The effects of export and R&D strategies on firms’ markups in downturns: The Spanish case

ABSTRACT

The Spanish economy was one of those most hit by the Great Recession in the euro area. It suffered a huge decrease in gross domestic product (GDP), affecting especially internal demand, and in business and enterprises’ research and development (R&D) expenditures, but experienced an important increase in exports as regard to the precrisis years (the so-called Spanish “miracle”). The incorporation of Spanish small and medium enterprises (SMEs) into exports has been spectacular since 2008. Further, this has coincided with a drop in markups (stronger for SMEs). Our main objective is uncovering whether SMEs’ export and R&D participation strategies have aided in offsetting their fall in markups. We found that exporting helps SMEs to balance the decrease in markups, especially in downturns, and increases the likelihood to continue in operation. To the contrary, financial constraints and recessive demand have a negative impact on SMEs’ continuing prospects. Finally, we found evidence that SMEs’ export participation was a response to the fall in domestic demand (the “venting out” hypothesis).

KEYWORDS:

• Export
• research and development
• markups

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ We compare Spain with Germany, as it is the largest economy in the European Union and traditionally considered an export champion, and with the euro area is the natural area of reference for Spain. Spain shares with the other countries in the area, among others, a common currency, monetary policy, and trade policy. The data on GDP and exports have been extracted from the AMECO database, European Commission (https://ec.europa.eu/info/business-economy-euro/indicators-statistics/economic-databases/macro-economic-database-ameco/ameco-database_en). The data on BERD expenditure correspond to the series of Intramural R&D expenditures of business and enterprises from EUROSTAT (https://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=rd_e_gerdtot&lang=en).

² It is important noting that many of the studies report the perceptions and actions of surviving firms, and therefore might suffer from a survivor bias (Kitching et al., 2009). This points to the need to consider all firms actions results conditional on surviving.

³ There are some theoretical and empirical papers supporting the joint analysis of R&D and exports. See Bustos (2011), Constantini and Melitz (2008), Atkeson and Burstein (2010), Long, Raff, and Stähler (2011), and Máñez-Castillejo et al. (2015), among others.

⁴ This spectacular growth in exports has been coined as the “Spanish miracle” (Eppinger, Meythaler, Sindlinger, & Smolka, 2017).

⁵ As in Almunia et al. (2018), internal demand is calculated as final consumption expenditure of households and nonserving institutions serving households plus investment plus acquisition of public administrations minus imports. We extract the data from the AMECO database.

⁶ Almunia et al. (2018) explore the possible link between domestic slump and export growth, and recognize that it is difficult to accommodate it in the benchmark of the modern trade models “a la Melitz” (Melitz, 2003), in which firms’ export participation is dependent on productivity and firms’ domestic sales and exports can be analyzed independently. De Lucio et al. (2019) argue that the observed evidence in export behavior could also be compatible with the capacity-constrained exporter model. This model predicts that in the event of a fall in demand, the resulting excess of capacity reduces the cost of producing goods for international markets, and consequently more firms find it profitable to export.

⁷ According to the opportunity cost theory (Aghion & Saint-Paul, 1998; Hall, 1991), it might be optimal for firms to invest in R&D activities in recessive periods since their opportunity cost in downturns will be at its lowest.

⁸ See Carbó (2009) for an analysis of the Spanish credit crunch.

⁹ The data on export participation and the percentage of firms undertaking R&D have been obtained from the ESEE. This is the dataset we will use in our empirical section. See the data section for a more detailed description of this database.

¹⁰ De Lucio et al. (2019), using data on the universe of merchandise exporters, also found an important increase of the number of exporters in Spain during the period 2008–2013.

¹¹ We calculate markups using the methodology proposed by De Loecker and Warzynski (2012).

¹² For Spain, Moreno Martín and Rodríguez Rodríguez (2004, 2010) using a different methodology and data for the 1990s conclude that exporters enjoy higher markups than nonexporters.

¹³ To get more detailed information on the dataset we use and how to access to it, please check this website: https://www.fundacionsepi.es/investigacion/esee/en/spresentacion.asp.

¹⁴ To measure capital, we use capital stock in real terms constructed using the perpetual inventory method and based on current replacement value net of depreciation and adjusted by capacity utilization. Capacity utilization refers to the percentage of the firm’s capacity utilized.

¹⁵ Output and materials in real terms are computed using firm-specific price indexes for output and materials, respectively. The price indexes, of the Paasche-type, are constructed starting from the percentage price changes in output and the percentage price changes in materials reported by the firm in the ESEE. Capital in real terms is obtained by deflating capital at current replacement values by a price index of investment.

