The Impact of Services on Economic Complexity: Service Sophistication asRoute for Economic Growth

Viktor Stojkoski1, Zoran Utkovski1,3, Ljupco Kocarev1,2,*

1 Macedonian Academy of Sciences and Arts, Skopje, Macedonia2 Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University Skopje,Macedonia3 Faculty of Computer Science, University Goce Delcev Stip, Macedonia

* E-mail: lkocarev@manu.edu.mk

Abstract

Economic complexity reflects the amount of knowledge that is embedded in the productive structure of aneconomy. By combining tools from network science and econometrics, a robust and stable relationship betweena country’s productive structure and its economic growth has been established. Here we report that not onlygoods but also services are important for predicting the rate at which countries will grow. By adopting aterminology which classifies manufactured goods and delivered services as products, we investigate the influenceof services on the country’s productive structure. In particular, we provide evidence that complexity indices forservices are in general higher than those for goods, which is reflected in a general tendency to rank countrieswith developed service sector higher than countries with economy centred on manufacturing of goods. Byfocusing on country dynamics based on experimental data, we investigate the impact of services on the economiccomplexity of countries measured in the product space (consisting of both goods and services). Importantly, weshow that diversification of service exports and its sophistication can provide an additional route for economicgrowth in both developing and developed countries.

Introduction

The concept of economic complexity has been recently introduced [1–5] to reflect the amount of knowledgethat is embedded in the productive structure of an economy. By leveraging tools from network science andeconometrics, it has been documented that a robust and stable relationship between a country’s productivestructure and its economic growth exists [1–8]. Central to the concept of economic complexity is the Method ofReflections (MR) [3, 4] according to which countries in the international market are ranked and the differencein their competitiveness is measured by using the information contained in the country-product matrix whichlinks products and countries according to Balassa’s [9] Revealed Comparative Advantage (RCA). The methodis based on the hypothesis of a linear relation (more precisely an arithmetic average) between the ubiquityof a product, and the competitiveness of its exporters. Being motivated by the triangular structure of realdata in the country-product matrix, in [10, 11] the authors propose an approach which, as they argue, is moreconsistent with respect to the economic arguments underlying this complexity approach to economics. At thefocus of their approach is the concept of fitness which serves as a measure for the competitiveness of a country.The theory is based on a highly non-linear relationship between the complexity of products and the fitness ofcountries producing them with the aim to better reflect the ideas underlying the arguments of a capabilitydriven economic competitiveness. Both methods are aligned with the observation derived from experimentaltrade data which suggests that countries tend to produce all the possible products they can, given their level oftechnology, thereby opposing (to a certain extent) standard economic theories which predict a high level of

1

arX

iv:1

604.

0628

4v1

[q-

fin.

EC

] 2

1 A

pr 2

016

specialization of countries in specific industrial sectors. While, in general, the results support the relationshipbetween a country’s productive structure and its economic growth, these concepts also introduce non-monetaryand non-income-based measures for a country’s economic complexity which uncovers its hidden potential fordevelopment and growth.

The concept of economic complexity, as originally introduced, focuses on the role of goods exported by acountry (also referred to as products in the relevant works) for the country’s competitiveness and economicgrowth. While manufacturing and the export of goods has traditionally been assumed to be the single mostimportant economic indicator, the role of services as drivers of growth has not been investigated to the sameextent. As per capita income increases, most countries witness a rising share of services in the total output,accounting to over 60 percent of global GDP in the last decade. The negligence of services in the policyand research debate may stem from earlier observations that services are associated with low productivityand are merely inputs in the production of goods. However, the mid 1990s saw two seemingly separate butrelated developments. First was the revolution in technology (in particular information and communicationtechnology-ICT) and, second, the rapid growth in transport, and trade – with the advent of globalisation andthe internet age [12]. As result, services are no longer exclusively an input for trade in goods but have becomea ”final export”, i.e. a product for direct consumption.

In view of these changes in the nature of services and their growing importance, we investigate their role inthe context of economic complexity. We adopt the nomenclature according to which products encompass bothgoods and services which is in line with current trends in economy where services are seen as products for directconsumption. Following this convention, here we refer to the amount of knowledge that is required to producea good or to deliver a service as productive knowledge. This knowledge can vary enormously from one good tothe next or from one service to the next. In a previous attempt to extend the concept of economic complexityto services, Hidalgo and Hausmann [4] used aggregated service data together with (highly) disaggregatedgoods data to evaluate the complexity of countries and products. For the reason that service data is still notdisaggregate enough to be included in the complexity calculations, they quickly dismissed the service inclusion.In order to circumvene the main problem associated with the inclusion of services in the model (which is thelack of highly disaggregated data on services, as opposed to goods), here we take a different path and aggregategoods data (to make it comparable to the service data), resulting in aggregated indices (i.e. indices based onaggregated data) as measures for the complexity of products and countries. We thereby show that, althoughsome information is inevitably lost in the process of aggregation, the inclusion of services in the model providesnew insights for countries and products. By analyzing the properties of the newly obtained indices, we reportthat not only goods but also services are important for predicting economic stability and growth. In fact, weprovide evidence that complexity indices for services are in general higher than those for goods, thus puttingmost services above goods in the complexity ranking. As result, countries with developed service sector rank ingeneral higher on the scale of aggregated economic complexity indices, as compared to the case when rankingis performed based on disaggregated goods data only.

While there remains the non-monetary and non-income-based dimension of the productive knowledge ashidden potential for development and growth, we also show that there is a relation between the economiccomplexity of a country measured in the product space (including goods and services) and the income percapita that country is able to generate measured with GDP per-capita adjusted for Purchasing Power Parity(PPP). Importantly, our results suggest that growth in service exports and its sophistication can provide anadditional route for economic growth in both developing and developed countries. We argue that the conversealso holds-the countries (including those with diversified goods portfolio, i.e. saturated goods space) which arenot able to accordingly populate the service space (i.e. diversify the service portfolio), may face diminishinggrowth prospects. These conclusions, which can not be inferred from the analysis based solely on goods data,are more consistent with the ”perceived” economic situation in some countries than otherwise suggested by theresults in [3, 4] and [10,11].

This paper is organized as follows. The Materials and Methods section starts by describing the data andhow it should be adjusted for adequate comparison of goods and services. In this section we also summarizethe Method of Reflections (MR), developed by Hidalgo and Hausmann [3], which we refer to as the ”LinearMethod” (LM), in order to distinguish from the ”Nonlinear Method” (NLM) proposed by Tacchella et al. [10].For consistency, we employ both methods to estimate the complexity of countries and products (including both

2

goods and services). The Results and Discussion section consists of five parts. In the first part we discussthe implications that arise as result of data aggregation and inclusion of services. Then, in the second partwe proceed by estimating few baseline regressions of growth on our aggregated complexity measures whichinclude both goods and services, and comparing them with standard (disaggregated) complexity measureswhich account for goods only. This allows us to study whether the aggregation of goods data and the inclusionof services still yields a complexity index that has a significant relationship with future growth. In the thirdpart we look at something that the branch of Economic Complexity has widely eschewed until now - thedifferences between the complexity of services and goods. We do this by comparing services and goods rankingsthrough both LM and NLM. The fourth part of this section examines how country rankings are affected bythe inclusion of services in the model. Finally, in the fifth part we address the additional information for theeconomic potential of the countries that our aggregated metrics create by providing examples for selectedcountries. The last section concludes our findings and gives suggestions for future research.

Materials and Methods

Data

To analyse the association between countries and products we collect international trade data from varioussources. On the one hand, there are not many service datasets; even those that exist do not provide qualitydata. The best dataset that can be found is the World Bank Trade in Services Database [13] based on theExtended Balance of Payments Classification. The dataset consolidates multiple sources of bilateral trade datain services and calculates the export trade flows of a reporter by using information on imports of the partnercountry. Since the nature of the trade in services makes it difficult to collect service data as accurately as goodsdata, the dataset is still far away from being comparable to the goods data. Therefore, in order to be usedproperly, we disaggregate it to 3-digit level. With this classification 12 different types of services are recognized.On the other hand, there is plenty of goods data. For the period of 1988-2000 we use the Feenstra et al. [14]dataset from the Center for International Data which disaggregates the goods to SITC rev.2 to 4-digit level.The dataset itself is an advanced version of the UN COMTRADE database as it cleans the data from it in thesame way as the Trade in Services Database does. We extend the dataset until 2010 by using the trade dataprovided by UN COMTRADE [15]. To make the goods data comparable to the services we aggregate it to theStandard International Trade Classification(SITC) rev.2 1-digit level, thus ending with 10 classes of goods.

