Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
2014 EU-SILC International Conference Conference, Lisbon, 16-17 October 2014 - Session 5 -
This paper was prepared as part of "Net-SILC2", an international research project funded by Eurostat
and coordinated by CEPS/INSTEAD (Luxembourg). It is a draft version subject to revision. It should
not be quoted or disseminated without the written consent of the author(s).
How do European citizens cope with economic
shocks? The longitudinal order of deprivation
Anne-Catherine Guio
(CEPS/INSTEAD, Luxembourg)
Marco Pomati
(University of Cardiff, UK)
1
How do European citizens cope
with economic shocks?
The longitudinal order of deprivation1
Anne-Catherine Guio2 and Marco Pomati3
Abstract
The recent economic crisis had a dramatic impact on European citizens, leading to more people
experiencing poverty and material deprivation. At EU level, the number of people suffering from
severe deprivation in 2012 reached more than 49.000.000, i.e. an increase of 8.7 millions people
since 2009.
The main contribution of this paper is to understand which items people have to go without as
their resources decrease, using the longitudinal component of EU-SILC. By definition,
curtailment is a temporal process which to be fully understood necessitates longitudinal data.
Although only a subset of deprivation items is available in the longitudinal dataset, this allows us
to compare the order of curtailment obtained by using longitudinal and cross-sectional data. An
IRT model is also estimated on cross-sectional data and used to confirm and aid the interpretation
of the results.
Interestingly, the results suggest a large degree of homogeneity across the EU in how households
curtail expenditure, despite the large differences in material and social contexts between Member
States.
1 This research was financed by the European Commission, Directorate General for Employment, Social Affairs and
Inclusion and by the second Network for the analysis of EU-SILC (Net-SILC2), funded by Eurostat. The European
Commission bears no responsibility for the analyses and conclusions, which are solely those of the authors. We would
to thank Tony Atkinson, Céline Thévenot, Isabelle Maquet and Serge Paugam for fruitful discussions and interesting
comments. 2 CEPS/INSTEAD, Esch-sur-Alzette, Luxembourg ([email protected]) 3 School of Social Sciences, University of Cardiff , UK ([email protected])
2
1. INTRODUCTION
The literature on budgeting strategies adopted by households on low income reveals a mixture of
resourcefulness and experience of mild to severe material deprivation. Households on low income
often rely on formal (such as state benefits) but also informal sources of financial/social assistance
and reciprocity exchange (Dean and Shah, 2002) to maintain at least some elements of their usual
lifestyles, ranging from borrowing money and exchanging favours with friends and relatives to
moving back with one’s parents and in-laws (Orr et al, 2006). However, despite the resilience of
many of these households in conditions of low income, there is considerable evidence that as
formal resources such as income drop, material deprivation is experienced (Yeung and Hofferth,
1998; Saunders et al., 2006; Berthoud and Bryan, 2011). Hence, although research on resilience
documents the resoluteness of those with low resources to maintain their usual style of living,
many have found that making ends meet means going without (Kempson, 1996; Anderson et al.,
2010; Batty and Cole, 2010; Pemberton et al., 2013). Orr et al (2006) argue that reductions in
resources caused by redundancies or illness (also known as income shocks or critical life events)
are easily absorbed only among high income households. At medium income levels households
begin to cut back on items such as holidays and rely on help from families and friends; through
minimal changes in living standards, physical assets and customary activities are maintained.
However, as resources drop even further, social capital is stretched to the limit, items previously
taken for granted become unaffordable and eventually even food consumption is reduced to a
minimum and a warm house becomes an unaffordable luxury (Wright, 2004). Hence, qualitative
evidence suggests that despite great heterogeneity in prices and consumer preferences and tastes,
the range of deprivations that households in poverty experience is relatively limited, and therefore
find similar deprivation patterns across households with similar levels of resources (Smith, 2005;
Fahmy and Pemberton, 2008). Similarly, large-scale expenditure studies also suggest that as
income rises among those who suffered from deprivation, commodity expenditure patterns
converge with those of higher-income households (Farrell and O’Connor, 2003; Gregg et al.,
2005).
Questions on material deprivations available in surveys such as the EU Statistics on Income and
Living Conditions (EU-SILC) data-set provide information on the types of goods and activities
that many households go without because of lack of resources. They enable policy-makers to
estimate how many people in a given country cannot afford a one-week annual holiday away from
home or keep their home adequately warm. Since 2009, the European Union portfolio of
commonly agreed social indicators includes measures of material deprivation (Guio, 2009),
defined as the enforced lack of (or the inability to afford, when desired) items and activities such
as holidays once a year, keeping one’s home adequately warm, facing unexpected expenses,
avoiding arrears, a washing machine, TV, telephone or a car. These indicators refer to “enforced
lacks”, i.e. lack of an item/activity due to insufficient resources and not lack due to choice (Mack
and Lansley, 1985) and they provide a snapshot of the many coping strategies of European
households. However, despite the large availability of deprivation data, little attention has been
given to the order in which certain spending curtailment strategies are adopted across the
European Union.
3
We argue that understanding how households cope with income shocks is important to assess
claims that poverty is the result of erratic spending or inefficient household budgeting: if this was
the case one would for example find a substantial amount of individuals who can afford to go on
holiday away from home but cannot afford to keep their houses warm or two pairs of all-weather
shoes. Understanding of the order in which deprivations are experienced also helps to establish a
common language across European welfare states to describe the severity of deprivation. If a
common order of curtailment is established across countries then deprivation indicators can also
be used to understand the severity of deprivation, despite the large international differences in
material and social resources and contexts. Overall, focusing on patterns of curtailment in
different countries also enables deprivation research to move towards a better understanding of
some of the key aspects of living conditions and the underlying processes of curtailment shared
across the EU. This is in line with the current definition of poverty, which following Townsend
(1979) defines the poor as those excluded from the minimum acceptable way of life in the Member
State to which they belong (Council of European Union, 1985).
