The Comparative Size of Countries Within Europe

The comparative size of countries within Europe Tom CrowardsSept. 2002

The comparative size of countries within Europe

by Tom Crowards

Abstract

Cluster analysis is used to identify different groupings, according to size, within 31 European countries. The variables used are population, land area, gross domestic product, foreign direct investment and trade openness. The results identify eight ‘small’ states (including two ‘micro’ states), seven ‘large’ states, and a remainder of 16 ‘medium-sized’ states. Countries that are candidates for accession to the European Union make up the majority of the ‘small’ states and also of a sub-group of medium-sized countries that are clearly ‘medium-small’. The accession of all candidate countries into the European Union would result in a far greater predominance of smaller states within the Union.

Introduction

This paper seeks to determine which countries within Europe can be classified as ‘small’ states. Methodology is adapted from that employed by Crowards (2002) to define a category of ‘small’ states from among the developing countries of the world. The countries included in the present study are the 15 European Union (EU) member states, the 13 states that are presently candidates for entry into the EU, and the three larger remaining members of the European Free Trade Association (EFTA). These countries are listed in Table 1, along with the data employed in this study. Excluded from this study are the exceedingly small European states of Andorra, Liechtenstein, Monaco, San Marino and the Vatican City. There is absolutely no question that these states are ‘small’ and the necessary data is hard to come by.

Which states within Europe are categorised as ‘small’ has emerged as a prominent issue relating to plans for expansion of the EU. The background and the arguments are clearly described in Archer and Nugent (2002) in a special issue of the journal Current Politics and Economics of Europe looking at small states and European integration. In essence, there are mechanisms that allow smaller states to exert a comparatively high degree of influence through the European Union, diminishing the relative influence of the larger states. Many of the countries that are candidates for accession to the EU have characteristics associated with small states. To inform further assessment of this issue it is important to determine which countries within Europe are ‘small’, which this paper seeks to do.

Choice of variables

As Archer and Nugent (2002) state, “There is no one wholly satisfactory definition of what constitutes a small state.” The most commonly used criterion is population size. However, a primary concern with determining size using any single parameter is that it involves subjective selection of largely arbitrary cut-off lines.

Land area is another common criterion in assessing the size of countries. However, huge differences in the resources available, utilisation of the land and density of population reduce the value of this parameter, certainly when employed in isolation.

A third useful parameter is that of national income, usually measured as Gross Domestic Product (GDP), giving an indication of comparative economic size and hence to some extent political weight.

Archer and Nugent list a range of features that might characterise small states. Some that might be relatively easily measured include greater trade openness, proportionately larger national administrations, and vulnerability to economic fluctuations (perhaps with more rapid recovery from adverse shocks).

To represent trade openness the variable ‘total exports plus imports as a percentage of GDP’ is used in this study. It has not been possible to identify and incorporate a measure for the size of the national administration. The value of this measure is questionable in the present context since it will be particularly distorted by the overhanging predominance of central administrations in previously eastern bloc countries that make up the majority of the ‘candidate’ countries. A brief analysis of the fluctuations in GDP growth for the countries in the sample over the period 1996-2000 showed no indication that volatility of income (measured as the average of absolute GDP growth over the period) followed any particular pattern. This variable has not, therefore, been used in the present study.

A fifth and final variable that has been introduced is that of net inflows of foreign direct investment. This provides some measure of the importance of international interactions and the ability of a country to influence some decisions made beyond its borders. It would be expected that smaller countries would attract smaller amounts of international investment.

How size is defined depends somewhat upon the use to which the definition is to be put. For instance, Crowards (2002) focussed primarily on economic size while Archer and Nugent also consider the issue of size in “diplomatic and international relations terms”. Of the five variables employed, the first three - population, area and total income - are considered the most important and relevant to most definitions of size. The additional two variables - trade openness and net foreign direct investment - are considered secondary in that they are not direct indications of size.