Table 1 shows the resulting goods and services. For further cleaning of unreliable or inadequately classifieddata, we restrict the dataset to countries that have population larger than 1.2 million and also exclude Chad,Iraq and Macau [4]. Even though that through this aggregated classification we end up with only 22 broadcategories of products (10 categories of goods and 12 categories of services) and a total of 130 countries, webelieve that they are detailed enough to provide new interesting facts about the complexity of goods, servicesand the countries exporting them. Therefore, we utilize it to calculate our aggregated complexity measures.

For the baseline regression analysis of growth on the aggregated complexity metrics we use data on GDPper capita, exports and population from the World Bank World Development Indicators Database, whereasdata from the Observatory of Economic Complexity [16] is taken to individually calculate the ComplexityOutlook Gain.

Methodology

Bipartite Network of Countries and Products: Following Hidalgo and Hausmann [1–5], the collecteddata are interpreted as a bipartite network in which countries are connected to the products (here both goodsand services) they export. According to this representation, the weights of the links in the bipartite networkare related to the Revealed Comparative Advantage (RCA) index introduced by Balassa [9]. RCA is widelyused in international economics for calculating the relative advantage or disadvantage of a certain country in acertain class of goods or services as evidenced by trade flows. Formally, it is defined as

RCAij =Eij/

∑j Eij∑

iEij/∑

i,j Eij(1)

3

Table 1. The included products and services

SITC Good EBOPS Service0 Food and live Animals 205 Transportation1 Beverages and Tobacco 236 Travel2 Crude materials, inedible, except fuels 245 Communication3 Mineral fuels, lubricants and related materials 249 Construcion4 Animal and vegetable oils, fats and waxes 253 Insurance5 Chemicals and related products, n.e.s. 260 Financial Services6 Manufactured goods 262 Computer and information7 Machinery and transport equipment 266 Royalties and License fees8 Miscellaneous manufactured articles 268 Other business services9 Commodities not classified elsewhere 287 Personal services

291 Government services983 Services not allocated

where E stands for exports, i is a country index, and j is a product index. In other words, the index RCAij

shows whether a country is significant exporter of a certain product or not. For example, RCAij > 1 indicatesthat the country’s share of a good or a service is larger than the good or service’s share of the entire worldmarket, thus ”revealing” a comparative advantage of the country with respect to that product.

The entries of the adjacency matrix M describing the links between countries and products in the bipartitenetwork (country-product space) may be related to the corresponding RCA index in different ways. In theoriginal, unweighted representation in [3, 4], the Mij entry of the adjacency matrix is defined as:

Mij =

{1, if RCAij ≥ 1

0, otherwise(2)

Tacchella et al. [10] extend the analysis to a weighted version of the bipartite country-product network byrewriting the Mij entry as:

Mij =Eij∑iEij

(3)

where the weight corresponds to the share of the world’s export of good or service j held by the country i.

Complexity of Products and Countries: The observations about the structure of the country-productmatrix have motivated a series of works, including [3,4] and [10,11] which interpret economic complexity byspecial fundamental endowments, called capabilities. Simply put, they represent the mixture of all availableresources in a given economy and the features of the national social organization which make possible theproduction and export of tradable products. As such, capabilities may be seen as intangibles assets which drivethe development, the wealth and the competitiveness of a country. In practice they determine the complexityof a productive system as each product requires a specific set of necessary capabilities which must be owned bya country in order to produce it and then to export it. Due to the difficulty in categorizing and quantitativelyanalysing capabilities, exported products by each country become in such a scenario the main proxy to inferthe endowment of capabilities, i.e. the level of complexity of a productive system.

In this context the simplest measures for country i’s and product j’s productive capabilities are introduced.The country measure is called diversity and can be estimated by calculating the country’s degree in the bipartitenetwork:

di =∑j

Mij (4)

whereas, the ubiquity of the product j, is given by the number of countries exporting that product:

uj =∑i

Mij (5)

4

Diversity and ubiquity are inversely related. Strictly speaking, higher diversity means that a country has anexport basket with many different products and it has a high amount of know-how. Whilst higher ubiquityrepresents that a good or a service is included in many countries’ export baskets, and thus it needs fewercapabilities to be produced. However, both diversity and ubiquity are simple graph characteristics (nodedegree) of the bipartite network represented by the adjacency matrix M which carry limited information aboutthe productive structure of a country or complexity of a product as they do not take into account who elseexports the same products. As result, a careful assessment is required if any of these simple measures is to beused for the explanation of economic phenomena.

A more accurate measure of the number of capabilities available in a country or required by a product(good or service) can be estimated through two different approaches (methods), summarized in the following.

Linear method for measuring product complexity: The first approach, due to Hidalgo and Hausmann[3, 4], is organised around an iterative linear method which is based on the premise that the number of acountry’s capabilities is equal to the average number of capabilities required by its exporting products, whereasthe number of capabilities required by a product is the average number of capabilities present in the countriesthat are exporting it. Because of the linear dependence between the complexity of a product and the complexityof the countries producing it we refer to this as the Linear Method (LM). The method can be written as:{

ci,n = 1di

∑j Mijpj,n−1

pj,n = 1uj

∑iMijci,n−1

where ci,n and pj,n respectively represent the complexity of a country and a product after n iterations,and the initial conditions are given as ci,0 = di and pj,0 = uj . By inserting the product (country) complexityequation into the country (product) complexity one can define ci,n (pj,n) as a linear transformation of ci,n−2(pj,n−2). When written in vectorial form, the linear operator characterizing this linear transformation representsthe transpose of an ergodic Markov transition operator [11]. This implies that all ci,n and pj,n convergeto a same constant with speed of convergence proportional to the second largest right eigenvalue. Thesemathematical properties of the linear method are discussed to a greater extent in [3, 11, 17] and are out of thescope of our research. For us it is important to note that the iterations can be evaded by generating two newmatrices for countries and products. Then, the capability measures can be obtained through ranking the nodesin each of these networks by calculating the second largest eigenvectors of the matrices. More formally, forcountries define the matrix [Cii′ ] as

Cii′ =∑j

MijMi′j

diuj

The country complexity measure, called Economic Complexity Index (ECI), is the normalized eigenvectorassociated with the second largest eigenvalue of C and is defined as

ECI =~c− 〈~c〉stdev(~c)

where 〈·〉 represents an average, stdev(·) stands for the standard deviation, and ~d is the eigenvector associatedwith the second largest eigenvalue of C = [Cii′ ].

Analogously, the linear product complexity metric is named Product Complexity Index (PCI) and is definedby exchanging the index of countries (i) with that for products (j) in the definitions above. In other words,the linear PCI is defined by analysing the matrix connecting product j to product j′, according to the numberof countries that export both products:

Pjj′ =∑i

MijMij′

diuj

The PCI is defined as

PCI =~p− 〈~p〉

stdev(~p)

where ~q is the eigenvector of Mjj′ associated with the second largest eigenvalue.

5

Nonlinear method for measuring product complexity: The second approach we address, introducedin [10,11], is conceptually similar to the approach due to Hidalgo and Hausmann. At the focus there is theconcept of Fitness which, as ECI, serves as a measure for a country’s competitiveness (complexity). Theirapproach is based on an iterative process which couples the Fitness of a country to the Complexity of a product,and defines the Fitness to be proportional to the sum of the exported products weighted by their number ofrequired capabilities. However, in contrast to the Linear Method, here the complexity of a product is no longerdefined as the average of the Fitnesses of the countries producing it. Instead, a strong nonlinear relationshipbetween the complexity of an exported product and the competitiveness of its producers is assumed. Dueto the present nonlinearity, we refer to it as the Nonlinear Method (NLM). Formally, the iterative process isdescribed as: ci,n =

∑j Mijpj,n−1

pj,n = 1∑i Mij

1ci,n−1

,

where ci,n and pj,n are respectively the intermediate values of the intensive Fitness (Complexity) of country iand the nonlinear complexity index of a product j after n iterations. After each step they are normalized:{

ci,n =ci,n〈ci,n〉

pj,n =pj,n

〈pj,n〉

The initial conditions for this method are ci,0 = 1 and pj,0 = 1 for all i and j. An attractive property of themethod is that the estimated metrics converge to a non-trivial unique fixed point after several iterations [18].

