Portal de Periódicos da CAPES

Beyond productivity slowdown: Quality, pricing and resource reallocation in regional competitiveness

2022; Elsevier BV; Volume: 101; Issue: 6 Linguagem: Inglês

10.1111/pirs.12696

1435-5957

Roberto Camagni, Roberta Capello, Giovanni Perucca,

Regional resilience and development

Labour productivity change at constant prices is the mainstream indicator of regional competitiveness. However, it hides and overlooks some relevant sources of competitiveness that may partly explain the Solow paradox. First, it mixes productivity improvements from technological progress with those from relocating activities to more productive sectors. Second, it partially overlooks novelties and qualities embedded in new products and the effects of market power. This paper proposes a methodology to disentangle the different effects and to apply it to recent development of European regions. Results highlight the highly heterogeneous competitive strategies of regions, and the persistent discrepancy between Eastern and Western ones. La variación de la productividad laboral a precios constantes es el principal indicador de la competitividad regional. Sin embargo, oculta y pasa por alto algunas fuentes relevantes de competitividad que pueden explicar en parte la paradoja de Solow. En primer lugar, mezcla las mejoras de productividad derivadas del progreso tecnológico con las derivadas de la deslocalización de actividades hacia sectores más productivos. En segundo lugar, pasa por alto en parte las novedades y las cualidades incorporadas a los nuevos productos y los efectos del poder del mercado. Este artículo propone una metodología para desentrañar los diferentes efectos y aplicarla al desarrollo reciente de las regiones europeas. Los resultados ponen de manifiesto la gran heterogeneidad de las estrategias competitivas de las regiones y la persistente discrepancia entre las orientales y las occidentales. 不変価格での労働生産性の変化は、地域の競争力の主要な指標である。しかし、それはソローのパラドックスを部分的に説明する可能性がある競争力の重要な源泉を隠し、看過するものである。不変価格での労働生産性の変化では、第一に、技術進歩による生産性の向上と、より生産性の高い部門に作業工程を移管することによる生産性の向上が混在しており、第二に、新製品に内在する斬新さや品質、市場支配力の影響が部分的に見落とされている。本稿では、異なる効果を分離し、ヨーロッパ地域における最近の発展に適用する方法論を提案する。結果から、地域の競争戦略の異質性が高いことと、東部における戦略と西部における戦略の相違が持続していることが強調される。 More than thirty years ago, Robert Solow conceptualized for the first time what is known now as the "productivity paradox" (Solow, 1987). This idea refers to the puzzling evidence according to which labour productivity in the United States slowed down from the 1970s on, in spite of the rapid technological change and widespread diffusion of information technology (IT). Some decades after, the paradox is still far from being resolved. Rather, the further progress of automation technologies and the outset of unprecedented possibilities of integration between artificial and human intelligence did not match a parallel increase in productivity (Brynjolfsson & McAfee, 2014). While the debate on the paradox originated in the United States, it fully applies to the European Union (EU) where, in the last fifty years, productivity grew at an average rate even lower than the one of other advanced world economies. In EU15 countries, labour productivity growth declined from an average yearly rate of 2.4% over the period 1973–1995 to 1.5% between 1995 and 2006 (van Ark et al., 2008). More recently, after the generalized slowdown unleashed by the economic crisis, this decline continued, and labour productivity in EU15 countries grew at a modest 0.71% average yearly rate between 2013 and 2018. From a regional perspective the evidence is even more puzzling, with increasing disparities in regional labour productivity within the EU, broadened in the last two decades (Gómez-Tello et al., 2020). The literature already suggested several explanations to this finding, pointing in general to the excessive expectations about the ICTs' impact on productivity, as a consequence of the concentration of the benefits of the new technologies in a relatively small fraction of the economy; the time-lag between the rise of a new technology and its widespread and effective implementation into production processes; and finally to measurement issues (Acemoglu et al., 2014; Brynjolfsson et al., 2017; Byrne et al., 2016; Cuadrado-Roura, 2020; Kim & McCann, 2020; Syverson, 2017). The present paper enters this debate by reflecting on two aspects. The first relates to Aghion et al.'s (2019) theorem stating that the usual statistical treatment of productivity growth at constant prices overstates inflation and assigns a too limited increase in real output reached through product novelty and quality. The paper proposes to resort to a less "manipulated" indicator given by the value added at current prices, supplying information on relative price trends on different sectors, mirroring their capacity to sell at increasing prices: this pricing capacity residing in a faster product innovation, superior and increasing product quality or in other marketing and delivery capabilities in monopolistic-competitive