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 (anne-catherine.guio@ceps.lu) 3 School of Social Sciences, University of Cardiff , UK (Marco.Pomati@cardiff.ac.uk)

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.