The question of what drives exports at the country level is one of the longest-standing discussions in international macroeconomics. Especially in Germany, which has traditionally been one of the countries with the highest exports in the world, fluctuations in merchandise exports indeed represent a substantial component of domestic economic growth volatility. Therefore, policy makers closely monitor these fluctuations.
Given Germany's export dependence and recent changes in the world economy, it seems highly relevant to analyze which factors other than the traditional export determinants (price competitiveness and foreign demand) are crucial in explaining Germany's export performance. To find out more about the driving forces of Germany's exports, our study goes beyond the standard export demand function and tests a wider range of factors and a longer time period than is typically analyzed. To assess the relative importance of the determinants, we employ the multivariate estimation technique of Johansen [1988, 1992a,b]. This approach allows for simultaneous estimation of the long-run relationship and the short-run adjustment process of the variables under consideration. Our study also examines the contribution of different determinants to the development of German exports over the more recent past.
The remainder of the article is structured as follows. Section 2 provides a literature overview. Section 3 explains the econometric methodology and the dataset. Section 4 presents the empirical results. Section 5 summarizes the main findings of the study and concludes.
Research papers based on firm-level data show that exporting firms are more productive than non-exporting firms and that they boost economic growth in the respective economy [De Loecker, 2007, 2011; Hansen, 2010]. Traditionally, the demand for exports is specified as a function of a country's price competitiveness and a foreign economic activity variable [Goldstein and Khan, 1985]. Both an improvement in price competitiveness and a rise in foreign demand are expected to lead to a rise in exports. Several studies estimate export demand elasticities for Germany and find the absolute values of the long-run price elasticity to range between 0.2 and 1.0 and the long-run income elasticity to range between 0.8 and 1.6 [Senhadji and Montenegro, 1999; Hooper et al., 2000; Strauß, 2000]. Although the results of these studies suggest that traditional determinants are essential in explaining German exports, substantial unexplained residuals remain. This outcome points towards the existence of crucial unobserved or omitted variables in the traditional export demand function. Therefore, this article aims at identifying some of these additional factors.
So far, there are only very few studies that go beyond the traditional export demand function and test the relevance of variables other than price competitiveness and foreign economic activity. Based on the arguments of the new trade theory [Krugman, 1983, 1985] and the endogenous growth theory [Grossman and Helpman, 1991], such studies mainly concentrate on factors such as globalization and quality differentials of traded goods. To capture the effects of globalization, introducing linear trends and dummy variables into the regression models has become predominant in the literature [Stephan, 2005; Barrell and Pomerantz, 2007]. Although there seems to be a consensus about the overall positive effect of globalization on German exports, the interpretation of the estimated coefficients is largely intuitive and indirect. For instance, a statistically significant trend in the export demand function can be attributed to different causes and, therefore, leaves significant room for interpretation. Hence, it is inevitably important to concentrate on isolated aspects of globalization to provide specific economic reasoning for the estimation results. In this study, we focus on the ongoing fragmentation of production processes as one potential determinant.
Concerning quality differentials of traded goods, Grossman and Helpman [1991] highlighted the importance of technological competitiveness in explaining trade flows. They argued that spending more on research and development enables firms to improve their product quality leading to an increased market share relative to competitors. To control the quality differentials, some empirical studies introduced specific R&D measures or patenting activities to the traditional export demand function [Madsen, 2004]. The findings of the existing studies, however, did not yield conclusive results for the case of Germany. Therefore, it seems useful to further investigate the importance of quality aspects for German export performance.
So far, empirical studies have generated little consensus about which factors are most important in explaining German exports. Most of them have examined either the effects of globalization or the effects of quality differentials as potential determinants. The objective of this article is to combine the two strands in the literature in one empirical analysis to derive the results that are more conclusive and contribute to explaining part of the large residuals of the traditional German export demand function. Besides, our analysis goes beyond the already existing explanations and includes the aspects of energy efficiency which is a characteristic often ascribed to German products.
As we are interested in the key drivers of German exports, we first discuss potential determinants before turning to the empirical analysis of factors that affect export performance in Germany. In the following, we list the factors that our analysis takes into account.
For robustness purposes, we also use the real effective exchange rate based on unit labor cost as an alternative. Based on the assumption that Germany is a small country, the advantage of this variable is that the estimated elasticity indicates whether German exports have grown to the same extent as export markets. A value of one indicates a constant market share. A value that is smaller than one reflects a loss in global export market share. For the construction of the variable, we use constant weights that are calculated as German export share averages over the period 2000–2003. The relationship between trade, foreign direct investment, and the activities of multinational corporations is studied in detail by Kleinert [2001]. We use the FDI stock variable rather than the flow variable, since stocks allow us to treat the FDI strategy as a long-run phenomenon, while looking at year-on-year evolution of FDI might blur the proper FDI strategy. The Renewable Energy Sources Act (EEG) draws on more than 15 years of experience. It has its origin in The Electricity Feed Act (StrEG) which became effective in 1991. In 2004, this act was replaced by The Renewable Energy Sources Act (EEG). According to the International Enegry Agency [International Energy Agency (IEA), 2008], it is the most important and successful instrument to promote the expansion of renewable energies in the electricity sector.