There is potential concern that addition of these secondary variables might simply be reinforcing an aspect of size captured by one of the other three variables. This could be identified as a strong correlation between these secondary variables and the other variables. Linear regression was carried out using two of the variables in turn, with the resulting R2 statistics being presented in Table 2. An R2 of close to 1 illustrates strong correlation between two variables, while an R2 of close zero suggests little correlation.

Table 2 shows the strongest correlation to be between GDP and Population. Population is somewhat correlated with Land Area, and GDP is less so. The strongest correlation with the secondary variables is GDP with Exports Plus Imports. However, Exports Plus Imports is also correlated with Land Area, while the other secondary variable FDI shows reasonable correlation with Population. It is clear that the secondary variables as a group are not correlated with any other particular variable. Moreover, the lowest correlation of all is between these two secondary variables, indicating that they are unlikely to influence the grouping of countries in a very similar fashion.

The sample of countries employed in this study is considered reasonably homogenous on the basis that they are all members of the EU or the EFTA, or are candidates for membership of the EU. A comparison of relative size within the sample is therefore considered appropriate.

Cluster analysis

The methodology employed is based on that of Crowards (2002). First, an observational assessment is made for each variable in turn to see if there are any obvious breaks in the series. Charts of each of the five variables, with the countries ranked in ascending order, are presented in Figures 1 to 5. A summary of the observed breaks in the series is provided in Table 3. The breaks in the series show no clear pattern, so there is little guidance from this exercise as to the number of countries that might be expected within size groupings determined by simultaneous assessment of all variables[1].

The second part of the methodology involves applying cluster analysis to the data. Cluster analysis is a statistical technique for identifying relatively homogenous groupings with a set of data, based on distances between data points. It enables a number of variables to contribute simultaneously to the determination of groupings.

There is no single decision rule that can be applied to definitively determine whether the results of cluster analysis based on one set of parameters (e.g. the number of clusters and their initial cluster centres) is better than the results based on an alternative set of parameters. A general rule is that the outcome should be of practical use and satisfy common sense criteria. In the absence of guidelines that might have been provided by the observational analysis, a range of combinations of parameters was employed in the cluster analysis so that any pattern in the results could be assessed.

Acher and Nugent (2002) point out that a simple distinction between ‘small’ and ‘large’ states may have little meaning and that ‘medium’ and ‘micro’ size categories also need to be considered. This makes sense based on cursory observation of the series, with some extreme maximum and minimum values and generally a relatively large central grouping. It suggests that analysis to determine 4 clusters might be most appropriate, identifying groups of micro, small, medium and large countries.

Non-hierarchical cluster analysis is employed, in preference to hierarchical cluster analysis, for reasons outlined in Crowards (2000). Essentially, non-hierarchical cluster analysis fixes the number of clusters to a pre-determined ‘reasonable’ target and enables the membership of each cluster to be refined through a series of iterations.

In addition to identifying 4 clusters, separate analyses were carried out with 2 clusters, 3 clusters and 5 clusters. Selecting the number of clusters to be employed is an important and essentially subjective decision to be made in non-hierarchical cluster analysis. Varying the number of clusters employed enables assessment of the robustness of the results to changes in this parameter.

Another parameter that can be varied is the starting position of each of the cluster centres. Crowards (2002) chose to apply equally-spaced cluster centres. For the considerably smaller sample size in the present study, with variables that are far from uniformly distributed, no initial cluster centres were specified. Rather, the default option was employed of allowing the computer software to automatically determine approximate cluster centres as the first iteration in the process.

The computer software employed was SPSS, using its K-means cluster analysis to determine final cluster centres that minimise the Euclidean distance (the sum of the squared distances over all variables) between objects and the centre for each cluster. The software performs a non-hierarchical cluster analysis with a process that refines membership of each cluster based on a series of iterative steps.

The analysis for each number of clusters was repeated for alternative ranges of variables: just the three primary variables (population, land area and GDP); four variables (including either FDI or trade openness); and all five variables. Again, this was to test the robustness of the results to altering the parameters of the analysis.

Results

A summary of the results is provided in Table 4. The primary results are considered to be those from 4-cluster analysis. The a priori goal is to identify 4 country groupings. The groupings of ‘small’ and ‘micro’ countries are identical when 4 clusters are determined, no matter how many variables are employed, producing relatively unambiguous results for what is categorised as ‘small’.