markets. This reasoning was already suggested by some authors (Acemoglu et al., 2014; Denison, 1967). The identification of such missing element in GDP is rather crucial at the regional level. When erroneously attributed to inflation, this element hides important market powers, spatially localized (e.g. in large metro-regions), provoking heavy inter-regional redistributive effects. In fact, the relevant advantage of some areas, selling goods and services at favourable terms-of-trade in inter-regional exchanges is not captured, generating a misperception about the relative growth trends and the degree of competitiveness of the single areas. Moreover, statistical offices, at least in the EU, do not compute specific regional price deflators and use national sectoral deflators, missing the effects of inter-regional differences in product innovation capabilities, product differentiation and market power. Third, a specific specialization in highly dynamic sectors, with price/quality increases and high Schumpeterian profits would be overlooked by productivity statistics, with a major effect at the regional level, given that regions are much more specialized than countries. Fourth, different regions sell at different prices similar goods due to the presence of specific location costs, namely, land and building rents. Rents represent both production costs and incomes, and, in particular regions like metropolitan ones, when they are passed on to prices, they generate an advantage that should be highlighted in inter-regional income distribution (Camagni, 2020). For these reasons, the quality/monopolistic competitive effect on prices needs to be (partially) re-introduced in regional productivity growth analyses: the value added of this paper is to propose a measurement for this effect. Moreover, the paper suggests that aggregate productivity gains associated to technological progress contain in reality a hidden element, namely inter-sectoral "reallocation of resources" following some suggestions already present in the prewar and afterwar studies on the "sources of economic growth" (Denison, 1962, 1967; Fabricant, 1942), transposed into a formalized shift-share model on productivity growth starting from 1980 (Camagni, 1980; Camagni & Cappellin, 1985; Ledebur & Moomaw, 1983). This same modelling idea was widely replicated thereafter, and is very popular nowadays (Bernard & Jones, 1996; Caselli & Tenreyro, 2006; Enflo & Rosés, 2015; McMillan & Rodrik, 2011; Paci & Pigliaru, 1998). The distinction between pure technological progress taking place inside industry and productivity increases through sectoral shifts towards higher value added sectors reveals two completely different competitive trajectories. This is even more true at the regional level, where such a hidden element registers a high heterogeneity: it can be in fact particularly high in less developed regions, since they generally go through a rapid structural change accompanying the early phases of industrialization, with massive reallocation from agriculture to manufacturing. All the previous reflections are of paramount importance today, when we are confronted with the difficult predictions of the effects of the 4.0 technological revolution (Capello & Lenzi, 2021; Rullani & Rullani, 2018). The paper proposes methodologies to disentangle the hidden element and to identify the overlooked one (Section 2), providing the appropriate statistical formula for their measurement (Section 3). In the case of the overlooked element (the pricing effect linked to quality and/or monopolistic power), the empirical approach is an original one. In the case of the hidden element (the reallocation effect), a well-known decomposition method is applied, and the novelty resides in its application not just to the traditional increase in real output per capita, but also to the proposed pricing effect (quality/monopoly effect) with the necessary adjustments in the formula. Then, the paper applies the methodologies to European regional data, so as to build an original taxonomy of regional development patterns through the different sources of regional competitiveness. Interesting results do emerge (Section 4). Productivity growth has been the topic of an extensive literature. 1 Traditionally, productivity growth analysis aimed at measuring the increase of output produced by different geographical units (countries, regions, firms, etc.) separating the contribution of the growth of single production factors, alone and in conjunction with each other, from that of external factors, called "technological progress" (Solow, 1957). In the last decades, since 1980–1990, the analysis of productivity growth at the country and regional level pointed to the paradoxical result that, in spite of unprecedented technological developments, productivity gains were relatively low. While we are still far from the full explanation of what has been labelled as the "productivity paradox" (Solow, 1987), previous studies summarized some of the elements