Weighted vs. unweighted representation As we will see in the Results section, the aggregation of thedata and the inclusion of services affect nontrivially the rankings of countries and products. Furthermore, thereis a qualitative and a quantitative difference between the results of the analyses with weighted, and respectively,unweighted representation of the bipartite network. In the following we summarize the main aspects of theweighted and the unweighted representation.

According to the unweighted representation (2), the element Mij of the adjacency matrix (corresponding tothe link between country i and product j in the bipartite network) is either 0 or 1, depending on whether thecountry i has a revealed comparative advantage with respect to the product j (RCAij ≥ 1) or not. This metricsis used by both Hausmann and Hidalgo [3, 4] (in the context of the economic complexity analysis), and byTacchella et al. [10] and Cristelli et al. [11] (in the context of the Fitness analysis). A simple motivation for theappropriateness of the metrics is reflected in the fact that a change in Mij from 0 to 1 is more important for theevolution of an individual country than the change in the world export share. The unweighted representationaims to provide non-monetary information about the long term potential of a country, by evaluating thecountry’s productive structure solely by investigating what is included in the country’s export basket and whoelse exports the same products. As such, this metric, also referred to as intensive in [11], is expected to followthe behavior of living standard indicators (such as GDP per capita).

As a generalisation of this (binary) metric, weights different from 0 and 1 may be given to the linksconnecting countries and products in the bipartite networks. A possible way to introduce weights is proposedby Tacchella et al. [10], as given by (3). With this particular choice of the weights in the adjacency matrix,the weighted metric gives a stronger emphasis to the monetary aspects and is more associated with economysize measures (such as the total GDP), than with non-monetary information such as the one contained in theunweighted metric. The weighted metric, also referred to as extensive, carries information about the short-termcompetitiveness of a nation, as it is initially structured to identify the most influential countries in the globalmarket dynamics in different product sectors [11]. We note that an attempt to introduce weights by the directuse of the RCA matrix is not justified, as this choice of the adjacency matrix coefficients would favour thecountries with a low diversification which, by chance, have natural resources in excess. In this sense, theweighted (i.e. the extensive) metrics does not extend (at least not in a straightforward way) to the economiccomplexity analysis in [3, 4].

6

Results and Discussion

Data Aggregation and Inclusion of Services: Implications

Before we proceed with a detailed analysis of the individual characteristics of the aggregated complexity ofcountries or compare services to goods, in this part we provide an initial intuition on how the aggregationof goods and the inclusion of services may affect economic complexity indicators and country ranking. Weaddress both the unweighted (binary) and the weighted representation of the adjacency matrix in the bipartitecountry-product network, as both models behave differently in light of data aggregation and addition of services.

Data aggregation: Intuitively, as data aggregation results in reduction of the number of products (now 22in total, including services), any change in the coefficients of the adjacency matrix in the unweighted (binary)representation (i.e. introduction/deletion of a link in the bipartite network), affects country rankings moresignificantly than in the case of the original model with 774 disaggregated goods. Indeed, small changes inthe RCA index may result in a ”jump” from 0 to 1 (or vice verse) in the unweighted representation wherematrix coefficients ”round off”. The impact of this effect is evident when diversity of countries based onaggregated goods is compared to diversity obtained on the basis on disaggregated goods (services are excludedto quantify the effect of aggregation only). For example, in the SITC 4-digit classification developed countriessuch as Germany, Italy and the United States have the most diverse export basket, whereas in the SITC 1-digitclassification they are replaced with not as developed countries. Nonetheless, this characteristic does not extendto the ubiquity analysis since in both representations chemicals and machinery are the least ubiquitous goods.With a smaller adjacency matrix (i.e. a smaller number of products) as result of aggregation, country rankingsare also expected to be affected. Indeed, country rankings based on both ECI and intensive fitness (IF) (whichrely on the unweighted representation) are more volatile (susceptible to higher variations over time) whenproducts are aggregated, compared to the original findings in [3, 4] (for ECI), and in [10,11] (for IF) based ondisaggregated goods data. This observation is consistent with the results presented later in this section, wherethe dynamics of country rankings is analysed over time.

Similarly to the unweighted representation, changes in the coefficients of the adjacency matrix in theweighted representation have bigger influence on the country rankings than in the case of the model withdisaggregated goods. However, this effect is here less profound, due to the absence of the ”rounding” of thematrix coefficients in the weighted representation. As result, country rankings based on the concept of extensivefitness (EF) presented in [10] (which relies on the weighted representation), are expected to be more stablewhen compared to the rankings based on ECI and IF (incorporating the unweighted representation). Thisobservation is supported by the results for the dynamics of country rankings (based on EF) presented later inthis section. Another property of the weighted representation (and of the concept of EF extracted therein) isthat ranking based on EF dynamics usually puts large economies on a higher position. This does not come notas a surprise, if we have in mind the monetary aspects associated to EF and the relation between EF and thetotal (not per capita) GDP, as documented in [10]. This is true for both the model based on disaggregated andaggregated data, as discussed in the next parts of this section.

Definitely, in the process of aggregation, some of the information revealed by the analyses based ondisaggregated goods data is inevitably lost. This is particularly true for the bipartite network model withunweighted (binary) adjacency matrix where, as result of aggregation, some countries loose the revealedcomparative advantage that they previously had in some goods (when goods were disaggregated), which isreflected in more profound changes in their overall export basket structure and their rankings (based on bothECI and IF). While this may be seen as a weakness of the model relying on aggregated goods data, we recallthe discussion that, in the absence of finer (i.e. disaggregated service data), the most plausible way to includeservices in the model is to aggregate goods data to a level comparable with available service data. Furthermore,the volatility in the rankings which comes as result of aggregation, is compensated for in the aggregated EFrankings which have relatively stable movements over time and, as such, are more similar to the disaggregatedones. The stability is a result of the initial structure of the weighted M matrix; a change in a weight has by farlesser impact on the final rankings than a binary change, especially with aggregation of data.

7

The role of services: To illustrate the effect of the inclusion of services in the model, in Fig 1a we presenta simple visualization of the unweighted country-product network for the year 2005.

According to this representation, goods and services are located at the center and counties are groupedinto regions (continents). The network consists of 116 countries, 10 products, 12 services, 450 links connectingcountries and goods, and 233 links connecting countries and services when RCAij > 1. Initially, the differencebetween the number of connections of services and goods suggests that there is a significant margin betweentheir complexity. This is also supported by the information contained in the binary M matrix for the sameyear after rows and columns are sorted according to decreasing diversity and ubiquity, as displayed in Fig.1b. We observe that the matrix preserves the triangular structure present in [2], thus implying that ubiquityand diversity are also inversely related in the model with aggregated goods and services. A more thoroughexamination shows that, among the countries, United Kingdom has the highest diversity (13), followed by theNetherlands (12), while Algeria, Saudi Arabia and Venezuela are least diversified with comparative advantagein only 1 product (Mineral fuels). The least ubiquitously exported products are Financial services, while Foodand live animals chiefly for food are mostly exported by the countries. By dividing the diversity into servicesand goods diversity, we discover additional information about the difference in the productive capabilitiesembedded in the services and the goods. Developed and highly diversified countries, such as United Kingdom(10), Switzerland (9) and the United States (9) have the most diverse service export baskets, whereas not asdeveloped and diversified countries, specifically Kenya (8) and Lebanon (8), have the highest goods diversity.However, this diversity analysis should be taken with caution since, as previously noted, diversity alone is notvery informative for evaluating the complexity of a productive system.

