Sage Journals: Discover world-class research

Abstract

The induced innovation hypothesis, initially proposed by Sir John Hicks, posits that as the cost of energy rises compared to other input factors, firms are motivated to engage in innovative practices to counteract the increased expenses related to energy consumption. This innovation can manifest through the development and implementation of technologies, processes, or methodologies that enhance energy efficiency or diminish overall energy dependency. In this study, we empirically examine and validate this hypothesis. By theoretically modeling how innovation responds to elevated energy costs, we exploit China’s substantial surge in energy demand as an external shock to global demand, to empirically test the predictions associated with our theoretical framework. We test these predictions using firm level data in Belt and Road Initiative (BRI) countries. Our findings strongly support the induced innovation hypothesis, revealing that, on average, a 1 percent rise in the relative cost of energy corresponds to a 2.1 to 5.1 percent increase in the likelihood of innovation in energy-exporting countries and a 0.5 to 3.6 percent increase in non-energy-exporting countries. These results are robust to various methodological variations and data restriction exercises.

JEL Classification: D22, D24, O13, O14

Keywords

energy energy cost energy trade innovation competition

Twitter summary

As shown in Figure 1, China’s oil imports have grown exponentially in the last two decades and this has pushed up global energy prices. With rising global energy prices, are firms more likely to innovate? We examine this question and find strong affirmative evidence in our paper. A 1 percent increase in energy costs prompts a noteworthy surge in the likelihood of innovation, ranging from 2.1 to 5.1 percent in energy-exporting countries and 0.5 to 3.6 percent in non-energy-exporting nations.

Figure 1.

China’s imports and exports.

1. Introduction

. . . change in the relative price of the factors of production is itself a spur to invention, and to invention of a particular kind directed to economize the use of the factor which has become relatively expensive . . . (Hicks 1932)

In a world with increasing energy demand and finite supply, firms must compete for energy. This competition places an upward pressure on energy prices. The induced innovation hypothesis, proposed by Sir John Hicks suggests that increases in energy costs relative to other firm costs should lead to innovations that reduces energy usage.¹ This hypothesis has been widely tested and largely confirmed (see Popp [2002], Johnstone, Haščič and Popp [2010], Newell [2010], and Wurlod and Noailly [2018] as examples) but one glaring limitation of these studies is that although the induced innovation hypothesis isn’t exclusively tied to energy-saving frontier technologies, much of the existing research has predominantly taken a narrow view of innovation, often fixating on quantifiable metrics like the number of (registered or granted) patents (Marshall and Parra 2019).

While this focus is understandable, it overlooks the fact that firms may undergo significant changes in production processes and market orientation that are only tangentially related to energy reduction technologies. Such innovations, though crucial for economic development, often go unpatented. As such, we deviate from the existing literature by focusing on product innovation, process innovation, and investments in research and development (R&D), providing a more holistic view of innovation beyond patent-centric measures when evaluating the induced innovation hypothesis.

As a further extension to the literature, we develop a novel theoretical framework inspired by the work of Pindyck (1978), to demonstrate how an energy shock can lead to increased energy prices that catalyzes innovation. Our theoretical approach is also similar in spirit to the well-established agglomeration economics literature (e.g., Duranton and Puga 2004; Glaeser and Gottlieb 2009) and recent contributions in the energy economics literature that documented that firms benefit from knowledge spillover due to their proximity (backward and forward linkages) to the national oil and energy production sector (Badeeb, Lean and Smyth 2016; Fang and Chen 2017; Lim and Morris 2022). We add further context—in terms of the possible transmission mechanisms—by focusing on the firm level implications of shocks to domestic energy supply. This approach we argue is of great interest in the context of increased bilateral trade relations that are premised on “resource-for-capital” arrangements in developing economies, especially given the Arrow-Romer-Marshall spillover externality that promotes innovation.

To empirically operationalize an energy shock, we leverage the unique circumstances surrounding China’s rapid emergence as a manufacturing export powerhouse following its trade and investment liberalization in the early 1990s (see Figure 1). The surge in China’s manufacturing exports significantly impacted economies producing labor-intensive goods, creating a substantial rise in global competitive supply of manufactured goods. Conversely, economies supplying raw materials and other inputs witnessed a parallel surge in global demand for these inputs due to China’s increased demand for these inputs (Autor, Dorn and Hanson 2016). Of interest to us, the rise in China’s manufacturing exports corresponded with a substantial spike in the country’s energy imports, contributing to its role as a primary driver of the commodities super cycle from 2000 to 2014 (Gauvin and Rebillard 2018). Leveraging this, we investigate whether China’s increased energy demand had a discernible impact on global energy markets, subsequently influencing innovation at the firm level in other countries. As such, this approach employs the China trade shock as an exogenous instrument for import competition, a well-documented justification in the literature (Autor, Dorn and Hanson 2013, 2016; Utar 2014; Utar and Torres Ruiz 2013). This allows us to make a stronger case for a causal interpretation of our results compared to the existing literature. Moreover, we use the broadest cross-country firm level dataset to examine this relationship to date.