The main contribution of this paper is to assess the most frequent deprivation sequence at the EU
and national levels, using the longitudinal component of EU-SILC. By definition, curtailment is
a process that happens over time and whose study ideally necessitates following up the same
individuals as they become more or less deprived across time. Even if deprivation sequences may
be assessed on the basis on the cross-sectional data, by comparing the deprivation patterns of
people with different deprivation levels at one point in time (see Deutsch and Silber, 2008 and
Deutsch et al., 2014), longitudinal data allows us to assess the fit of different deprivation
sequences by also using information on how deprivation evolves for each individual case across
time. Although only a subset of deprivation items is available in the longitudinal dataset4, this
allows us to compare the order of curtailment obtained by using longitudinal and cross-sectional
data. An Item Response Theory model is also estimated on cross-sectional data and used to
confirm and aid the interpretation of both cross-sectional and longitudinal Deprivation Sequence
results.
2. DATA
In order to estimate the order in which different items and activities are curtailed in different
countries, both cross-sectional (2009) and longitudinal (2009-2011) components of the EU-SILC
are used.
The cross-sectional analysis is conducted on a set of 13 deprivation items available for the first
time in the 2009 EU-SILC cross-sectional data, collected through a thematic module on material
deprivation. These 13 items were proposed by Guio, Gordon and Marlier (2012) as suitable, valid,
reliable and additive measure of deprivation at the EU level and in each individual Member State.
4 We opted to use three waves of the longitudinal data set, instead of the four waves available in order to increase the
sample size. Indeed EU-SILC is a rotational panel, i.e. each wave a quarter of the sample quit the panel. Following people during three years instead of four years allows working with 50% of the sample, instead of 25%.
4
The longitudinal analysis is estimated on six items because only six items (out of the 13-item list)
are available in the EU-SILC longitudinal data (2009-2011)5.
The deprivation items conform to the definition of enforced lack (Mack and Lansley, 1985)
outlined above and are listed below in Table 1. The last column indicates which deprivation items
are used in the longitudinal analysis (see also Annex 1 for a description of material deprivation
indicators used in this analysis and at the EU level).
Table 1 Deprivation rates (based on 2009 EU-SILC data) for items used in cross-sectional
and longitudinal analysis
Deprivation (Enforced lack) % Longitudinal
“Household items”, i.e. items collected at household level. The household deprivation
information is assigned to all household members (including children) when the household
cannot afford the item.
Afford one week annual holiday away from home 38
Face unexpected expenses 35
Replace worn-out furniture 31
Avoid arrears (mortgage or rent, utility bills or hire purchase
instalments)
12
Afford a meal with meat, chicken, fish or vegetarian equivalent every
second day
10
Keep home adequately warm 10
Afford/ have access to a car/van for private use (but would like to have) 9
Afford a computer and an internet connection (but would like to have) 5
“Adult items”, i.e. items collected at individual adult level (people aged 16+, living in
private households). The adult deprivation information is assigned to all household
members (including children), if at least half the adults in the household cannot afford the
item.
Have regular leisure activities 18
Spend a small amount of money each week on oneself without having
to consult anyone (pocket money)
17
Get together with friends/family for a drink/meal at least monthly 13
Replace worn-out clothes by some new (not second-hand) ones 12
Afford two pairs of properly fitting shoes, including a pair of all-weather
shoes
3
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, authors’computation.
Reading notes: In 2009, 3% of EU-27 citizens suffered from the enforced lack of two pairs of all-weather
shoes.
Using the six items available in the longitudinal components of EU-SILC, Table 2 divides the
weighted longitudinal sample according to individual deprivation trajectories across two
consecutive years (2010 and 2011). The table shows that among those who were deprived in 2010
and 2011 (3rd to 6th columns from the left), the majority experienced an increase or decrease in
the number of deprivations, although a non-negligible minority experienced exactly the same
5 We opted to use three waves of the longitudinal data set, instead of the four waves available in order to increase the
sample size. Indeed EU-SILC is a rotational panel, i.e. each wave a quarter of the sample quit the panel. Following people during three years instead of four years allows working with 50% of the sample, instead of 25%.
5
number and types of deprivations (around 8% of the sample for Austria) or experienced the same
number but different types of deprivations (2% in Belgium). Overall, Table 2 suggests for the vast
majority of countries there is a substantial amount of change in deprivation profiles across just
two years. Moreover, when looking at individual trajectories only a small minority of cases
experience the same number of deprivations and switch between types of deprivations
(i.e. items). However, to fully tackle the role of consumer choice and relative prices much
more detailed data on the quality and price of good owned and not owned by respondents,
together with international market prices is needed. We return to these issues in the
conclusion.
This paper focuses on the individual deprivation patterns of people from twenty Member
States who have shared information (for up to three consecutive years) on the deprivations
they endure Hence, in the next sections, we will begin to open the black box of the
deprivation transitions and see whether we can identify a shared pattern of curtailment across
countries and methodologies. For doing so, we will focus on the people who lacked at least one
item in one of the last three years of the panel (2011, 2010 and 2009).
6
Table 2 Deprivation transitions, 6 items, 2010 and 2011
% of people
Not deprived
of any item in
2010 & 2011
Deprived of the
same item(s) in
2010 & 2011
Deprived of more/fewer/different items in
2010 & 2011
More
items
Fewer
items
Equal number but
different items
AT 61 8 10 18 2
BE 63 10 13 12 2
BG 8 22 27 35 9
CY 26 16 30 23 5
CZ 39 21 18 18 4
DK 75 5 10 8 2
EE 27 21 23 23 5
ES 42 13 17 23 4
FI 63 9 13 12 2
HU 12 35 27 22 4
IT 37 7 31 20 4
LT 12 23 26 31 9
LU 71 7 11 11 1
LV 8 23 33 29 8
MT 31 34 11 23 1
NL 74 8 8 8 2
PL 25 36 17 19 3
PT 25 21 25 25 5
RO 12 45 20 21 3
UK 54 10 16 17 4 Source: EU-SILC 2011 longitudinal data, Users’ database - August 2013.