that contributed to this unexpected empirical finding. These elements refer to the time-lag between the emergence of new technologies and their application into production processes (David, 1990), to the relatively narrow field of application of new technologies (Triplett, 1999), to the rather poor performance of ICT-using sectors in Europe compared with other advanced economies (Dahl et al., 2011; Van Ark et al., 2008), and to the limitations of national accounting statistics (Aghion et al., 2019; Moulton, 2000). Without neglecting the importance of the abovementioned factors, this paper aims at presenting a new perspective on the issue of regional labour productivity growth, strictly related to its measurement: in particular, to the traditional measurement of increases in real output, namely the increase in value added at constant prices. As mentioned in the introductory section, the main idea in this work is that measures of labour productivity growth contain, on the one hand, hidden elements that have to be explicitly recognized, and, on the other, overlook other important sources of regional competitiveness. Given the fact that sectoral weights θ i are defined by the initial sectoral structure of the economy, the last term—which is the most appropriate to interpret technical change stricto sensu, as it applies to single productions—does not consider the effect of shifts in the sectoral structure of the economy. A redistribution of employment in more or less advanced productive sectors has an effect on the regional aggregate productivity growth. 2, 3 The result reached by Solow (1957) using an aggregate production function included the strong reallocation effects from agriculture, typical of those years. This fact can partly explain his paradoxical results of a relevant productivity increase with respect to later analyses. In a subsequent work (Solow, 1987), he turned to sectoral analyses. In this case, sectoral reallocation is by definition excluded. The effect of resource reallocation on labour productivity growth was already clearly stated and estimated at least since the works of Denison (1962, 1967): in his growth accounting exercise, he added ad hoc procedures to "squeeze down the residual" found by Solow (Nelson, 1981), isolating the effects of an "overallocation to agriculture and non-farm self-employment" (Denison, 1967, p. 332). Subsequently, some pioneering studies proposed methods for the decomposition of labour productivity growth in a shift-share context, in order to set apart the role of its different sources/processes (Camagni, 1980; Camagni & Cappellin, 1985; Ledebur & Moomaw, 1983). The first two studies proposed a shift-share analysis with three effects: the usual differential, sectoral competitive effect (nowadays generally called "within-sector productivity growth"); a MIX effect, indicating a higher (lower) initial share of sectors with a relatively high (low) productivity growth (today called "static shift effect") and a reallocation effect showing an increasing share of high and increasing productivity sectors (today called "dynamic shift effect"). In the same years, Ledebur and Moomaw (1983) proposed a slightly more complex model with the same components, disentangling, inside the reallocation effect a high concentration from an increasing concentration of high-productivity industries capturing an increasing share of national employment. In subsequent years, the method consolidated with minor adjustments (Caselli & Tenreyro, 2006; Enflo & Rosés, 2015; Paci & Pigliaru, 1998). 4 The most relevant adjustment refers to the use of absolute rates of change instead of the relative ones with respect to the aggregate (the national, the EU, etc.), typical of the shift-share analysis (Bernard & Jones, 1996; McMillan & Rodrik, 2011; OECD, 2018). The three effects maintain their meaning: with-sectors, static shift and dynamic shift effects (OECD, 2018). This absolute growth form is the one utilized in the specification of the present paper (Section 3). The reallocation phenomenon is likely to be more significant in developing and transition economies, for instance with massive reallocation from agriculture to manufacturing. Between 2000 and 2017, 8.9% of total employment in the Eastern countries that joined the EU since 2004 moved away from the agriculture towards more productive sectors. The early empirical analysis for the European Commission already quoted (Camagni & Cappellin, 1985) found—for the countries in which a sufficient sectoral breakdown of productivity was available, namely Italy and France—that a reallocation effect from agriculture and commerce towards industry in the period 1970–1978 was strong in many "follower" and catching-up regions, often stronger that the "differential" or "competitive" effect in single sectors. A similar positive reallocation effect was present also, more rarely, in some advanced regions, particularly in Ile-de-France, moving in the direction of advanced industrial and service sectors. Paci