Fig 1. The Aggregated Country-Product Network. a) Visualizes the country-product network for2005. Services are given in blue and goods in red. Country nodes are colored according to the continents theybelong, Europe is in green, Asia in purple, Africa in yellow, and nodes representing American countries are incyan. b) Shows the M matrix. Here rows represent countries, ordered accordingly to their diversity, whilecolumns represent products ordered according their ubiquity.

8

The role of services seems to be particularly evident in the model based on the weighted representation,where fluctuations in country rankings based on extensive fitness may be attributed more to the inclusion ofservices than to aggregation of goods data itself, as suggested later on in this section. Altogether, the effects ofdata aggregation and inclusion of services significantly impact the productive structure of the countries. Inparticular, the addition of services may help us extract new valuable information about both the long-termand the short-term situation of the nations.

Economic Complexity as Indicator of Future Growth

The significance of economic complexity stems from the observation that the productive capabilities possessedby a country carry information about its long term potential. Ergo, the information contained in the adjacencymatrix describing the bipartite country-product network, may be used to predict country’s future growth. Inthis context, the performance of the Method of Reflections (here referred to as the Linear Method) has beenextensively evaluated through econometric tools in [3,4]. Similarly, the proponents of the nonlinear methodalso argue that the economic complexity indicators (such as intensive and extensive fitness) influence economicgrowth [10,11]. However, intensive and extensive fitness have distinct exploratory features: intensive fitnessis argued to provide information about robustness and long-term potential for economic growth; extensivefitness, on the other hand, is more tied to monetary indicators (such as total GDP) and is related to short-termgrowth predictions. In order to resolve inconsistencies appearing when standard regressions are used forgrowth prediction, in [8] it is argued that there is a heterogeneous pattern of evolution of country dynamics inthe fitness-income plane. Depending on whether a country has a lower or higher level of income comparedto expectations based on its level of fitness, two regimes are observed which correspond to very differentpredictability features for the evolution of countries: in the former regime one finds a strong predictablepattern, while the second regime exhibits a very low predictability. As result, the authors propose a selectivepredictability scheme which incorporates concepts from dynamical systems to deal with the heterogeneity ofthe evolution pattern.

According to these observations, we argue that one may position the complexity approach in the middlebetween two economic concepts - one which traditionally describes economic systems in purely GDP-orientedterms, and another which substitutes GDP with new economic indicators [19,20]. With this respect, we favorthe stand presented in [8] that instead of simply substituting monetary based information (such as GDP)with new, alternative indicators [19, 20], it is more appropriate to complement GDP-based measures with newdimensions. The role of GDP thus remains important in our analysis, with the remark that the results shouldbe interpreted based on the understanding that the country GDP (respectievly GDP per capita) is only a partof the picture or, more precisely, only one of the parameters in the multi-dimensional model.

While being aware of the potential weakness when used for prediction of future growth, attracted by itssimplicity, we stick to the standard regression analysis and examine the relationship between growth and theaggregated complexity measures. To do this, we first partially reproduce the main regressions presented in [4](where the complexity of a country is based on disaggregated data for goods exports) and then compare themto their aggregated counterparts. By examining the statistical properties of the indices, we demonstrate thatthe aggregation of goods and the addition of services in the model not only produces results for the EconomicComplexity Index (based on the linear method) which remain consistent with the predictions in [3, 4], but alsoremoves (to a certain extent) the inconsistency observed in the predictions obtained from the model based onthe Intensive Fitness (the nonlinear method) and disaggregated goods data. We believe that this behaviorcomes as result of the aggregation of the goods data and, more importantly, of the inclusion of services which,as we know, account for a large portion of the GDP.

In the following we provide details for the regression analysis performed for the model which relates linearlyproduct complexity with the complexities of the countries producing it (i.e. the Linear Method), and for themodel which incorporates a non linear relation between the complexity of a product and the complexities(fitnesses) of the countries that produce it.

9

The Linear Method: We examine the relationship between the aggregated complexity measures and longterm economic growth by estimating the following panel regression with fixed time effects:

log (GDPi,t+T

GDPi,t) = α+ βci,t + φit + ηt + εit

where the dependent variable is measured as the ten-year GDP (PPP) per capita growth of country i, from1988-1998 and 1998-2008, meaning that T is 10. The coefficient of β is of particular interest to us as it is anestimate of the marginal effect of ci,t, the economic complexity of the country at the initial time t; φi,t is a setof three controls; ηt is the time effect and εit is the error term. The three controls that are included are thelog of the initial GDP (PPP) per capita, the increase in natural resource exports and the initial ComplexityOutlook Index (COI) of the country. The first control helps us explain the catch-up effect, which states thatthe developing countries have potential to grow at faster rates, when compared to the developed ones. Sinceeconomic complexity does not capture the effects of exporting natural resources, we include the increase innatural resource exports in constant dollars as a share of initial GDP (natural resources match SITC categories0, 1, 2, 3, 4 and 68, same as Sachs and Warner [21]) as the second control. In addition, in every estimationci,t is accompanied by the COI of the country. The Complexity Outlook Index measures how many differentproducts are near a country’s set of productive capabilities. Countries with a low complexity outlook have fewnearby products and will find it difficult to acquire new capabilities and increase their economic complexity.

Table 2 gives the results from the performed regressions for the effect on growth of the measure calculatedthrough the linear method. The first column shows the estimated parameters when growth is regressed onlyon the first two controls, while the second adds the disaggregated ECI and COI to the regression. The thirdcolumn replaces the disaggregated ECI with the aggregated ECI. As a consequence, the value of R2 decreasesalmost by 10 percentage points. This is expected because the aggregated ECI takes less information about theproductive capabilities contained in the goods. Nevertheless, the index is a significant predictor of economicgrowth at any level. The fourth column combines the two indices. In this case, the aggregated ECI is no longersignificant. This appears as a consequence of the high colinearity between the indices - they have a PearsonCorrelation of 0.51 even when controlling for the income, natural resource exports and COI. The robustness ofour results is checked with two additional regressions where we include the initial export of goods and servicesas a percent of GDP and the initial population. The first variable is a measure of trade openness, and a highervalue means that a country is more willing to trade with others. By including the log of population in theregression, we account for the size of the country. Column V and V I present the results from the robustnesscheck. When adding our measure of trade openness, the aggregated ECI remains highly significant, whereaswhen adding the size of the country the significance slightly drops to 5% level. Nonetheless, the significance ishigh enough to conclude that our regression passes the robustness check.

The Nonlinear Method: In the original works which describe the nonlinear relation between productcomplexity and the fitness of the countries producing the product [10, 11], the authors use the HarmonizedSystem 2007 nomenclature for goods in order to estimate the (disaggregated) Intensive Fitness (IF), whichis slightly different from the one used in [3, 4] and in our work. Therefore, for a fair comparison, we firstre-estimate the disaggregated IF with the SITC rev.2 4-digit classification. Then, in Table 3 we redo thegrowth regressions with the logs of the estimated disaggregated IF and afterwards perform a regression analysisbased on the aggregated IF as an economic complexity measure. Our results for the country rankings withrespect to IF, based on the disaggregated data, are in a similar line with the findings in [8]. As suggested bythem, we find that the disaggregated version of the log of the initial IF is not as good predictor of growth asthe disaggregated ECI. In fact, the aggregated ECI with included services, explains income changes better by aslight margin. This may serve as a confirmation for the observation in [8] that, when used in the context ofthe nonlinear method, regression analysis is not the most appropriate tool for growth prediction. However,differently from [8] in these regressions the log of the initial IF is statistically significant at any level, and whenthe log of the aggregated IF is added to the regression both variables preserve their significance.

10

Table 2. Aggregated Economic Complexity Index and Growth.