Reading notes: Among the people present in the panel in 2010 and 2011 in Austria, 61% did not lack any
of the six items in 2011 and in 2010, 8% lack exactly the same items in both years, 10% of people lacked
more items in 2011 than 2010, 18% lacked less items in 2011 than in 2010 and 2% lack the same number
of items, but which were different.
3 DEPRIVATION SEQUENCE
3.1 Descriptive analysis
Table 1, based on the cross-sectional incidence of the 13 deprivation items available in EU-SILC
2009, suggests that the enforced lack of two pairs of all-weather shoes is only experienced by a
small minority (3%), particularly when compared to more common deprivations such as the
enforced lack of one week holiday away from home. Our data also suggest that around 90% of
those who can’t afford two pairs of all-weather shoes also cannot afford a holiday, while fewer
than 10% who can’t afford the latter can’t afford shoes (EU-SILC 2009). This would suggest that
as resources (such as income) begin to decrease people tend to curtail their holidays first and it is
7
only when their resources are extremely low that they lose the ability to afford even very basic
goods like shoes. One way to corroborate this claim visually is to divide respondents according
to how many items they can’t afford (number of deprivations) as shown in
Figure 1, here ranging from 1 to 13. This shows that holidays and unexpected expenses
deprivations are much more widespread than arrears and shoes deprivations across the deprivation
scale. More than half of those who can’t afford two necessities can’t afford holidays or unexpected
expenses, and this proportion grows gradually with the number of deprivations. In contrast, only
a small proportion cannot afford to pay arrears or two pairs of shoes. However, this small
proportion grows gradually with the number of deprivations. Most importantly, the order
(holidays, unexpected expenses, arrears and shoes) is constant across the deprivation scale.
Figure 1: Percentage who can’t afford each item, by level of deprivation, EU level, 2009
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, authors’computation.
Reading notes: More than half of those who can’t afford two necessities can’t afford holidays or unexpected
expenses,
A very similar pattern emerges by dividing respondents into income quintiles (see Figure 2). In
this case, the most likely order of curtailment at the EU level is clearly holidays, unexpected
expenses, arrears and finally shoes.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1 2 3 4 5 6 7 8 9 10 11 12 13Deprivation score
Holidays Unexpected expenses Arrears Shoes
8
Figure 2: Percentage who can’t afford each item, by income quintile, EU level, 2009
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, authors’computation.
Reading notes: More than 60% of those who are in the first (lowest) income quintile can’t afford holidays
or unexpected expenses,
This order is nevertheless probabilistic: although on average respondents will conform to this
pattern, it does not necessarily apply perfectly to all respondents. Similarly to a model prediction,
there is always some degree of difference between observed and predicted orders: even when
considering the four items above there is a small minority of people who can’t afford to pay arrears
nor afford two pairs of shoes but who are able to afford holidays. This could be the result of
misreporting and/or unique individual factors and particular resources which set this rare group
of cases apart from the vast majority of the population. As the number of deprivation items
increases the relative frequency order will become more uncertain and the number of cases that
do not confirm exactly to the best order of curtailment will also increase. As shown in Figure 3,
the order for holidays, unexpected expenses and shoes remains constant across the deprivation
scale, while the order is less clear for other items (such as car and arrears) across the deprivation
scale.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
5th 4th 3rd 2nd 1st
Income quintile
Holidays Unexpcted Expenses Arrears Shoes
9
Figure 3: Proportion of people who can’t afford the item, by level of deprivation, EU level,
2009
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, authors’computation
Finding the most representative order of curtailment for 20 countries visually would be unfeasible,
and more advanced methods are therefore needed.
3.2 The Deprivation sequence methodology
Understanding the determinants of the individual consumption level and the relative shares of its
components is a long standing issue in economics (Engel (1895), Working (1943), Leser (1963)).
According to classical microeconomic theory of consumption behaviour, consumers are supposed
to allocate their income to the purchase of products so as to maximize utility, given a set of prices
for a group of products. Econometric studies usually use detailed individual data from household
budget surveys to estimate systems of demand equations, where the share of expenditures depend
on the relative prices of different goods, disposable income and individual characteristics (see
among others the model proposed by Deaton and Muellbauer (1980)). Some research focused
specifically on how consumers prioritise their acquisition of durables over time, as their income
increases and to whether people tend to have similar priority of acquisition patterns for sets of
consumer durables (Roos and Von Szelisk, 1943; Paroush, 1965, 1973; McFall, 1969; Hebden
and Pickering, 1974). Drawing on the work by Paroush (1965, 1973) and Guttman (1950),
Deutsch and Silber (2008) used for the first time the same approach to look at the mirror image:
whether individuals facing the threat of poverty are curtailing their consumption of various goods
in a given order.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1 2 3 4 5 6 7 8 9 10 11 12 13Deprivation score
Holidays Unexp.Exp Furniture Leisure Pocket money
Drink/meal out New clothes Meat Home warm Car
Arrears Computer/Int Shoes
10
This methodology compares the deprivation order of each case in a dataset to all the possible
orders. For example, if the questionnaire contains two questions on whether respondents can
afford a one-week holiday a year and two pairs of shoes, there are two possible orders of
curtailment: as resources decrease households could decide to curtail holidays first and then two
pairs of shoes. Alternatively they could curtail two pairs of all-weather shoes, but still go on
holiday. Assuming that data was collected on only these two items and that being able to afford
or not wanting an item is scored as 0 and being deprived (unable to afford) is scored as 1, it is
possible to test which order best approximates the one found among all cases in our sample. For
example, if holidays were curtailed first followed by shoes we would expect three possible
patterns in the data consistent with this order:
Table 3 Expected patterns for order 1 (Holidays, Shoes)
Holidays Shoes
0 0
1 0
1 1
We could then compare each case in our dataset to this pattern, and allocate errors to each case as
we did above. We would then aggregate the total amount of error for each possible order. There
are K! possible orders, where K is the number of deprivation variables. In the simple example
above there are only 2!=2 possible patterns.