and Pigliaru (1998) showed how, between 1980 and 1990, workforce reallocation explained a large part of the convergence of the relatively less productive southern European regions towards the EU average. This effect seemed to slow down in the subsequent years, and Martino (2015) demonstrated that in the period from 1990 to 2007 the contribution of employment shifts to labour productivity growth in the EU was generally negative, with the exclusion of eastern transition regions. The weak effect of workforce sectoral change on labour productivity is consistent with the findings of Fotopoulos (2008) and Fiaschi and Lavezzi (2007), both focused on EU15 regions in the two decades after 1980. In the same vein, in their city-level analysis on UK, Martin et al. (2018) argued that the employment shifts from manufacturing to services had a negative effect on labour productivity. Economic systems do not compete just on their technical efficiency in producing a given quantity of standardized output, but also on the quality of what they produce (Moulton, 2000; Van Biesebroeck, 2003). Not just product innovation, but also horizontal differentiation represents an important way to enhance regional competitiveness, generating a localized temporary monopoly power and allowing firms to increase the price of their output, also in those sectors typically characterized by low productivity growth (Libery & Kneafsey, 1998). Many studies, for instance, showed how, without any technological improvement, the agricultural sector was able in some regions to increase the price of the local products thanks to institutional innovation such as designations of origin and other strategies of product differentiation (Macedo et al., 2020). Such price increase can be the result of an increase in quality, or a monopolistic competitive behaviour or the synergies of the two: a monopolistic competitive behaviour can in fact stimulate quality increases, as well as a novelty in the product may lead to monopolistic competitive behaviours. Independently from which of the two causes prevails, the result is that growth is mis-measured. This is what we call a "quality/monopolistic competitive effect," which needs to be (partially) reintroduced in productivity growth analyses. Our proposal concerning such measurement is the following. We consider the already analysed productivity increases at constant prices ∆ Y as encompassing normal, "business as usual" quality increases in existing products and add a second component, namely the sectoral differential increases in prices with respect to the national average inflation rate, to be intended as the expression of the "quality/monopolistic competitive effect" of new products, in line with Acemoglu et al. (2014). 5 Isolating this component allows us to identify those regions that were able to sell their output at increasing prices thanks to a monopoly power created by novelty/quality/differentiation of products appreciated by the market. If regional sectoral deflators were available, there would be a more precise and direct way to measure the "quality/monopolistic competitive" effect. Unfortunately, this is not the case, since only national sectoral deflators are available, imposing their use at regional level through the regional sectoral specialization. The consequences of this is that we can only capture "quality/monopolistic competitive" effect through the presence of a favourable mix of sectors in regions and not through a specific sectoral regional price variation. 6 Measured in this way, the "quality" effect may mirror other mechanisms leading to an increase in prices that have nothing to do with quality increases or monopoly power. In particular, it may capture increases in cost of inputs, that may characterize either general resources, largely used across sectors, like oil and natural resources, or specific intermediate inputs used only in specific economic activities. In the former case, the effect on the differential increases in prices across sectors is expected to be limited, as the shock in the market of input spreads to the whole economy. In the latter case, instead, the sectoral difference in intermediate cost increases would raise a concern in empirical analyses on specific products or well specified sectors. In our case, due to data limitations, the manufacturing sector is analysed at a rather aggregate level, and we do not register input–output exchange. Furthermore, our analysis shows that the sectors driving the regional quality /monopolistic competitive effect are mostly services, whose use of intermediate inputs is relatively small. However, as a robustness check for our assumptions explained so far, we analysed the trend in oil and gas process over the period of analysis: its negative trend guarantees that our "quality/monopolistic competitive price" index is not influenced by pervasive energy price increases (Appendix A, Table A1). Moreover, a further source of relevant price changes may be a shock on the demand-side of the economy. 