Having outlined the potential factors affecting Germany's exports, we now turn to the estimation technique. We employ the multivariate cointegration estimation approach to assess the relative importance of the factors discussed earlier. Theoretically, the export demand function reflects a long-run steady-state relationship. From an econometric point of view, this would imply a cointegration relationship between the variables under consideration. We apply the Johansen's [1992a,b] cointegration technique to determine the number of cointegration vectors in each model. The models under consideration entail more than two variables. Therefore, more than just one linear independent cointegration vector could exist. The Johansen's [1992a,b] procedure allows to test for more than just one cointegration vector, whereas the Engle and Granger's [1987] methodology only allows to test for one cointegration relationship and would, therefore, yield inconclusive results.
The list in the Appendix details the variables used in our study. All estimations use quarterly observations between 1992Q1 and 2016Q4. Data before 1992 are either missing or are dropped due to unification-related fluctuations. We use all variables in logs. This procedure allows us to interpret the result as export elasticities. Since our data do not include zero values, we do not need to adjust the data before converting them into logs.
We consider five alternative model specifications of our export demand function to derive the final cointegration relationship. This final specification (Model 6) comprises all significant variables, which were identified in the previous estimations. Table 1 shows the different specifications considered in our analysis.
Variables included in the different models
1 | x | x | x | ||||
2 | x | x | x | x | |||
3 | x | x | x | x | |||
4 | x | x | x | x | |||
5 | x | x | x | x | |||
6 | x | x | x | x | x |
Model 1 includes of two explanatory variables,
Johansen's cointegration test
0 | 44.653*** [0.003] | 25.327** [0.018] | 0 | 72.332*** [0.000] | 37.686*** [0.003] |
1 | 19.326* [0.067] | 12.013 [0.185] | 1 | 34.646* [0.057] | 18.102 [0.174] |
2 | 7.313 [0.110] | 7.313 [0.111] | 2 | 16.545 [0.150] | 12.020 [0.185] |
0 | 64.402*** [0.005] | 37.484*** [0.003] | 0 | 058.588*** [0.004] | 33.641*** [0.007] |
1 | 26.918 [0.293] | 12.644 [0.592] | 1 | 24.947 [0.163] | 19.285* [0.089] |
2 | 14.274 [0.271] | 8.305 [0.512] | 2 | 5.661 [0.735] | 3.801 [0.880] |
0 | 64.249*** [0.000] | 35.174*** [0.004] | 0 | 111.458*** [0.000] | 51.920*** [0.000] |
1 | 29.076* [0.060] | 18.923* [0.099] | 1 | 59.539 [0.110] | 27.113 [0.181] |
2 | 10.153 [0.269] | 8.465 [0.333] | 2 | 32.425 [0.366] | 19.588 [0.268] |
The table shows the estimates of the Johansen's cointegration test for Models 1–6 as specified in Table 1.
***, (**), [*] denote significance at 1, (5), [10%] level, respectively.
Together with the weak exogeneity of the explanatory variables leads us to continue with the analysis of a single equation error correction model. This allows us to interpret the long-run relationship as a structural export function. As in standard I(1) cointegration, the timing of variables in the cointegrating relation does not interfere with the cointegration property (Nielson, 2005). Therefore, we continue with the estimation of the conditional single equation error correction model of the following general form:
Table 3 contains the reduced rank cointegration relations with the numbers in parentheses and brackets representing the standard errors (SE) and the statistical significance levels, respectively. Table 2 also shows the adjusted
Reduced rank cointegration relations
βExports | 1 | 1 | 1 | 1 | 1 | 1 |
βDemand | –0.973*** (0.067) | –0.976*** (0.057) | –0.815*** (0.041) | –0.943*** (0.122) | –0.910*** (0.056) | –1 [0.113]a |
βREER | 0.630*** (0.237) | 0.746*** (0.217) | 0.682*** (0.131) | 0.687*** (0.229) | 0.605*** (0.210) | 0.426*** (0.142) |
βR&D | 0.030 (0.835) | |||||
βEnergy prices | –0.081*** (0.022) | –0.071*** (0.021) | ||||
βFDI | –0.013 (0.079) | |||||
βDomestic VA | 1.304* (0.796) | 1.837*** (0.643) | ||||
αExports | –0.256*** (0.071) | –0.294*** (0.048) | –0.377*** (0.063) | –0.257*** (0.049) | –0.213*** (0.065) | –0.424*** (0.069) |
Adj. R2 | 0.638 | 0.600 | 0.645 | 0.581 | 0.614 | 0.651 |
S.E. | (0.027) | (0.028) | (0.027) | (0.029) | (0.028) | (0.027) |
LM(4) | [0.658] | [0.359] | [0.701] | [0.341] | [0.058] | [0.169] |
JB | [0.739] | [0.349] | [0.949] | [0.603] | [0.139] | [0.942] |
The table shows the estimates of the long-run coefficients for Models 1–6 as specified in Table 1.