Respondents would either be able to afford both holidays and shoes (as in the first row), or be
unable to afford holidays and able to afford shoes (second row), or be unable to afford either (third
row). Respondents who can’t afford shoes but can afford holidays (pattern 0,1) are in this case
not consistent with the considered order and would need one change (an error) to be converted to
the closest expected pattern (from 0,1 to 1,1). An error of 1 (or a residual in modelling terms)
would then be allocated to this case. If the expected order was the opposite of the one above (shoes
are curtailed first), we would expect the opposite patterns in the data, as shown in Table 4.
Table 4 Expected pattern for order 2 (Shoes, Holidays)
Shoes Holidays
0 0
1 0
1 1
In the presence of longitudinal data we can extend the Deprivation Sequence Methodology
explained above by looking at information over multiple episodes (waves) for the same person
(see 5). Each individual deprivation pattern found in the data is scored against a number of
expected patterns given an order (e.g. Holiday, Shoes). The main difference is that the expected
patterns also allow increase, decrease and no change in deprivation scores across time. Each case
is therefore compared against the expected pattern and allocated an error. As explored above, an
error is the smallest change between the deprivation pattern of a given dataset case and an
11
expected pattern. All cases that match any of the expected patterns of a given order are allocated
an error of 0.
Table 5 shows the longitudinal extension of order 1 shown in Table 3. Similarly to the cross-
sectional methodology, an aggregate error is calculated at the national/EU level, and the order
with the smallest aggregate error is selected as the “best” national order.
Table 5 Expected longitudinal patterns for order n. 1 (Holidays, Shoes)
WAVE 1 WAVE 2
Holidays Shoes Holidays Shoes
0 0 0 0
0 0 1 0
1 0 0 0
0 0 1 1
1 0 1 0
1 1 0 0
1 0 1 1
1 1 1 0
1 1 1 1
3.3 The Deprivation sequence: results
A. Best EU order
The best order is simply the order with the lowest aggregate error. Using the 6 items available in
the longitudinal dataset (see Table 1), this order is:
1) Holidays
2) Unexpected expenses
3) Meat/chicken/fish
4) Home warm
5) Arrears
6) Car
As their resources increase, households on average tend to first cut back on their annual holidays
and use up their savings (resulting in inability to face unexpected expenses), and as their resources
decrease even further they are even unable to afford meals, a warm house and paying bills, and
eventually even a car.
The results from the longitudinal analyses show a substantial amount of overlap with those based
on cross-sectional data, as shown in Table 5 (see Deutsch et al., 2014, for a discussion of the
cross-sectional results). At the national level, the hierarchies show either a perfect or very close
match. This suggests that the cross-sectional 13-item order can be considered a good predictor of
the longitudinal deprivation sequence. This order was, at the EU level:
12
1) Holidays*
2) Unexpected expenses*
3) Furniture
4) Pocket Money
5) Leisure
6) Drink/meal out
7) Clothes
8) Meat/chicken/fish*
9) Home warm*
10) Arrears*
11) Car*
12) Computer/Internet
13) Shoes
National results of the cross-sectional 13-item order are provided in Annex 3.
*Items only available in the longitudinal data set, see Table 1.
B. Homogeneity of national deprivation orders across the EU
In Table 6, there is a large degree of overlap between national hierarchies: holidays and expenses
are generally the first items to be curtailed across all countries. As for the other items, most
countries have an order similar to the EU one, but the variation is much more noticeable. Bulgaria
and Portugal for example are the only countries where the enforced lack of an adequately warm
house is first and second respectively. Similarly, access to a car is the second (cross-
sectional)/third (longitudinal) item in Romania.
Focusing on the differences between national best orders however hides the fact that the EU order
fits most countries relatively well. A more sensible strategy is to store the aggregate errors for
each of the 720 (6!) possible hierarchies and then rank them. As shown in Table 7, out of 720
possible longitudinal hierarchies the EU order has a rank of less than 55 in all countries apart from
Denmark and Finland. This means that the EU order may not be the best fitting one but it fits
better than 92% (i.e. 55/720) of all the other possible orders in most countries. The orders that fit
marginally better are substantially small variations of the EU order. For Denmark and Finland the
EU order is still better than the vast majority of orders but the rank is much lower (134th and 154th
respectively). The fourth column also shows that any order with holidays and unexpected
expenses at end of the order fits all countries badly.
13
Table 6: Best order of curtailment, longitudinal and cross-sectional data, EU countries
EU-27 AT BE BG CY CZ DK EE ES FI HU IT LT LU LV MT NL PL PT RO UK
Holidays
CS 1 2 1 2 1 1 2 1 1 2 2 1 2 2 2 1 2 1 1 1 2
LONGI 1 2 1 2 1 1 2 1 1 2 2 1 2 2 2 1 2 1 1 1 2
Unexp. expenses
CS 2 1 2 3 2 2 1 2 2 1 1 2 1 1 1 2 1 2 3 3 1
LONGI 2 1 2 3 2 2 1 2 2 1 1 2 1 1 1 2 1 2 3 2 1
Meat/
chicken/
fish
CS 3 3 5 4 5 3 4 4 6 4 3 4 3 4 3 3 6 3 5 4 4
LONGI 3 3 5 4 5 4 4 4 5 5 3 5 4 4 3 3 5 3 6 6 5
Home warm
CS 4 6 4 1 3 5 6 6 4 6 6 3 4 5 6 4 5 4 2 5 3
LONGI 4 6 4 1 3 5 5 6 4 6 6 4 3 5 6 6 4 4 2 5 4
Arrears
CS 5 4 3 5 4 6 3 5 3 3 4 5 6 3 5 5 3 5 6 6 5
LONGI 5 4 3 5 4 6 3 3 3 3 4 3 6 3 5 4 3 5 5 4 3
Car
CS 6 5 6 6 6 4 5 3 5 5 5 6 5 6 4 6 4 6 4 2 6
LONGI 6 5 6 6 6 3 6 5 6 4 5 6 5 6 4 5 6 6 4 3 6
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, and EU-SILC 2011 longitudinal data, Users’ database - August 2013. Notes: CS: cross-
sectional order; LONGI: longitudinal order. See Annex 2 for country abbreviations.