Dependent Variable: Growth in GDP pc1988-1998, 1998-2008

Variable (I) (II) (III) (IV) (V) (VI)Income per capita, logs −0.031∗∗ −0.112∗∗∗ −0.046∗∗∗ −0.143∗∗∗ −0.054∗∗∗ −0.043∗∗∗

(0.013) (0.020) (0.015) (0.020) (0.017) (0.015)Increase in NR exports 0.334∗∗ 0.334∗∗∗ 0.346∗∗ 0.329∗∗ 0.343∗∗ 0.350∗∗

(0.160) (0.127) (0.160) (0.128) (0.169) (0.160)Initial COI 0.005∗∗∗ 0.004∗∗ 0.004∗∗∗ 0.004∗∗ 0.004∗∗

(0.002) (0.002) (0.002) (0.002) (0.002)Initial ECI 0.120∗∗∗ 0.127∗∗∗

(0.023) (0.027)Initial aggregated ECI 0.041∗∗∗ −0.010 0.043∗∗∗ 0.037∗∗

(0.014) (0.016) (0.014) (0.015)Initial Export of GS 0.001∗∗

(0.000)Initial Population, logs 0.008

(0.015)Constant 0.510∗∗∗ 1.237∗∗∗ 0.671∗∗∗ 1.238∗∗∗ 0.706∗∗∗ 0.520∗

(0.122) (0.180) (0.138) (0.180) (0.149) (0.301)Observations 210 210 210 210 202 210R2 0.208 0.331 0.240 0.332 0.241 0.242Year FE Yes Yes Yes Yes Yes Yes

Notes: Liberia in both periods and Malawi in 1998 were excluded since they wereextreme outliers. Robust standard errors in parentheses. ***p < 0.01, **p < 0.05,*p < 0.1

Table 3. Aggregated Intensive Fitness and Growth.

Dependent Variable: Growth in GDP pc1988-1998, 1998-2008

Variable (I) (II) (III) (IV) (V) (VI)Income per capita, logs −0.031∗∗ −0.032∗∗ −0.043∗∗∗ −0.044∗∗ −0.051∗∗∗ −0.040∗∗∗

(0.013) (0.013) (0.014) (0.020) (0.016) (0.015)Increase in NR exports 0.334∗∗ 0.296∗ 0.336∗∗ 0.313∗ 0.332∗ 0.340∗∗

(0.160) (0.174) (0.161) (0.172) (0.169) (0.160)Initial COI 0.003∗ 0.004∗ 0.004∗∗ 0.003∗ 0.004∗∗

(0.002) (0.002) (0.002) (0.002) (0.002)Initial IF, logs 0.002∗∗∗ 0.002∗∗∗

(0.000) (0.000)Initial aggregated IF, logs 0.001∗∗∗ 0.001∗∗∗ 0.001∗∗∗ 0.001∗∗

(0.000) (0.000) (0.000) (0.000)Initial Export of GS 0.001∗∗

(0.000)Initial Population, logs 0.008

(0.016)Constant 0.510∗∗∗ 0.550∗∗∗ 0.732∗∗∗ 0.746∗∗∗ 0.770∗∗∗ 0.564

(0.122) (0.121) (0.144) (0.144) (0.152) (0.352)Observations 210 210 210 210 202 210R2 0.208 0.239 0.236 0.253 0.238 0.237Year FE Yes Yes Yes Yes Yes Yes

Notes: For consistency with the aggregated ECI regressions, Liberia in bothperiods and Malawi in 1998 were excluded from the sample. Robust standarderrors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1

11

The log of the initial IF and initial aggregated IF have a Pearson correlation of 0.08 after controlling for theinitial income, natural resource exports and COI, thus eliminating the collinearity which appeared in the LinearMethod and signalling that these measures provide different information about the long term potential of acountry. We believe that this appears as a consequence of the nonlinearity of the method. Of greater interest tous are the results regarding the explanatory power of aggregated IF. Columns III, V and V I show that theyare very similar to the ones for aggregated ECI - the regressions have nearly identical R2 and the aggregatedIF continues to have a positive and significant relation with growth even when adding the controls and doingthe robustness check. To make a more adequate presentation of the marginal effects of both aggregated ECIand the log of the aggregated IF, we standardize their estimated coefficients by multiplying them by the ratioof the standard deviations of the independent and dependent variables [22]. The standardized coefficientsimply that, a one standard deviation increase in the aggregated ECI is associated with an increase in thegrowth variable equal to 17.6% of a standard deviation in that variable, while a one standard deviation increasein the aggregated IF increases the same growth variable by 26.3% of its standard deviation. From this, itcan be said that the aggregated complexity measures have similar effect on economic growth, and, due tothe nearly identical R2, it appears that the aggregation and the inclusion of services lessen the discrepancybetween the explanatory power of the two metrics. More importantly, the magnitude of the standardizedcoefficients is only slightly smaller than the increase in natural resource exports (which is between 25% and 27%of a standard deviation in both regressions), thus allowing us to conclude that an increase in the aggregatedeconomic complexity has an economically large effect.

Services vs. Goods

Here we investigate the dynamics of the (aggregated) linear and nonlinear Product Complexity Index, estimatedthrough the unweighted representation of the bipartite network. We exclude the weighted Product ComplexityIndex, since it is less associated with the productive capabilities embedded in the goods or services [11].

We begin our analysis by presenting a visualization of the aggregated product complexity ranking in Fig 2.It can be easily noticed that most services have higher complexity than goods. This result can be statisticallysupported by regressing the aggregated PCI against a dummy variable that captures the effect of services andfixing the time effects

pjt = µg + ∆µsDs + ηt + εjt

where pjt is either the linear or the nonlinear product complexity index for good or service j at time t, µg isthe average good complexity, ∆µs is an estimate of the additional average service complexity; Ds is the dummyvariable; ηt the time effects and εjt is the error component.

Table 4 gives the estimated results. From column I it can be concluded that for the linear method, aftercontrolling for the year, on average the service complexity is higher than the good complexity by 1.58 standarddeviations. The nonlinear method produces similar results, as presented in column II – the average servicecomplexity is 1.77 times the average product complexity, whilst the average good complexity is only 0.03 timesthe average product complexity.

But why does producing services require so much more capabilities?As Hausmann and Hidalgo [2] argue, the production of a certain good requires the country to have the

necessary capabilities associated with it. In other words, a link between a country and a product (be it a goodor a service) exists if the country has the capabilities required to produce it. The framework which relatesproducts and capabilities may be used to qualitatively describe complexity of services. In particular, by lookingat the bottom of the product complexity rankings we find that SITC 3: Mineral Fuels, Lubricants and relatedmaterials is the least complex product in almost all years, and in both rankings (only in 2004 the NLM ranksSITC 3 higher). Disaggregating it, we can get various goods based on coal, petroleum, gas, or electrical power.All of these represent natural resources whose production does not rely as much on the presence of capabilities,as it does on the geographical luck of the country.

By moving further up in the rankings we observe that in the majority of the examination period SITC 7:Machinery and transport equipment represents the most complex aggregated good. For a country to be able toproduce machineries such as combustion engines or x-ray equipment, besides using natural resources as inputs,it has to combine other capabilities, for example the capabilities embedded in its workers. The positioning of

12

the linear aggregated products is well reflected in the SITC rev.2 4-digit ranking provided in the Observatoryof Economic Complexity [16].

At the very top we can find services such as EBOPS 260: Financial Services. We believe that, in the senseof [2], the production of these services requires the combination of the complex knowledge embedded in thegoods and in the other services that a country makes. In addition, we argue that as a country’s product spacebecomes saturated with goods, the contribution of economic complexity to future economic growth begins todecrease, which is in line with the observations in [4]. In this context, our results suggest that probably theeasiest way for maintaining the effect of the evolution of the productive structure is by introducing serviceswhose set of required capabilities is similar, but a little more complex; for instance, the production of modernfinancial services requires the presence of sophisticated machineries and an additional input of capabilities. Itis worth noting that our argument is in accordance with the conclusions in [23] where it was indicated thatservices tend to grow in importance for the economy as the level of a country’s development rises.

Fig 2. Evolution of services and goods complexity (1995-2010). a) Product ranking estimatedthrough LM. b)Same as a) only estimated through NLM. a-b Services: blue, goods: red.

13

Table 4. PCI Regressions.