¹⁶ The law of motion for capital is as follows: $k_{it}=\left(1-\delta \right)k_{it-1}+I_{it-1}$. It implies that the capital used by a firm in a specific period t was contracted in year t–1 (this means assuming that the firm needs a full production year for capital to be ordered, received, and fixed before being in operation). Labor and materials (unlike capital) are decided in year t, that is the period in which they are utilized by the firm (hence, they can be a function of tfp_it). These timing assumptions suggest that both labor and materials are taken as nondynamic inputs (differently from capital).

¹⁷ To invert the materials demand function, we assume that this function is strictly monotonic in unobserved productivity.

¹⁸ The nonparametric components, as the function f(.), involved in the implementation of Wooldridge’s (2009) method are specified by third-degree polynomials.

¹⁹ In the ESEE, there are 20 manufacturing sectors according to the NACE classification.

²⁰ Considering all sectors in Table 3, the average elasticity for materials ($e_{its}^m$) is 0.639, for labor ($e_{its}^l$) 0.235, and for capital ($e_{its}^k$) 0.050.

²¹ In a more standard Cobb-Douglas specification of the production function, within industry markups variation depends only on revenue shares.

²² This variable can be considered as a firm-specific demand shifter. Doraszelski and Jaumandreu (2013) also use this information to proxy for the business cycle in Spain. They show that in the 1990s, this variable mirrored the macroeconomic cycle; that is, in periods where the economy was growing, firms tended to report that their markets were in expansion. We confirm this for the period 2000–2008, as the percentage of firms declaring a recessive demand is 17.88 percent, while in the period 2008–2014 the percentage grows to 46.84 percent. We use the dichotomous indicator of recessive demand instead of the continuous recessive index (from 0 to 100) to ease the interpretation of the cross-product variables of the dummy recessive demand and the firms’ business strategies of exporting and/or performing R&D (that are also dichotomous variables). Further, replacing the recessive demand indicator by the continuous recessive index does not qualitatively change our results.

²³ Blundell et al. (1999) suggest that permanent individual effects might be captured by the entry presample mean of the dependent variable, which acts as a sufficient statistic for unobserved firm heterogeneity.

²⁴ We have also experimented by considering as presample years 1997, 1998, and 1999, or 1996, 1997, and 1998. However, as results were similar, we opted for a more parsimonious approach.

²⁵ This term is generically calculated as the ratio of the density over the distribution function of a normal distribution (ϕ(Zθ)/ϕ(Zθ)), in which the argument (Zθ) is the index function from a probit model with a vector of regressors Z.

²⁶ Moreno Martín and Rodríguez Rodríguez (2004, 2010) using data from the ESEE for the period 1990–2000, and a different methodology to calculate markups, also found that exporters enjoy higher markups.

²⁷ The higher prices set by exporting firms may arise from: quality differences in the products they export and/or a different elasticity of demand in foreign markets. Hallak and Sivadasan (2009) and Kugler and Verhoogen (2012) report evidence in favor of the higher quality of products sold by exporting firms.

²⁸ For instance, product innovation R&D may result in product quality upgrading or product differentiation, what may contribute to reduce the price elasticity of demand.

²⁹ We have also tried with the alternative measure of calculating the cost of loans deviation with respect to the average of the sector in which the firm operates. In spite of results being qualitatively similar, we believe that our choice is more reliable since there may be sectors particularly affected by adverse borrowing conditions and this would not be reflected in a measure that uses sector averages for comparison.

³⁰ We find that foreign capital participation has a negative effect on firms’ continuing in operation. This result has been found in previous studies with the ESEE data (see, for instance, Beneito et al., 2015).

³¹ We name this estimation method as the Heckman-bivariate probit since the two strategies are estimated jointly and we correct for nonrandom selection into exports and performing R&D. We correct for sample selection as we observe firms’ choices only for those continuing in operation.

³² As in Almunia et al. (2018), we also control for supply-side factors such as firms’ TFP and financial restrictions.

³³ As we start estimation in 2000 and our regressors are lagged one period, we use as presample years 1997 and 1998.

³⁴ At the bottom of Table 6, we report the results for the Rivers-Vuong endogeneity test. As the retrieved residual from the first step instrumentation procedure is nonsignificant in the export equation, we may conclude that the variable Growth_domestic_sales does not suffer from endogeneity. We reinforce this conclusion by testing the second condition for their validity. We reestimate our specification augmenting the export equation with the residual and the l-1 instruments (in our case, l-1 is 1, as the number of instruments is 2). We provide an F-test for the exclusion of the instruments (IVs) in the export equation, and report this result at the bottom of Table 6. As before, we cannot reject the null of excluding them in the export equation.

Additional information

Funding

This work was supported by the Generalitat Valenciana [PROMETEU/2019/095] and the Spanish Agencia Estatal de Investigación (Spanish Ministerio de Economía, Industria y Competitividad) co-financed with FEDER funds, European Union [ECO2017-86793-R].