Note: The cross-sectional orders are based on the original results from Deutsch et al (2014) calculated on the full 13-item list. The seven items not available in the
longitudinal dataset were omitted, and the rank reallocated to the six remaining items. The longitudinal order was in contrast estimated directly on the six items. All cases
with no deprivations (a sum score of 0) and the few cases who suffered from all deprivations (sum score of 6) were excluded during estimation, as they provide no
information for the purposes of this model.
14
Table 7: Rank of the EU order in each country
Country Rank of EU
order6 Highest rank of order with holidays and unexpected
expenses as last (5th and 6th respectively)
Poland 1 515
Czech Republic 3 435
Malta 4 517
Italy 6 478
Bulgaria 8 483
Romania 13 498
Estonia 15 520
Hungary 16 519
Belgium 17 478
Lithuania 17 429
Spain 19 431
Austria 20 478
Cyprus 30 425
Latvia 30 541
United Kingdom 33 466
Portugal 46 381
Luxembourg 53 343
Netherlands 54 415
Denmark 134 355
Finland 162 251 Source: EU-SILC 2011 longitudinal data, Users’ database - August 2013
Reading notes: Out of 720 possible longitudinal hierarchies, the EU order (1) Holidays; 2) Unexpected
expenses; 3) Meat/chicken/fish; 4) Home warm; 5) Arrears; 6)Car) is the best 20th rank in AT.
The key message from the results above is that whereas the order of curtailment for holidays and
unexpected expenses is very similar across all countries, the other four items (meat, warm, arrears
and car) show more variability (both in cross-sectional and longitudinal analysis). Nevertheless,
the EU order revealed by the cross-sectional and longitudinal deprivation sequence methods
provides a good approximation of the order of curtailment of these four items.
Looking at aggregate trajectories into deprivation also corroborates the results above. Figure 4 is
obtained by looking at those individuals that were not experiencing Material Deprivation
according to the EU definition (i.e. lack two items or less, see Annex 1)7 in the previous year and
6 1) Holidays, 2) Unexpected expenses, 3) Meal, 4) Warm , 5) Arrears, 6) Car 7 The official Material Deprivation indicator identifies those suffering from three or more of the following deprivations:
holidays once a year, keeping one’s home adequately warm, facing unexpected expenses, avoiding arrears, a washing machine, TV, telephone or a car.
15
are now lacking three or more items. It shows the percentage of those who upon entering Material
Deprivation report not being able to afford the item reported at the top of each pane. The majority
(75% or more) of these are unable to afford holidays, with the exception of Denmark, where only
half suffer from this deprivation. A very similar pattern is found for unexpected expenses. Overall,
as we move down the EU hierarchy from holidays and unexpected expenses towards the items
lower down the hierarchy, fewer and fewer countries are left between the second and first vertical
lines on each graph, indicating that those who have just entered Material Deprivation are less
likely to curtail these items. For example, only in Denmark and Finland more than half of the
population experience Arrears, in contrast with deprivation patterns for Holidays and Unexpected
expenses.
Figure 4 Percentage of people lacking each item for those who have just (re)entered into
Material Deprivation (i.e. lack 3 or more items), pooled data
Source: EU-SILC longitudinal data, Users’ database - August 2013, authors’computation. Reading notes:
Each dot shows the percentage of people who (re)entered into deprivation (defined as lacking at least three
items out of nine (official EU commonly agreed indicator, see Annex 1) in year T, but not in year T-1) who
are deprived of the item on top of each pane. For example, the top-left pane shows that in DK 50% of those
who just (re) entered material deprivation cannot afford one week annual holiday away from home.
Similarly, Figure 5 shows that a large proportion of those entering Material Deprivation in a given
year (T) were already lacking holidays and could not face unexpected expenses the previous year
(T-1), but the majority of them did not experience the other four deprivations.
16
Figure 5 : Percentage of people who can’t afford each item, in T (year of entry into Material
Deprivation), split between those who already lack the item in T-1 (year before the entry
into MD) or not, pooled data
Source: EU-SILC longitudinal data, Users’ database - August 2013, authors’computation. Reading notes:
Reading note: Among those who entered into deprivation in T, more than 90% lacked holidays. A large
majority of these people already lacked holidays in T-1 (orange bar).
Figure4 and 5 also reiterate the greater rank variability across countries of items such as meat,
warm, car and arrears. The next section explores this issue further using Item Response Theory,
a methodology that ranks the deprivation severity of each item on a common deprivation scale.
3.3 Item Response Theory
Item Response Theory (IRT) models have been used in the measurement of deprivation by, among
others, Dickes (1983, 1989), Gailly and Hausman (1984), Pérez-Mayo (2004), Cappellari and
Jenkins (2006), Ayala and Navarro (2007 and 2008), Dickes and Fusco (2008), Guio, Gordon and
Marlier (2012) and Szeles and Fusco (2013). Also known as Latent Trait Analysis, IRT is a set of
statistical models that describe the relationship between questionnaire item responses and an
unobserved latent trait, such as academic ability, level of happiness or material deprivation. IRT
postulates a relationship between each item and the underlying deprivation trait, and this is best
represented using Item Characteristic Curves (ICCs). Figure 6 shows thirteen ICCs, which
illustrate the relationship between the underlying deprivation trait (comparable to a standardised
version of the deprivation score shown in
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Car Meat Arrears Warm Unexp Holidays
lacking the item in T but not in T-1 Lacking the item in T and T-1
17
Figure 1, or the sum of deprivations) and the probability of being deprived of each item: as
deprivation (shown on the X-axis, expressed in standard deviations (s.d.) from the mean)
increases, the probability of being deprived of an item (shown on the Y-axis) increases. The
further to the right the ICC the more severe the deprivation. The curves are ordered according to
the EU (cross-sectional) deprivation order (see table 4, first column), and the dotted curves
indicate item not present in the longitudinal element of the EU-SILC. The ICCs for the first two
items in the order (i.e. holidays and unexpected expenses) show variation between -1 and 1 s.d.:
as shown above these items detect the first signs of a drop in resources such as income, and the
vast majority of those who suffer from more extreme levels of material deprivation (e.g above 1
s.d.) cannot afford these. Looking at the horizontal distance between the curves (which reflects
the range of deprivation severity and is shown by the dashed horizontal line) shows that the ICCs
for these two items are close together but far apart from the other four items (meal, warm, car and
arrears) which were included in our longitudinal deprivation order; the severity of deprivation
associated with these two deprivations is distinctively lower than that of the other four items.