Dependent Variable: Product Complexity Index1995-2010

Variable (I) (II)Service dummy 1.584∗∗∗ 1.773∗∗∗

(0.225) (0.676)Constant −0.864∗∗∗ 0.033∗∗∗

(0.203) (0.009)Observations 352 352R2 0.652 0.086Year FE Yes Yes

Note: Robust Standard errors in parentheses. ***p < 0.01

Country Dynamics

In order to assess the effects of goods aggregation and inclusion of services on the productive structure of thecountries, we perform a static and dynamic comparison of the country rankings based on aggregated (withgoods and services) and disaggregated (goods) data.

Static Comparison: For the static comparison in Fig 3 we plot the SITC 4-digit level disaggregated versionsof ECI, IF and EF against their aggregated counterparts for the year 2005. First, it can be observed that thereis a very strong and positive Spearman’s rank correlation between both EFs. This is due to the nature of theweighted index – it is strongly connected to the size of the economy. Moreover, Fig 3a-b shows that the positiverank correlation remains even for both ECIs and IFs, although their strength falls. In fact, the aggregatedand disaggregated ECI have a Spearman correlation of 0.76. We argue that this discrepancy appears becausethe complexity of a country is proportional to the sum of the complexities of its exporting product and, sinceservices are more complex, the countries that are more diverse in producing services are supposed to be rankedhigher in these aggregated rankings.

Dynamic comparison based on the unweighted representation: To provide a more detailed analysisof the country rankings, in Fig. 4 we present the dynamics of the aggregated and disaggregated ECI and IF, andhighlight several countries(China, Estonia, Ethiopia, Malaysia, Switzerland and United Arab Emirates(UAE),United Kingdom (UK) and the United States (USA)) with different economic attributes. Here countries areranked according to the change in ECI (Fig 4a), aggregated ECI (Fig 4b, IF (Fig 4c) and aggregated IF (Fig4d) experienced between 1995 and 2010. Because of data availability, the ranking is limited to 94 countries.

Fig 3. Scatterplots of the aggregated and disaggregated versions of ECI, IF and EF (2005). a)Plots ECI against aggregated ECI.b)Plots IF against aggregated EF. c) Plots IF against aggregated IF.

14

From the figure several conclusions can be drawn. First, it can be noticed that there are by far more changesin the aggregated rankings than in the disaggregated ones. We believe that the aggregation of the data andthe nature of the services are behind this observation. Although the calculation of the indices does not accountfor the volume of trade, it still accounts for its presence. Since the aggregated ECI and IF rely on aggregateddata (effectively decreasing the number of products-i.e. the dimension of the adjacency matrix), it can be saidthat, the disappearance or appearance of a link in the bipartite country-product space has a far bigger impacton the complexity of the country than in the case with disaggregated goods data. However, after 2007 there iseven larger volatility in the rankings, which cannot be attributed solely to the aggregation. An explanation forthe occurrence of this aberration we find in the emergence of the Global Financial Crisis. As discovered in [24]for USA, the recent reduction in trade relative to overall economic activity is far larger than most changesexperienced in the past.

Based on these results we argue that services are most probably more susceptible to economic crises (ascompared to goods), resulting in more frequent but also bigger changes in the rankings. One interpretation ofthis phenomenon would be that in light of financial crises, the financial services (which are usually the mostcomplex services according to our model) are subject to a major decrease in trade, thus significantly affectingcomplexity indices and country rankings. Additionally, it may be that austerity measures, which are sometimesassociated with financial crises, affect service exports different than goods exports. However, additional effectsmight take place. For example, in regard to [25], it may be argued that the production of a service requires amore complex combination of capabilities as a means to be produced, and a country which adds a service to itsexport basket is at a greater risk of failure (in terms of delivering the service) due to the lower probability thatall of the required tasks to construct the service are going to be performed correctly. Nevertheless, as we willsee later from our results, countries with developed service sector seem to be relatively robust on long termand succeed to exit the crises relatively quickly.

Second, clearly all rankings discriminate between developed and developing countries. For example, thereis an obvious distinction between Switzerland and Ethiopia. Switzerland is a country that has a highlydiversified trade structure, is developed and is one of the wealthiest nations. Although it starts fairly low in theaggregated rankings, the country grows to become one of the most complex economies after 1999, matching thedisaggregated ECI rankings and overcoming the IF rankings. Ethiopia represents the other side of the story; allindices rank this country as one with the lowest capabilities over the years. Obviously the capabilities structureof Ethiopia reveals that it does not have a diversified trade structure and it is one of the poorest and leastdeveloped countries. Nevertheless, there are some countries that are wealthy and have witnessed tremendousdevelopment over the years, but do not have a high quantity of productive capabilities. For instance, UnitedArab Emirates is ranked in the bottom half of all unweighted economic complexity rankings. The wealthdiscrepancy is explained through the export structure of the country. When we look at it, we can see that UAEmostly exports natural resource based products which production mostly depends on the geographical positionof country. It is said that the countries who export those products are ”geologically lucky” since they are richnot because of their level of complexity, but because of the abundance of natural resources. The aggregatedECI and IF rank UAE even lower in some years as the revealed comparative advantage that UAE had in somecomplex products is lost by the aggregation of the data.

Most importantly, it can be seen that those countries whose GDP structure is dominated by services(See [26]) are usually ranked higher in the aggregated rankings. These are mostly developed nations like UKand USA. Both aggregated complexity indices almost always rank these two countries among the top 5 (Onlyin the last two years of examination their positioning drops in the aggregated ECI rankings), yet they arenot positioned as high the disaggregated rankings, especially UK. On the other hand, countries that have agoods based economy are normally ranked lower in the aggregated rankings. We can undoubtedly say thatChina has one of the most developed goods industries and has amazed an immense economic growth over theyears. Nonetheless, both aggregated rankings rank the country fairly low just because the service sector is notas developed. The distinction between the rankings is even more pronounced when examining countries likeEstonia and Malaysia which have similar wealth and disaggregated rankings through the years, but differentGDP structure. Estonia’s economy relies heavily on services, whereas Malaysia’s is almost equally reliant onboth goods and services. Therefore, Estonia is usually ranked far higher in the aggregated rankings, and hasexperienced bigger rises and falls over the years. For comparison, at the same time Estonia averaged a bigger

15

Fig 4. Unweighted Country Dynamics (1995-2010). a) ECI rankings movement of the countries. In itwe highlight China, Estonia, Ethiopia, Malaysia, Switzerland, the United Arab Emirates, United Kingdom andthe United States. b) Same for aggregated ECI. c) IF rankings movement of the countries. d) Same foraggregated IF. a-d Three digit iso codes are used as country abbreviations.

yearly GDP per capita growth (4.99%, whilst Malaysia had 2.65%) and had bigger standard deviation in it(7.07 to 4.34).

Weighted Dynamic Comparison: In the Materials and Methods section we interpreted the measuresconstructed through the weighted version of the network as metrics who ignore the size of an economy. Instead,they emphasize the world export share of a product that the country possesses for evaluation of its short-termcompetitiveness [17]. For that reason, both the disaggregated and aggregated EF dynamics, pictured in Fig 5,usually put large economies on a higher position. Furthermore, distinctively from the aggregated ECI andIF rankings, the aggregated EF rankings have relatively stable movements over time and are more similar tothe disaggregated ones. We argue that the small differences that appear can help us extract new informationabout the short-term situation of the nations.

As in the unweighted country dynamics, the complexity of service export oriented economies is increasedin the aggregated Extensive Fitness; USA and the United Kingdom are always the two most competitiveeconomies, and Switzerland is generally ranked higher. Yet, their different positioning is not as noteworthy asin the unweighted analysis. Initially, this can be concluded from the ranking dynamics of China, who is the

16

Fig 5. Weighted Country Dynamics (1995-2010) a) EF rankings movement. In it we highlight China,Estonia, Ethiopia, Malaysia, Switzerland, the United Arab Emirates, United Kingdom and the United States.b) Same for aggregated EF. a-b Three digit iso codes are used as country abbreviations.

most competitive economy according to the disaggregated EF rankings: in the aggregated rankings China’scomplexity is lower, but it steadily grows and even surpasses Switzerland’s in 2010. The contrast betweenthe rankings can be explained in more detail by again discussing the positions of Estonia and Malaysia. Asexpected, in both EF rankings Malaysia is almost always ranked higher than Estonia because it has a largereconomy. However, Malaysia is not as influential country in the global market dynamics in the service sector asEstonia, and consequently there is a smaller range between the both countries in the aggregated EF rankings.This range between Estonia and Malaysia is even smaller when compared to the one observed in Fig 4.