The groupings of ‘small’ and ‘micro’ countries are very similar when either 2 clusters or 5 clusters are determined. The minor indiscrepancies in the results involve the Slovak Republic which for three of the twelve analyses reported in Table 4 is classified as small, and Ireland and Bulgaria which are classified as small in just one of the twelve analyses.

No results are presented for the 3-cluster analysis. This is because the requirement to define 3 distinct groupings tended to split the sample along lines that are inconsistent with the other results and are not particularly useful in the present context.

There are seven countries that are consistently classed as ‘large’, forming a clearly defined group. An attempt was initially made to distinguish between medium-small and medium-large states within the relatively large ‘medium-sized’ grouping. However, the results of various 5-cluster and 4-cluster analyses provide ambiguous results, attributable in particular to the sometimes contrasting influence of the secondary FDI and Trade Openness variables. A sub-group of countries was identified that is reasonably unambiguously ‘medium-small’. However, there are other countries that could also fit into this category - unlike the other groupings (‘micro’, ‘small’ and ‘large’) there is no clear dividing line between medium-small and medium-large states.

The country groupings that have been identified are presented in the final column of Table 4, to enable comparison with the outcome of each cluster analysis. They are also summarised more clearly in Table 5.

Figures 6 to 9 illustrate the groupings of countries on scatter graphs of two of the variables at a time: population and land area; population and GDP; population and FDI; and population and ‘exports plus imports’. Population is used as a common variable in each of these Figures since it is the single variable most commonly used to estimate country size. These Figures confirm the clear grouping of ‘small’ states and also of ‘large’ states.

Conclusion

The final determination of what are ‘small’ European states (including the two ‘micro’ states) includes just one EU member state, six candidate countries, and one other country[2]. The sub-group of clearly medium-small states includes four more candidate countries. This leaves just three other candidate countries in the medium and large categories. It is clear, therefore, that the accession of all candidate countries would dramatically shift the balance of EU membership away from larger states towards smaller states.

References

Archer, C. and Nugent, N. (2002), ‘Introduction: Small States and the European Union’, Current Politics and Economics of Europe.

Crowards, T.M. (2000), ‘Defining the category of ‘small’ states’, Staff Working Paper No. 5/00, June, Caribbean Development Bank, Barbados.

Crowards, T.M. (2002), ‘Defining the category of ‘small’ states’, Journal of International Development, 14(2), March, 143-179.