However, at higher levels of deprivation the probabilities of being deprived of these four items at
the bottom of the order (meal, warm, car and arrears) are very similar; the curves are so close
together that it is difficult to tell them apart, and therefore the order of curtailment for these items
is much harder to establish. These results give a potential explanation of why there is greater
variability in the order of curtailment of these items at national level and yet the EU order shows
on average a good fit across most countries. These four items indicate more severe levels of
deprivation than holidays and unexpected expenses, but their respective ranks in the order seem
interchangeable.
The ICCs also corroborate the results presented above: not being able to afford two pairs of shoes
is associated with extremely high levels of deprivation (the probability of enforced lack begins to
rise only at levels of deprivation above 1 s.d.), and therefore this represents the very last item to
be curtailed for most countries and population subgroups. The ICCs also reiterate the need to have
a broad range of items that capture all levels of deprivation, in both the cross-sectional and
longitudinal components of the survey. Among the 13 items proposed by Guio, Gordon and
Marlier (2012), those which were not yet collected in the core EU-SILC survey are crucial to
capture adequately the full range of deprivation severity.
18
Figure 6: Item Characteristic Curves (ICCs), 13 items (cross-sectional data), EU level
Source: EU-SILC 2009 cross-sectional data, Users’ database - August 2011, authors’computation
4 CONCLUSION
The Deprivation Sequence methodology developed by Deutsch and Silber (2008) in the context
of deprivation research proves to be an insightful methodology to detect orders of curtailment. As
shown in this paper, its simple and data-driven logic can easily accommodate longitudinal data.
Item Response Theory can also be used to explore some of these findings further and identifying
the overlap in national deprivation ranks across the EU.
19
The analysis presented in this paper shows that deprivation data can be used to build an insightful
narrative of the way people on low income are gradually excluded from some of the key aspects
of living conditions of each Member State. Cross sectional analyses showed that as their resources
decrease people generally first cut back on their annual holidays, their savings to face unexpected
expenses, new furniture, leisure and social activities. As their resources decrease even further,
they are unable to afford meals, a warm house and paying bills, and eventually even two pairs of
all-weather shoes. Although using a smaller set of items constrained by data availability, the value
added of this paper is to extend the cross-sectional methodology used in by Deutsch et al. (2014)
using longitudinal data. The analysis confirms that the same deprivation pattern is also found
when following the same people across time.
Across the European Union, the bad fit of a deprivation order in which expenditure on holiday
away from home is given priority over other goods and activities provides clear evidence against
claims that poverty is the result of erratic spending or inefficient household budgeting: the vast
majority of those without basic items such as shoes cannot afford holidays nor have enough
money to face unexpected expenses. It also highlights the importance of social activities such as
a monthly drink or meal with friends or family and reiterates the importance of seeing poverty as
a form of exclusion from ordinary living patterns, customs and activities (Townsend, 1979).
This type of analysis is also extremely important to confirm the validity and reliability of the EU
deprivation measures in general, and of the 13 item scale proposed by Guio, Gordon and Marlier
(2012) in particular. The analysis shows that these 13 items can be used to capture a large range
of material deprivation severity, which is not perfectly well captured by the items currently
collected in EU-SILC. This paper also furthered the use of Item Response Theory to understand
deprivation sequences with cross-sectional data and more specifically drew attention to how Item
Characteristic Curves can be used to build a more nuanced view of the overlap between the orders
of deprivation items identified by the deprivation sequence method. Item Response Theory
analysis also shows that questions on extreme deprivation such as two pairs of shoes are needed
in the longitudinal element of the EU-SILC to further corroborate the cross-sectional results and
more generally for capturing extreme levels of deprivation.
Although theories of consumption behaviour and relative prices may be useful to analyse detailed
expenditure studies which collect information on the cost and quality of household goods, we
argued that deprivation items available in surveys such as the EU-SILC seem to be less suited to
empirical exploration of such theories; detailed expenditure data for example may show how the
quality of certain goods is cut down as resources decrease and how individuals are (un)able to
find cheaper goods, while deprivation items simply signal the (enforced) lack of these. The
strength of much of the available deprivation items lie in their ability to detect the exclusion from
shared living patterns, customs and activities because of lack of resources. Nevertheless, future
research could use relative price theories to track changes in deprivation across several years when
more data is available and could use deprivation data in conjunction with expenditure data to
unify these two subject areas.
Finally, we acknowledge that the issues we have raised deserve further exploration, particularly
in understanding which formal and informal resources prevent people from sliding into extreme
levels of deprivation.
20
Annex 1: material deprivation indicators – definitions
Official EU material deprivation indicators: standard and severe material deprivation
Based on the information available from the EU Statistics on Income and Living Conditions (EU-
SILC) data-set, the “standard” EU MD rate is currently defined as the proportion of people living
in households who cannot afford at least three of the following nine items:
1. coping with unexpected expenses;
2. one week’s annual holiday away from home;
3. avoiding arrears (in mortgage or rent, utility bills or hire purchase instalments);
4. a meal with meat, chicken, fish or vegetarian equivalent every second day;
5. keeping the home adequately warm;
6. a washing machine;
7. a colour TV;
8. a telephone;
9. a personal car.