In summary, the aggregated rankings reveal that the unweighted complexity of a country is dominated byits ability to produce services. Countries whose export basket includes many different types of services havebigger long-term growth potential. Conversely, the aggregated EF dynamic analysis shows that on the shortterm, the complexity of an economy does not rely as much on the production of services as it relies on theinvolvement of the country on the global market of both goods and services.

17

Country Specifics

Countries with developed service sector: As observed from the presented results, the inclusion of servicesaffects significantly the complexity indices and the rankings of individual countries. Specifically, the fact thatthe services appear to be on average more complex than goods (i.e. rank higher on the product complexityscale), positions (in general) countries with developed service sector higher in the economic complexity ranking.This holds in particular for countries such as USA and UK. The aggregated extensive dynamics in each yearranks these countries as the two most competitive economies on the short term which is not the case withthe disaggregated EF. In addition, until 2008 UK is ranked higher in both aggregated unweighted rankings(aggregated ECI/aggregated IF), when compared to the disaggregated rankings with only goods, whereasUSA is also ranked higher in the aggregated ECI rankings and keeps the same positioning in the aggregatedIF rankings (always among the 3 most complex economies). After 2008 there is a drop in their aggregatedECI rankings which is not reflected in the aggregated IF rankings. Hence, it may appear that due to theaveraging process of the Linear Method, the overall structure of the aggregated ECI rankings is more affectedby the financial crisis. The differences between the complexity dynamics based on aggregated goods andservices and the complexity dynamics based only on (disaggregated) goods of UK and USA are even morepronounced when we compare our aggregated complexity metrics (which account for both goods and services)to the ones calculated by only using the SITC Rev.2 1-digit aggregated goods. Therefore, we argue that byincluding the services, the aggregated metrics presents a more realistic picture about the complexity of thesetwo countries. We base our argument on the fact that USA and UK were among the first countries to experiencethe post-industrial transition from a manufacturing-based economy to a service-based economy during the lastfew decades of the twentieth century [27]. This transition, characterized with a declining manufacturing sector,yields an effective decrease of the complexity of a country, when estimated only by using goods data.

While the same conclusions hold for some of the developed countries, as in the case of Switzerland forexample, they appear to not be true for Germany, Japan and Italy. In fact, their rankings are reversed: in therankings based solely on goods they are among the most complex economies, whereas they are placed on averagelower in the rankings where services are included. We believe that this discrepancy is due to the difference inthe speed of transition. It is known that over the years, Germany, Japan and Italy had the lowest increase inservice employment among the developed countries [28], which implies that they preserved the diversity inthe goods sector and did not develop as diversified service sector as other developed countries. One one side,it may be the case that economic complexity indices based only on goods are slightly biased towards thesetype of economies. This could especially be true for Italy, a country which exhibits great discrepancy in therankings: in the aggregated country rankings it is always placed outside the top ten, meaning that it definitelydoes not have as competitive service sector as other economies, and yet, according to the disaggregated IF, isconstantly among the five most complex countries. However, on the other side, it may also be that aggregatedrankings which include services are biased towards the economies with sophisticated service sector, as it isthe example with the US and, in particular, with UK. More precise insights would definitely be obtained byperforming an analysis based on finely disaggregated goods and service data, something which is not possibledue to current service nomenclature and data availability.

The PIIGS nations: PIIGS is a derogatory term which refers to the nations of: Portugal, Ireland, Italy,Greece and Spain; five European Union member states that were unable to refinance their government debt orto bail out over-indebted banks on their own during the sovereign debt crisis of 2009. Their downturn wasdiscussed from an economic complexity perspective in [11]. According to the authors, the fragility of the PIIGSis mostly of financial origin and is less related to their productive structure. They strengthen their argumentby pointing out that there are different regimes in the fitness-income plane, and that the PIIGS are in the onewhere the developed nations are stationed. The economies in this regime are less reliant on their productivestructure for future growth since they already have a diversified goods export basket; i.e. they have saturatedtheir product space. While we agree with this argument, we argue that the fragility of the PIIGS countriesmay be, to a certain extent, attributed to the structure of their service sector. Therefore, in the following weconcentrate on the economic complexity of these countries where services are added in the model.

Fig 6 illustrates the disaggregated and aggregated dynamics of the PIIGS countries. Even though theaggregated ECI and IF dynamics (Fig 6b and Fig 6d) moderately match their disaggregated counterparts (Fig

18

6a and Fig 6c) with the partially stable (ranking) movements of Italy and Spain, there are some significantdifferences in the movements of the other countries. For example, in the aggregated rankings Portugal exhibitsa steady decline, whereas Greece has very big fluctuations which lead to a dramatic drop in its complexity inthe period of 2009-10. Additionally in contrast to [11], our aggregated rankings reveal that Ireland has greatlytransformed its productive structure and is presently one of the most complex economies. We believe thatthese dynamics explain better the long term economic expectations of the PIIGS nations than the projectionsbased solely on goods data. As a matter of fact, Ireland was the first country to exit its bailout program bythe end of 2013 [29], and in the following years experienced significant economic growth, thus implying thatthe Irish economic downturn was more short to medium termed, and is better explained by the EF rankingsdepicted in Fig 6e-f. On the other hand, the economies of Greece and Portugal did not experience the sameimprovements and, as of 2016, are yet to make their full recovery; Greece was even reclassified as an emergingmarket by some equity index providers [30].

It should be emphasized that we do not dispute the fact that the PIIGS crisis may be attributed, at least inpart, to the structural problems in the countries’ economic systems and the inflexibility of their policies. Inaddition, as the authors in [8] argue, different regimes may exist for the economic complexity leading to thethere introduced Selective Predictability model. Instead, we provide an additional explanation for the recenteconomic situation in these countries. In particular, we address the re-emergence of Ireland’s economy and theongoing stagnation of Greece and Portugal. Unlike Ireland, who over the years managed to expand its set ofcapabilities and diversify its product space to other more complex services, Greece and Portugal maintainedtheir service productivity only in industries such as shipping and tourism. These services, respectively includedunder EBOPS 205: Transportation and EBOPS 236: Travel in our aggregated classification, rank lower on theproduct complexity scale, and are shown to be especially sensitive to financial crises.

Conclusion

In this paper we extended the relatively new branch of Economic Complexity to include services. By includingservice data and aggregating goods data (to make it comparable to the service data), we introduced aggregatedindices as measures for the complexity of products and countries. We thereby showed that, although someinformation is inevitably lost in the process of aggregation, the inclusion of services in the model provides newinsights for countries and products. Importantly, the statistically significant relationship between the complexitymeasures and the economic growth remains even with the aggregated product nomenclature (including goodsand services). Nonetheless, these aggregated metrics are not as good predictors of growth as their disaggregatedversions for, although we add service information, through this setup we extract less information about theproductive capabilities embedded in the goods. Regardless of that, the aggregation allows us to correctly graspthe differences between the complexity of goods and services. We discovered that services require far morecapabilities as a means to be produced, as compared to goods.

Next, we investigated the impact that the inclusion of services has on the overall structure of the countryrankings. We concluded that adding the services exports increases the complexity of economies with developedservice sector when they are estimated based on the model with binary (unweighted) entries of the adjacencymatrix describing the country-product network (unweighted representation). Following standard interpretationof the unweighted representation which relates economic complexity with the potential for future growth,one may conjecture that the economies whose export basket is based on (sophisticated) services have biggerlong-term growth potential as opposed to economies centred on export of goods.

