth. So when we do our standard model of expectations, we want to keep as close to the existing literature as possible. So there's nothing in there, I think, which is particularly different to what other researchers have used in the past, including myself. We then introduce these splitting equations, and in these equations, we, we use a wide range of,
00:07:20 - to psychological factors, cognitive, factors, things which might contribute to generating an excess of observations within a particular category. So we still have variables such as age, gender, and education in there, but we also have information such as individual's, personality traits, the big five, agreeableness, extroversion, neuroticism, op, openness to experience
00:07:50 - We also, control for things such as the number of times within this 18 year period that individuals, are correct in their financial expectations. Remember, it's longitudinal data. So I can look at the first year of my panel, I can see whether the individuals think they're going to be better off financially, then I can use my longitudinal data to find out whether they're correct, they were correct in their prediction.
00:08:22 - which are relevant for today, is we use information related to interview conditions. So when the individual was making these expectations, were there problems related to language reading, interpretation, how long were the interviews? Because in a sense that there's a literature out there which states sat, the longer our interviews, the the more trust an individual has in the person that's interviewing them, and the more likely they are to give accurate responses.
00:08:55 - that we allow individuals to be steered away from their initial expectations. And the word allow is very important. We don't enforce it if the data doesn't support it. We just allow this mechanism whereby individuals may be steered away from a particular category. And once we have modeled our expectations appropriately allowing for this middle inflation, we can then look at the implications of this

Impact of Middle Inflation on Financial Expectations

00:09:29 - So what we find is that when we, set up our generalized ordered probate model and we do our analysis, we find that 27% of responses that we observe in our expectations data can be attributed to the impact of middle inflation. So just to repeat, 27% of responses can be attributed to the impact of middle inflation.
00:09:58 - of the data that we are looking at. We find that the probability associated with being steered away from the pessimistic category, the worse off category, is greater than the probability of our splitting equations, steering people away from the better off the optimistic category. So given that middle inflation seems to be important in shaping individuals' financial expectations, the second stage of our analysis is
00:10:29 - to look at the, patterns or model, the patterns of UK household consumption behavior. So we know that however we model financial expectations, there is a statistical relationship between financial expectations and consumption regardless of the modeling approach. But our results reveal that specifications models, statistical models, which do not explicitly control
00:11:00 - of sentiment consumer expectations on expenditure. And this could be expenditure as measured by the probability of, for example, buying durable goods and the amount of the, the amount of, expenditure that households incur on, for example, a durable good. So our findings suggests that if you ignore the role of middle inflation, we get a situation whereby the impact
00:11:32 - are overstated. So if we just take one example, there's, there's, there's lots of different types of expenditure, which are looked at in the paper, but if we just look at the amount of expenditure on durable goods, the overestimate from a standard modeling approach compared to our new statistical framework is approx is approximately 38%. So it is a large effect. when I talk about the fact that the relationship
00:12:02 - and consumption is, is overstated, it's overstated by about 38%. So why is this important? Well, if we think about policymakers, we can think about the, the recent budget, when we are thinking about the effect of, tax changes, tax cuts and so on. In order to model the effects of such policy changes, what we're interested in is thinking about how individuals and households respond. So for example, if the budget takes place,

Impact of Expectations on Household Consumption

00:12:36 - as I said before, I may go on a spending spree. If on the other hand, I think that's not gonna happen, I may cut down in terms of my expenditure. So we're trying to work out the relationship between, expectations and things like household consumption. What we're trying to do is to work out the, the true relationship between these two things. And our paper shows. Our research shows that you need
00:13:07 - to get an accurate prediction of the extent to which expectations may affect things like consumption, and things like changes in monetary policy, fiscal policy, all of these sorts of things, are introduced in order to affect, for example, households, behaviors, their financial, decisions and so on. And if we think about the current living standards crisis, I think, you know, going forward in terms
00:13:37 - as an economist is over the last, year or so, there's been a lot of discussion about inflation in the news. So in a sense, households, because terms like inflation, are, are now common terms, which, which households are faced in the news and media on a regular basis going, going forward. I think it will be interesting to see whether or not households, financial expectations on more or less accurate going forward.
00:14:07 - of highlight possible areas for future research.

Faculti AI can make mistakes

Research

Policy

Analysis