In June 2010, EU leaders launched the new “Europe 2020 Strategy” and set in this context an EU
social inclusion target, which consists of lifting at least 20 million people out of the risk of poverty
or social exclusion in the EU by 2020, which is based on three indicators. One of them is a
measure of “severe” deprivation, which is built in the same way as the “standard” measure but
with a threshold set at four rather than three enforced lacks.
Proposed revised material deprivation indicator based on 13 items collected in the cross-
sectional EU-SILC survey (2009)
In view of the revision of the current material deprivation indicator, Guio, Gordon and Marlier
(2012) analysed the full set of material deprivation items included in the 2009 thematic module
on material deprivation and the core survey and proposed a selection of 13 material deprivation
items which passed various robustness tests. These items, presented below, cover some key
aspects of living conditions which appear to be customary in the whole EU and from which some
people are excluded due to a lack of resources (and not because by choice – enforced lack).
a) “Adult items”, i.e. items collected at individual adult level (people aged 16+, living in
private households). The adult deprivation information is assigned to all household
members (including children), if at least half the adults in the household cannot afford
the item. The five items are:
1. to replace worn-out clothes by some new (not second-hand) ones;
2. to afford two pairs of properly fitting shoes, including a pair of all-weather shoes;
3. to spend a small amount of money each week on oneself without having to consult
anyone (hereafter referred to as “pocket money”);
4. to get together with friends/family for a drink/meal at least monthly;
5. to have regular leisure activities;
21
b) “Household items”, i.e. items collected at household level. The household deprivation
information is assigned to all household members (including children) when the
household cannot afford the item. The 8 items are:
6. to replace worn-out furniture (but would like to have);
7. to afford a meal with meat, chicken, fish or vegetarian equivalent every second day;
8. to face unexpected expenses;
9. to keep home adequately warm;
10. to afford one week annual holiday away from home;
11. to avoid arrears (mortgage or rent, utility bills or hire purchase instalments)
12. to afford/ have access to a car/van for private use (but would like to have)
13. to afford a computer and an internet connection (but would like to have)
Subset of 6 items available in the longitudinal EU-SILC survey
The longitudinal element of EU-SILC contains six of the original 13 items, which measure the
affordability:
1. to have a meal with meat, chicken, fish or vegetarian equivalent every second day;
2. to face unexpected expenses;
3. to keep home adequately warm;
4. to have one week annual holiday away from home;
5. to avoid arrears (mortgage or rent, utility bills or hire purchase instalments)
6. to have access to a car/van for private use (but would like to have)
22
Annex 2: EU countries’ official abbreviations
“Old” Member States “New” Member States
BE Belgium 2004 Enlargement
DK Denmark CZ Czech Republic
DE Germany EE Estonia
IE Ireland CY Cyprus
EL Greece LV Latvia
ES Spain LT Lithuania
FR France HU Hungary
IT Italy MT Malta
LU Luxembourg PL Poland
NL The Netherlands SI Slovenia
AT Austria SK Slovakia
PT Portugal
FI Finland 2007 Enlargement
SE Sweden BG Bulgaria
UK United Kingdom RO Romania
23
Annex 3: Best order of curtailment, cross-sectional data, by country, 2009
Source: The cross-sectional orders are based on the original results from Deutsch et al (2014) calculated on the full 13-item list available in EU-SILC 2009 cross-sectional
data, Users’ database - August 2011
EU-27 AT BE BG CY CZ DE DK EE EL ES FI FR HU IE IT LT LU LV MT NL PL PT RO SE SI SK UK
Holidays 1 2 1 3 2 2 2 2 1 2 1 2 3 2 2 1 2 2 2 1 3 1 1 1 2 3 1 2
Unexp. expenses 2 1 2 4 3 3 1 1 3 3 3 1 2 1 1 2 1 1 1 3 1 2 8 7 1 1 3 1
Furniture 3 5 3 1 1 1 6 3 2 1 2 3 1 3 3 11 3 3 3 2 2 3 2 2 6 2 2 6
Leisure 5 3 4 8 6 6 4 6 7 6 5 7 5 5 7 4 4 5 5 5 4 4 5 4 5 5 7 4
Pocket money 4 4 6 6 8 5 5 5 5 8 4 6 4 6 5 3 5 4 6 6 5 5 4 5 4 7 6 3
Drink/meal out 6 6 5 7 9 10 3 7 8 9 6 8 8 4 6 5 6 6 7 4 6 6 6 3 8 9 8 5
Clothes 7 8 7 5 7 9 8 4 6 7 8 5 7 7 13 6 7 7 4 7 7 7 7 8 7 6 9 7
Meat/chicken/fish 8 7 10 9 10 4 7 10 9 10 13 9 9 8 11 8 8 10 8 8 12 8 12 9 11 8 4 9
Home warm 9 12 9 2 4 11 9 12 13 5 9 12 11 11 9 7 9 11 11 9 11 9 3 10 12 10 13 8
Car 11 10 11 11 12 7 12 11 4 12 12 10 12 10 8 13 10 12 9 11 9 11 10 6 9 12 5 12
Arrears 10 9 8 10 5 12 10 8 10 4 7 4 6 9 4 9 11 9 10 10 8 10 13 11 3 4 11 11
Computer/Internet 12 11 12 12 13 8 13 13 11 11 10 11 13 12 10 10 12 13 12 12 13 12 11 12 13 11 10 13
Shoes 13 13 13 13 11 13 11 9 12 13 11 13 10 13 12 12 13 8 13 13 10 13 9 13 10 13 12 10
24
REFERENCES
Anderson, W., White, V. and Finney, A. (2010) "You Just Have to Get by" Coping with Low
Incomes and Cold Homes. Bristol: Centre for Sustainable Energy.
Ayala, L. & Navarro, C. (2008). Multidimensional indices of housing deprivation with
applications to Spain. Applied Economics, 40(5), 597–611.
Batty, E., Cole, I. and Green, S. (2011)Low-Income Neighbourhoods in Britain. The Gap
between Policy Ideas and Residents Realities. York: Joseph Rowntree Foundation.