However, we also found out that the newly obtained aggregated country dynamics experience bigger changesover the years when compared to the disaggregated ones. While this effect may be attributed in part to theaggregation of goods data, the large volatility of the rankings of the service-based economies following theFinancial Crisis of 2007-08 suggests that this effect is also due to the inclusion of services in the model. Asresult, it may be argued that service exports are more susceptible to (financial) economic crises (as comparedto goods), resulting in more frequent but also bigger changes in the rankings. One interpretation of thisphenomenon would be that in light of financial crises, sophisticated services (such as financial, which are usuallythe most complex services according to our model) are subject to a major decrease in trade, thus significantlyaffecting complexity indices and country rankings. In addition, and in regard to [25], it may be argued that

19

Fig 6. Dynamics of the PIIGS nations (1995-2010). a) ECI rankings movement b) Same foraggregated ECI. c) IF rankings movement d) Same for aggregated IF. e) EF rankings movement d) Same foraggregated EF. a-f Three digit iso codes are used as country abbreviations.

20

the production of a service requires a more complex combination of capabilities as a means to be produced,which increases the risk that of failure in the delivery of the service. Nevertheless, the dynamics suggest thatcountries with sophisticated service sector seem to be relatively robust on long term and succeed to exit thecrisis relatively quickly. This is true for example for the US economy, and in particular, for Ireland which wasthe first country to exit its bailout program by the end of 2013, and experienced a significant economic growthafterwards.

Similar conclusions hold for the aggregated country dynamics derived on basis of the model where theadjacency matrix is not binary (i.e. has weighted entries), although to a lesser extent. As result of theintroduced weights, there is a lower volatility in the aggregated dynamics and the difference between them andthe disaggregated dynamics are less pronounced. This yields the conclusion that Extensive Fitness developedon basis of the weighted representation may be an appropriate measure for the economic complexity of acountry in a model based on aggregates goods and service data (after accounting carefully for the fact that itputs accent on the size of the economy.)

Lastly, by producing a country specific analysis, we argued that the model with services yields economiccomplexity measures which, for some countries, are more consistent with their ”perceived” economic situation,as otherwise suggested by an analysis performed solely on disaggregated exported goods. Examples include thecountries with developed service sectors (US, UK and Switzerland) and, in particular, the PIIGS countrieswhere these metrics provide additional information in light of their sovereign debt crisis.

A more detailed analysis is certainly needed to further uncover the effects of services on the complexitystructure of the countries. This, however, requires a highly accurate and disaggregated service data which iscurrently not available. Another fruitful research direction would be to fully exploit the concept which relatesproducts (goods and services) with the capabilities required for their production. Building an explanatorymodel for the underlying capabilities would bring novel insights about the relations between products in theproduct space and, importantly, about the productive knowledge embedded in a country’s economy. This is asubject of our current work.

References

1. Hidalgo CA, Klinger B, Barabasi AL, Hausmann R. The product space conditions the development ofnations. Science. 2007 Jul 27;317(5837):482-7. doi: 10.1126/science.1144581

2. Hausmann R, Hidalgo CA. The network structure of economic output. Journal of Economic Growth.2011 Dec 1;16(4):309-42.

3. Hidalgo CA, Hausmann R. The building blocks of economic complexity. proceedings of the nationalacademy of sciences. 2009 Jun 30;106(26):10570-5. doi:10.1073/pnas.0900943106

4. Hausmann R, Hidalgo CA, Bustos S, Coscia M, Simoes A, Yildirim MA. The atlas of economic complexity:Mapping paths to prosperity. Mit Press; 2014 Jan 17.

5. Bustos S, Gomez C, Hausmann R, Hidalgo CA. The dynamics of nestedness predicts the evolution ofindustrial ecosystems. PloS one. 2012 Nov 19;7(11):e49393.

6. Frenken K, Van Oort F, Verburg T. Related variety, unrelated variety and regional economic growth.Regional studies. 2007 Jul 1;41(5):685-97.

7. Saviotti PP, Frenken K. Export variety and the economic performance of countries. Journal of EvolutionaryEconomics. 2008 Apr 1;18(2):201-18.

8. Cristelli M, Tacchella A, Pietronero L. The heterogeneous dynamics of economic complexity. PloS one.2015 Feb 11;10(2):e0117174.

9. Balassa B. Trade liberalisation and “revealed” comparative advantage. The Manchester School. 1965May 1;33(2):99-123.

21

10. Tacchella A, Cristelli M, Caldarelli G, Gabrielli A, Pietronero L. A new metrics for countries’ fitnessand products’ complexity. Scientific reports. 2012 Oct 10;2.

11. Cristelli M, Gabrielli A, Tacchella A, Caldarelli G, Pietronero L. Measuring the intangibles: Ametrics for the economic complexity of countries and products. PloS one. 2013 Aug 5;8(8):e70726.doi:10.1371/journal.pone.0070726

12. Mishra S, Lundstrom S, Anand R. Service export sophistication and economic growth. World BankPolicy Research Working Paper Series, Vol. 2011 Mar 1.

13. Francois J, Pindyuk O. Consolidated Data on International Trade in Services v8. 9. IIDE Discussion Pa-per 20130101. 2013 Jan. Available: http://data.worldbank.org/data-catalog/trade-in-service.Accessed 8 September 2015.

14. Feenstra RC, Lipsey RE, Deng H, Ma AC, Mo H. World trade flows: 1962-2000. National Bureau ofEconomic Research; 2005 Jan 17. Available: http://cid.econ.ucdavis.edu/nberus.html. Accessed 8September 2015.

15. UN COMTRADE, International Trade Statistics Database. Available: comtrade.un.org. Accessed 8September 2015.

16. Simoes AJ, Hidalgo CA. The Economic Complexity Observatory: An analytical tool for understandingthe dynamics of economic development. Workshops at the Twenty-Fifth AAAI Conference on ArtificialIntelligence 2011. Available: https://atlas.media.mit.edu/about/team/

17. Caldarelli G, Cristelli M, Gabrielli A, Pietronero L, Scala A, Tacchella A. A network analysis of countries’export flows: firm grounds for the building blocks of the economy. PloS one. 2012 Oct 19;7(10):e47278.doi:10.1371/journal.pone.0047278

18. Pugliese E, Zaccaria A, Pietronero L. On the convergence of the Fitness-Complexity Algorithm. Preprint.Available: arXiv:1410.0249. Accessed 15 January 2016.

19. Costanza R, Hart M, Talberth J, Posner S. Beyond GDP: The need for new measures of progress.

20. Costanza R, Kubiszewski I, Giovannini E, Lovins H, McGlade J, et al. Development: Time to leave GDPbehind. Nature. 2014; 505: 283–285. pmid:24436983 doi: 10.1038/505283a

21. Sachs JD, Warner AM. Natural resource abundance and economic growth. National Bureau of EconomicResearch; 1995 Dec 1.

22. Herzer D, Vollmer S. Inequality and growth: evidence from panel cointegration. The Journal of EconomicInequality. 2012 Dec 1;10(4):489-503.

23. Saez S, Taglioni D, Van der Marel E, Hollweg CH, Zavacka V. Valuing services in trade: a toolkit forcompetitiveness diagnostics. World Bank Publications; 2015 Jan 15.

24. Levchenko AA, Lewis LT, Tesar LL. The collapse of international trade during the 2008–09 crisis: insearch of the smoking gun. IMF Economic review. 2010 Dec 1;58(2):214-53.

25. Ferrarini B, Scaramozzino P. Complexity, Specialization, and Growth. Asian Development Bank Eco-nomics Working Paper Series. 2013 Mar 1(344).

26. The World Factbook. 2011 Nov 26. Central Intelligence Agency [Internet]. Available:https://www.cia.gov/library/publications/the-world-factbook/fields/2012.html

27. Bell D. The Coming of Post-industrial Society. A Venture in Social Forecasting. With a New Introd. bythe Author. Basic Books Incorporated; 1976.

22

28. Castells M. The rise of the network society: The information age: Economy, society, and culture. JohnWiley & Sons; 2011 Aug 24.

29. Ray M. Eurozone Debt Crisis. 2015 Jul 17. [Cited 16 February 2016]. In: Encyclopedia Britannica[Internet]. Available: http://www.britannica.com/topic/euro-zone-debt-crisis/.

30. MSCI reclassifies Greece to emerging country status. Reuters. 11 Jun 2013 Available:http://www.reuters.com/article/msci-equity-reclassifications-idUSL2N0EN20K20130611. Accessed 10February 2016.

23