***, (**), [*] denotes significance at 1, (5), [10%] level, respectively; standard errors are given in parentheses.
Results of the Wald test:
The traditional determinants are highly significant and have the expected signs in all models. The estimation results are also robust to changes in model specifications. Furthermore, the estimation results indicate that two out of four additional determinants are crucial in explaining German exports.
The energy variable turns out to be highly significant. A rise in the oil price leads to an increase in German exports in the long run. This positive long-run relationship between the oil price and German exports can be interpreted as evidence of increased global demand for energy-efficient products and alternative energy technologies.
The estimation results further suggest that increased specialization by the ongoing fragmentation of production processes benefits the German export sector. In our analysis, we use the share of domestic value added in total output as a proxy of specialization. The sign of the coefficient indicates that a decline in the share of domestic value added in total output improves Germany's export performance. This indicates that increased regionalization of production processes contributes to the export performance in the long run.
While the energy variable and the domestic value-added variable turn out to be statistically significant determinants, the quality variable and the FDI variable seem to be statistically insignificant in explaining German exports. Hence, we do not find support for the hypothesis according to which Germany-specific quality aspects are crucial in explaining German export performance. Even in the case when the R&D variable enters the export demand equation in its lagged representation because of the argument that the contemporaneous effects of quality variables are unlikely, the results do not change. One reason for this outcome might be linked to the fact that R&D expenditure only entails information about the input to innovation but not information about successful research and therefore does not appropriately capture all quality aspects. In this respect, a more disaggregated approach might shed more light on the relationship between innovation and export performance [Cassiman et al., 2010].
The FDI variable also turns out to be statistically insignificant in explaining German exports. This outcome might be linked to the fact that the results of the Wald test for weak exogeneity suggest that the α-coefficient of the FDI variable is significantly different from zero indicating that FDI cannot be treated as weakly exogenous. However, if we treat the FDI variable as endogenous as well, we end up with a multivariate system which cannot be interpreted as a structural export demand function.
Based on the before-mentioned conditions, our final export demand model is given by the following explanatory variables:
In this study, we examine different hypotheses for their ability to explain German export performance during the period 1992–2016. The estimation results of the single-equation error correction model indicate that world demand, price competitiveness, energy prices, and the fragmentation of the production process are the main factors explaining Germany's exports. By contrast, we do not find conclusive evidence of Germany's exports being determined by R&D expenditure or FDI activities. The econometric results suggest that world demand is one of the key drivers of German exports. According to the estimation output, it is possible to treat demand as unit elastic. Hence, given that everything else remains unchanged, a 1% increase in global demand leads to an increase in German exports of the same size.
Further, the results indicate that German exporters have benefited from the ongoing specialization. Because German companies have optimized their value chain of production in such a way as to make use of the comparative advantages of individual firms and locations, they were able to increase their exports.
Price competitiveness plays a comparatively smaller role in explaining export growth. Even though prolonged effort in containing costs through wage moderation was significant, the effect is, in particular, diluted by the nominal appreciation of the euro before the outbreak of the worldwide financial crisis.
Further, the results of the study indicate that Germany particularly benefited from the ongoing globalization and the possibility of optimizing the value chain of production globally. Hence, the EU enlargement and the worldwide liberalization of capital and goods markets enable German manufactures to make use of the comparative advantages of different locations. Therefore, Germany's policy makers should continue to safeguard open borders as well as strong trade and investment links between economies worldwide.
The results on the relationship between the oil price and exports are of particular interest. The positive long-run relationship supports the argument that with rising energy prices and natural resources becoming scarce, the demand for energy-efficient products and alternative energy technologies increases. Since Germany is leading in these industries, the result might support the view that German manufacturers and the export sector strongly benefit from this development. However, further research is required to address this link more specifically. An alternative interpretation of the result might be that German exporters benefit from the petrol dollar effect. For instance, an analysis of a disaggregated product level might lead to more insights on that topic.