Berthoud, Richard and Bryan, Mark L (2011) 'Income, Deprivation and Poverty: A Longitudinal
Analysis.' Journal of Social Policy, 40 (01). pp. 135-156.
Cappellari, L., & Jenkins, S. P. (2007). Summarizing multiple deprivation indicators. In S. P.
Jenkins & J. Micklewright (Eds.), Inequality and poverty: Re-examined (pp. 166–184). Oxford:
Oxford University Press.
Dean, H. and Shah, A. (2002) 'Insecure Families and Low-Paying Labour Markets: Comments on
the British Experience', Journal of Social Policy, 31 (1): 61-80.
Deaton, A. & Muellbauer, J. 1980. “An almost ideal demand system”. American Economics
Review, 70.
Deutsch, J., & Silber, J. (2008). The order of acquisition of durable goods and the
multidimensional measurement of poverty. In N. Kakwani & J. Silber (Eds.), Quantitative
approaches to multidimensional poverty measurement. Basingstoke: Palgrave-Macmillan.
Dickes, P. (1983). Modèle de Rasch pour items dichotomiques: Théorie, Technique et application
à la mesure de la pauvreté. Nancy: Université de Nancy II.
Dickes, P. (1989). Pauvreté et Conditions d’Existence. Théories, Modèles et Mesures. Document
PSELL n°8. Walferdange:CEPS/INSTEAD.
Dickes, P., & Fusco, A. (2008). The rasch model and multidimensional poverty measurement. In
N. Kakwani & J. Silber (Eds.), Quantitative approaches to multidimensional poverty
measurement (pp. 49–62). New York: Palgrave Macmillan.
Engel (1895), Engel, E., ‘‘Die Lebenskosten belgischer Arbeiter-Familien fruher and jetzt,’’
International Statistical Institute Bulletin 9 (1895), 1–74.
Fahmy, E. and Pemberton, S. (2008)Show and Tell: Multi-Media Testimony on Rural Poverty
and Exclusion. Hereford: Rural Media Company.
Farrell, C. and O’Connor, W. (2003) Low-income families and household spending. Norwich:
Department for Work and Pensions.
25
Gailly, B., & Hausman, P. (1984). Des Désavantages Relatifs à une Mesure Objective de la
Pauvreté. In G. Sarpellon (Ed.), Understanding poverty(pp. 192–216). Milan: Franco Angeli.
Gregg, P., Waldfogel, J. and Washbrook, E. (2005) ‘That’s the way the money goes: expenditure
patterns as real incomes rise for the poorest families with children’, A more equal society, pp.
251–275.
Guio, A. -C. (2009). What can be learned from deprivation indicators in Europe? Eurostat
methodologies and working paper, Luxembourg: Eurostat.
Guio, A. -C., Gordon D. & Marlier E. (2012). Measuring material deprivation in the EU:
Indicators for the whole population and child-specific indicators, Eurostat Methodologies and
working papers, Luxembourg: Office for Official Publications of the European
Communities(OPOCE).
Guttman, L. (1950). The basis for scalogram analysis. In A. S. Stouffer et al. Measurement and
Prediction. The American Soldier Vol. IV. New York: Wiley.
Hebden J. J. and Pickering, J. F. "Patterns of Acquisition of Consumer Durables," Oxford Bulletin
of Economics and Statistics, 36 (May 1974), 67-94.
Kempson, E. (1996) Life on a Low Income. York: Joseph Rowntree Foundation.
Leser, C.E. 1963. “Forms of Engel functions”. Econometrica, 31: 694-703.
Mack, J. and Lansley, S. Poor Britain, London, George Allen & Unwin.
McFall, J. (1969) "Priority Patterns and Consumer Behavior," Journal of Marketing, 33 (October),
50-55
Orr, S., Brown, G., Smith, S. J., May, C. and Waters, M. (2006)When Ends Don’t Meet:
Assets, Vulnerabilities and Livelihoods. An Analysis of Households in Thornaby-on-Tees.
Chuch Action on Poverty and Oxfam.
Paroush J. (1965), "The Order of Acquisition of Consumer Durables,'' Econometrica, 33 (January
1965), 225-235.
Paroush, J. (1973). Efficient purchasing behavior and order relations in
consumption. Kyklos, 26(1), 91–112.
Pemberton, S., Sutton, E. and Fahmy, E. (2013) A review of the qualitative evidence relating to
the experience of poverty and exclusion. Bristol, UK: Poverty and Social Exclusion in the UK
(Working Paper - Methods Series No.22), pp. 1–59.
Pérez-Mayo, J. (2004). Consistent poverty dynamics in Spain. IRISS Working Paper Series N0.
2004–09, Differdange, Luxembourg.
Roos, C.F. and Von Szelisk, V. S. (1943), The Demand for Durable Goods, Econometrica, Vol.
11, No. 2, Apr., 1943.
26
Yeung, W. J. and Hofferth, S. L. (1998) ‘Family adaptations to income and job loss in the US’,
Journal of Family and Economic Issues, 19(3), pp. 255–283.
Saunders, P., Sutherland, K. Davidson, P., Hampshire, A. King,S. and Taylor, J. (2006)
Experiencing Poverty: The Voices of Low-Income Australians. Towards New Indicators of
disadvantage Project. Stage:1 Focus Group Outcomes. Sydney: SPRC.
Szeles, M., & Fusco, A. (2013). Item response theory and the measurement of deprivation:
Evidence from Luxembourg data. Quality & Quantity,47(3), 15–45.
Smith, D. M. (2005) On the Margins of Inclusion: Changing Labour Markets and Social
Exclusion in London. Bristol: The Policy Press.
Townsend, P. (1979). Poverty in the United Kingdom. London: Penguin.
Working, H. (1943),. Statistical laws of family expenditure. Journal of the American Statistical
Association, 33: 43-56.
Wright, F. (2004) 'Old and Cold: Older People and Policies Failing to Address Fuel Poverty',
Social Policy & Administration, 38 (5): 488-503.