7 While we are not able to empirically separate these demand-side effects from the supply-side ones, we claim that the period under analysis is long enough to avoid the facts of demand volatility and to restore demand–supply equilibrium (in absence of major shocks like the ones determined by the recent pandemic or the war in Ukraine). 8 For these reasons, we think that the difference between a sectoral price change and the national inflation rate is valuable enough to capture a pricing effect linked to quality increase or other forms of market power. The aggregate regional "quality/monopolistic competitive effect" emerges as the sum of the contributions of all sectors, weighted by their relative importance and stems from the difference in the sectoral mix in each region compared with the one of the country. The availability of regional-sector deflators would improve substantially this analysis. Nevertheless, it is a wellknown fact that the most relevant difference in performance of different regions (e.g., between a global city-region and a small-city region) resides in the sectoral mix (defined at sufficiently detailed level) due to selective attraction/development of more advanced activities (the ones that are likely to introduce innovation and quality in products) and "sorting processes" of best workers (Combes et al., 2008, 2012). Given the high diversification of territorial structures, a regional analysis is able to reveal the importance of the pricing/quality mechanism as a major factor of income production, income distribution and income transfer across space, uncovering the limitations of analyses made on pure volumes or physical measures. Overlooking the role of price changes, unequal terms-of-trade among regions and market powers—which nowadays are solidly built on continuous innovation, high efficiency and no more on direct coercion or prince's decision—means neglecting the generation of huge real effects across territories, regions and countries (Camagni, 2020). Once the "quality/monopolistic competitive effect" is measured at the sectoral level by price increases, we decompose it between a "pure quality/monopolistic competitive effect" and a "sectoral reallocation effect towards high quality/monopolistic competitive sectors," able to sell their output at increasing prices. This type of reallocation represents a highly selective strategy, which could be implemented by already developed regions whose marginal gains from purely technological advances are relatively limited. The ∆ operator denotes the percentage change in employment or productivity between t 0 and t 1 . Technological progress stricto sensu is captured by the increase of labour productivity of all sectors ( ∆ y i ) weighted by the regional sectoral share of employment at time t0 ( θ i , t 0 ), implicitly imposing no sectoral reallocation in the period of analysis. Sectoral reallocation towards more (or less) productive sectors is, in its turn, measured by the shift of the sectoral share of employment ( ∆ θ i ) assigning to it the sectoral productivity at the end of the period ( y i , t 1 ). 10 The other major source of regional competitiveness concerns the output quality improvement. As said before, an approach based on value added at sector-specific constant prices partially overlooks quality measurement, especially for that part that belongs to the introduction of new products, for which statistical offices have little reference to previous virtual prices. 11 The rationale (Section 2) is to consider the productivity increases at constant prices ∆ Y as encompassing normal, "business as usual" quality increases in existing products; to add sectoral increases in prices as a measure of the "quality/monopolistic competitive effect" stricto sensu embedded in new or strongly improved products; to remove from this pricing effect the average inflation rate, encompassing different pure monetary effects (cost inflation, demand inflation, quantitative easing monetary policies, exogenous shocks) with no logical link with quality improvement. Empirically, this regional "quality / monopolistic competitive effect" is expected to register a significant regional variance. Urbanized regions, for example, are generally able to impose higher prices, since: (i) they supply higher functions; (ii) their sectoral mix favours sectors selling at higher and increasing prices with respect to rural or small city regions; and (iii) they register higher prices because they contain high urban rents, and not merely production costs. 