Table 1. List of Countries and Associated Data

Population / Area / GDP / FDI / XM
1000
(yr 2000) / km2 (yr2000) / current prices, yr 2000, 1000mn PPS / US$ billion,
net inflow (avg96-00) / X+M % GDP (1999 or 2000)
EU Member States
B / Belgium / 10,262 / 30,520 / 248.3 / 102.5 / 172.7
DK / Denmark / 5,349 / 43,075 / 176.5 / 113.1 / 79.5
D / Germany / 82,193 / 356,840 / 2,025.5 / 575.0 / 66.3
EL / Greece / 10,565 / 131,985 / 123.0 / 9.2 / 48.7
E / Spain / 39,490 / 504,880 / 608.8 / 153.3 / 62.2
F / France / 59,521 / 543,965 / 1,404.8 / 328.7 / 55.9
IRL / Ireland / 3,820 / 68,895 / 103.5 / 115.6 / 161.4
I / Italy / 57,844 / 301,245 / 1,165.7 / 60.0 / 55.6
L / Luxembourg / 441 / 2,585 / 20.9 / 218.4
NL / Netherlands / 15,983 / 33,940 / 401.1 / 322.1 / 116.4
A / Austria / 8,121 / 83,855 / 204.8 / 47.7 / 90.7
P / Portugal / 10,023 / 91,630 / 115.3 / 29.3 / 74.7
FIN / Finland / 5,181 / 337,030 / 131.7 / 58.1 / 74.8
S / Sweden / 8,883 / 449,790 / 246.6 / 233.4 / 89.5
UK / UK / 59,823 / 244,755 / 1,547.9 / 722.5 / 56.3
EFTA Member States
IS / Iceland / 283 / 102,820 / 9.5 / 1.2 / 73.1
NO / Norway / 4,503 / 323,895 / 175.5 / 49.8 / 77.1
CH / Switzerland / 7,206 / 41,285 / 259.6 / 102.5 / 79.1
EU Candidate Countries
BG / Bulgaria / 8,170 / 110,911 / 51.4 / 5.9 / 122.5
CY / Cyprus / 671 / 9,241 / 12.9 / 1.0 / 93.0
CZ / Czech Rep / 10,272 / 78,860 / 135.5 / 34.6 / 146.6
EE / Estonia / 1,436 / 43,431 / 12.4 / 3.4 / 172.2
HU / Hungary / 10,024 / 93,029 / 115.1 / 20.3 / 129.2
LV / Latvia / 2,417 / 64,589 / 15.9 / 4.0 / 100.1
LT / Lithuania / 3,696 / 65,300 / 27.6 / 4.6 / 96.7
MT / Malta / 390 / 316 / 4.9 / 4.2 / 216.7
PL / Poland / 38,649 / 312,685 / 342.1 / 64.8 / 61.8
RO / Romania / 22,443 / 238,391 / 117.3 / 11.2 / 73.9
SK / Slovak Rep / 5,401 / 49,035 / 58.1 / 7.0 / 149.6
SI / Slovenia / 1,989 / 20,273 / 31.0 / 2.3 / 121.8
TR / Turkey / 65,303 / 779,452 / 397.5 / 8.5 / 55.8

Notes:

Population, land area and GDP data come from the EU internet site.
Data for FDI (foreign direct investment) net inflows (US$ billions, in current prices) averaged over the years 1996-2000 come from the World Bank internet site.
Data for XM (eXports plus iMports as a percentage of GDP, the ‘trade openness’ variable) come from the World Bank internet site.

Table 2. Correlation between the variables

Population / Land Area / GDP / FDI
Land Area / 0.59
GDP / 0.85 / 0.48
FDI / 0.54 / 0.26 / 0.34
X+M / 0.36 / 0.61 / 0.76 / 0.15

Table 3. Summary of observable breaks identified in individual series

Number of countries / % of countries /

Population

/ Land Area / GDP / FDI / X+M
2 / 6% / 180
4 / 13% / 1,000 / 20,000
10 / 32% / 100
11 / 35% / 110
12 / 39% / 15
16 / 52% / 85
17 / 55% / 150
20 / 65% / 200,000 / 80
22 / 71% / 13,500
23 / 74% / 300 / 75
26 / 84% / 50,000 / 280
27 / 87% / 400,000 / 650

Table 4. Results of Cluster Analysis

Number of VARIABLES / 3 / 4fdi / 4xm / 5 / 3 / 4fdi / 4xm / 5 / 3 / 4fdi / 4xm / 5 / OVER-
Number of CLUSTERS / 2 / 2 / 2 / 2 / 4 / 4 / 4 / 4 / 5 / 5 / 5 / 5 / ALL
1 / B / Belgium / ML / ML / ML / ML / M / M / M / M / M / Ml / Ms / Ms / M
2 / DK / Denmark / ML / ML / ML / ML / M / M / M / M / M / Ml / Ml / Ml / M
3 / D / Germany / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
4 / EL / Greece / ML / ML / ML / ML / M / M / L / L / M / Ms / Ml / Ml / M
5 / E / Spain / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
6 / F / France / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
7 / IRL / Ireland / ML / ML / ML / ML / M / M / M / M / S2 / Ml / Ms / Ms / Ms
8 / I / Italy / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
9 / L / Luxembourg / S / S / S / S / m / m / m / m / m / m / m / m / m
10 / NL / Netherlands / ML / ML / ML / ML / M / M / M / M / M / Ml / Ms / Ms / M
11 / A / Austria / ML / ML / ML / ML / M / M / M / M / M / Ml / Ml / Ml / M
12 / P / Portugal / ML / ML / ML / ML / M / M / M / M / M / Ms / Ml / Ml / M
13 / FIN / Finland / ML / ML / ML / ML / M / M / L / L / M / Ms / Ml / Ml / M
14 / S / Sweden / ML / ML / ML / ML / M / L / L / L / M / Ml / Ml / Ml / M
15 / UK / UK / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
1 / IS / Iceland? / S / S / S / S / S / S / S / S / S1 / S / S / S / S
2 / NO / Norway / ML / ML / ML / ML / M / M / L / L / M / Ms / Ml / Ml / M
3 / CH / Switzerland / ML / ML / ML / ML / M / M / M / M / M / Ml / Ml / Ml / M
1 / BG / Bulgaria / ML / ML / ML / ML / M / M / M / M / S2 / Ms / Ms / Ms / Ms
2 / CY / Cyprus / S / S / S / S / S / S / S / S / S1 / S / S / S / S
3 / CZ / Czech Rep / ML / ML / ML / ML / M / M / M / M / M / Ms / Ms / Ms / Ms
4 / EE / Estonia / S / S / S / S / S / S / S / S / S1 / S / S / S / S
5 / HU / Hungary / ML / ML / ML / ML / M / M / M / M / M / Ms / Ms / Ms / Ms
6 / LV / Latvia / S / S / S / S / S / S / S / S / S2 / S / S / S / S
7 / LT / Lithuania / S / S / ML / S / S / S / S / S / S2 / S / S / S / S
8 / MT / Malta / S / S / S / S / m / m / m / m / m / m / m / m / m
9 / PL / Poland / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L
10 / RO / Romania / ML / ML / ML / ML / M / M / L / L / M / Ms / Ml / Ml / M
11 / SK / Slovak Rep / ML / S / ML / S / M / M / M / M / S2 / Ms / Ms / Ms / Ms
12 / SI / Slovenia / S / S / S / S / S / S / S / S / S2 / S / S / S / S
13 / TR / Turkey / ML / ML / ML / ML / L / L / L / L / L / L / L / L / L

Notes to Table 4:

All analysis involves variables converted to z-values, and includes as a minimum the variables for population, land area and GDP.
“4fdi” as the number of variables indicates 4 variables employed with foreign direct investment as the extra variable, excluding the trade openness variable.
“4xm” as the number of variables indicates 4 variables employed with trade openness as the extra variable, excluding the foreign direct investment variable.
‘S’ denotes a grouping of small states (that can be split into S1 and S2)
‘ML’ for 2-cluster analysis denotes a grouping of medium and large states
‘m’ denotes a grouping of micro-states
‘L’ denotes a grouping of large states
‘M’ denotes a grouping of medium-sized states
‘Ms’ denotes a grouping of medium-small states
The final “OVERALL” column gives a subjective classification based on other columns in the Table

Table 5. Summary of the Results of Cluster Analysis

‘micro’ / ‘small’ / ‘medum-sized’ / ‘large’
Malta / Cyprus / Austria / Ireland(m-s) / France
Luxembourg / Estonia / Belgium / Netherlands / Germany
Iceland / Bulgaria(m-s) / Norway / Italy
Latvia / Czech Rep. (m-s) / Portugal / Poland
Lithuania / Denmark / Romania / Spain
Slovenia / Finland / Slovak Rep. (m-s) / Turkey
Greece / Switzerland / UK
Hungary(m-s) / Sweden

Notes:

(m-s) denotes a state that is clearly ‘medium-small’
countries in bold are ‘candidate’ countries
countries in italics are not members of the EU nor are candidate countries
the remaining countries are EU member states

p.1

[1] Crowards (2002) was more fortunate in that initial observation indicated clear similarities between the variables in the number of countries that might be expected to be included within each size grouping . This provided guidance as to whether the final results were reasonable or not.

[2] The group of ‘small’ countries would clearly also include the even smaller states that are omitted from the study - Andorra, Liechtenstein, Monaco, San Marino and the Vatican City.