12 Such an expectation is empirically proved with an analysis of variance (ANOVA) witnessing a higher quality/monopolistic competitive effect in urban (capital city) regions with respect to other regions, as reported in Appendix A, Table A2. When applied in European countries, the "quality/monopolistic competitive effect" in terms of prices is found in almost all countries in tertiary sectors such as professional, technical and research activities, public administration and health, arts, entertainment and recreation, which are those in which, with a different methodology, Aghion et al. (2019) found "creative destruction," product innovation and the maximum "missing growth." As in the case of productivity growth at constant prices, also in this case the hidden reallocation effect emerges comparing the results obtained working on aggregate figures with those achieved aggregating sectoral figures. Our interest in disentangling the different sources of labour productivity growth discussed in the previous section relies on the fact that each of these dimensions and their combinations theoretically depict different modes on which competitiveness relies. The implementation of the indicators previously mentioned for the European regions allows to empirically identify in a map where the different regional competitiveness strategies take place (with the caveat just mentioned). Before presenting the results, data applied for the analysis are presented in the next sub-section. We created a database for 271 NUTS 2 regions from EU27 countries plus UK, with the exclusion of Ireland and Malta, for which data are missing. For each region, we collected information on both value added at current prices and employed persons in 11 sectors. 14 In order to disentangle the two sources of competitiveness associated to technological progress (Equation (3) we applied country- and sectoral-specific deflators provided by the EUKLEMS data repository (Stehrer et al., 2019). These deflators allow tracking over time the change in the price of output in the 11 sectors and for each of the 26 countries included in our analysis. The same source also provides the overall county-level deflator, used to calculate ∆ y * in Equation (9). Our analysis focuses on the period 2013–2017, and we chose 2013 as base year for the deflators. Years antecedent to 2013 were not considered since the economic crisis of 2007 significantly affected our variables of interest, namely, value added and occupation (Mazzola & Pizzuto, 2020). Since our goal consists in identifying regional competitiveness strategies, we aim at minimizing the potential influence of exogenous, macroeconomic factors. However, to verify that the results obtained are not influenced by the choice of the period, the indices of quality/monopolistic competitive effect and technological development have been calculated for the period 2008–2017. A strongly positive and significant correlation exists between the indicators calculated in the two different periods (p-value < 0.001), suggesting that the same results would have been achieved in a different period of time (more details are reported in Table C1, Appendix C). Regional competitiveness patterns have been detected by grouping regions with the following methodology. 15 European regions were first classified according to the simultaneous presence of positive and/or negative values of the two main sources of competitiveness, using the two indicators of overall technological progress and "quality/monopolistic competitive effect" (i.e., ∆ Y and ∆ Y * ). Through this first step, we were able to identify a first group of 28 regions characterized by a negative value in both indicators. The remaining regions were split according to the two indicators of reallocation, respectively towards higher productivity sectors and towards higher price-increase sectors (mirroring higher quality and/or monopolistic competition). The choice of these two indicators is explained by the relatively small value taken by these two indicators, confining them in ne,gligible roles when analysed and interpreted jointly with the other indicators (i.e., technological progress stricto sensu and "quality/monopolistic competition" stricto sensu). Table 1 reports the number of observations included in each of the possible four general groups. These four groups, jointly with the fifth one previously identified, are then compared in terms of productivity growth sources, in order to highlight the possible sources of regional competitiveness, and identify the prevailing competitiveness pattern. Table 2 reports the results of this analysis. It shows the distribution of the four indicators capturing the different sources of competitiveness (i.e., the rows of the table) across the five groups of regions (i.e. the columns of the table) in terms of mean values and their statistical difference with respect to the whole regional sample, whose (t-test) significance is reported with an asterisk. The sources of regional competitiveness characterizing each pattern are highlighted in bold. Starting from the left of Table 2, the first group is characterized by a pure effect of technological progress stricto sensu, independent from any kind of sectoral reallocation (indicated with a - in Table 2). The second group of regions (second column) is characterized by a significantly higher than average reallocation to higher value added sectors, accompanied by a positive, significant but weak pricing reallocation (+ in Tabl