We focus on firms in ninety countries that are part of the Belt and Road Initiative (BRI) with China.² Notably, as illustrated in Figure 2, China’s energy imports now originate from around the world. In 1992, China sourced energy inputs from just twenty-three countries, a number that increased to eighty-five by 2020. We therefore focus on the BRI countries because the energy demand component of the China shock is more targeted in these countries due to the unique resources-for-capital arrangements employed by China under the initiative (Huang 2016). Under the initiative, loans of Chinese origin are collateralized either directly by natural resources or indirectly through future revenue streams from natural resources. As an example, these loan arrangements account for approximately a quarter of all Chinese loan commitments in Africa (Chuku, Lang and Lim 2023). Baniya, Rocha and Ruta (2020) points out that the trade effects of the BRI present a large variation across individual countries and across sectors. This suggests that there is increased competition from China in these countries in both goods and input markets and as such, our analysis presents a reference point for these increased competitive pressures.

Figure 2.

Change in China’s energy imports 1992 to 2020 (%).

To preview our findings, we consistently show that rising energy costs act as a catalyst for innovation across diverse measures of innovation, including product innovation, process innovation, and R&D. We find that a 1 percent increase in energy costs prompts a noteworthy surge in the likelihood of innovation, ranging from 2.1 to 5.1 percent in energy-exporting countries and 0.5 to 3.6 percent in non-energy-exporting nations. These results underscore the significance of energy cost dynamics in steering innovation strategies, irrespective of a country’s energy export status. From a policy standpoint, governments can leverage these insights to design targeted initiatives that incentivize innovation in response to changing energy cost dynamics, fostering a more adaptive and resilient business landscape. These results are robust to diverse methodological modifications to our baseline strategy and data restrictions.

The rest of the paper is organized as follows. Section 2 provides an overview of the relevant literature. Section 3 includes a theoretical framework for our analysis. Section 4 reviews the empirical strategy and describes the data. Section 5 contains the main results on the innovation–energy cost relationship. We discuss our results in Section 6 before offering some concluding remarks in Section 7.

2. Relevant Literature

As energy prices have risen recently, there is growing evidence confirming the induced innovation hypothesis (Crabb and Johnson 2010; Johnstone, Haščič and Popp 2010; Newell 2010; Popp 2002; Triguero, Moreno-Mondéjar and Davia 2014; Wurlod and Noailly 2018). While increased energy prices may reduce R&D investment and innovation by undermining profits, the high input cost may induce firms to explore energy-saving efforts, the crux of the induced innovation hypothesis. This hypothesis suggests that with increased energy prices there is a growth of innovation in energy saving technologies. These technologies include technologies that reduce pollution, such as scrubbers for use on industrial smokestacks or catalytic converters for automobiles and changes to the production process, such as improved energy efficiency, that lead to reduced energy use and negative environmental impacts.

Newell, Jaffe and Stavins (1999) develops a methodology for testing the induced innovation hypothesis by estimating a product-characteristics model of energy-using consumer durables and show that the direction of innovation was responsive to energy price changes for some products but not for others. Similarly, Popp (2002) used United States patent data from 1970 to 1994 to estimate the effect of energy prices on energy-efficient innovations and find that both energy prices and the quality of existing knowledge influences the incentives to innovate. Using cross country data, Wurlod and Noailly (2018) analyzes the impact of green innovation on energy intensity and shows that green innovation has contributed to the decline in energy intensity in most sectors. Similar findings were reported by Johnstone, Haščič and Popp (2010) and Newell (2010) who show that public policy has a significant role to play.

Interestingly, almost all these analyses use patent count as the measure of innovation. Of course, this limits our overall understanding of these relationships because patents are an imperfect measure of technological innovation (Morris 2018). For instance, Johnstone, Haščič and Popp (2010) point out that patent count is a subset of product innovation, and it is highly correlated with firm size and is a crude measure of innovation outputs that captures only some types of invention. Additionally, filing for a patent is driven in large part by the extent to which a firm’s management/and or ownership deems it important given the nature of the business, the institutional environment with regards to enforcement and appropriability, among other challenges. For developing countries, where patenting is less common (Zahler, Iacovone and Mattoo 2014), estimating the impact of energy prices on patent count may suffer purely because of the limited numbers of patents recorded in these countries.

Although binary in nature, studies such as Cirera and Muzi (2016) show that in some instances, measuring innovation more broadly as product and process innovation offers some improvement on patent count measurements which mainly accounts for radical product innovation. Moreover, these studies tend to focus on developed economies, where investment in R&D is higher and innovation tends to be on the technological frontier (Crespi, Tacsir and Vargas 2016). Our extension of the literature in this regard is to focus on the association between innovation (product and process innovation) and energy demand in mostly developing countries.

We also include services firms in our analysis. Morris (2018) points out that characteristics such as intangibility, simultaneity of production and consumption, and perishability, make innovation in the services sector different to manufacturing firms. For example, Zahler, Iacovone and Mattoo (2014) shows that services firms out innovate manufacturing firms in some industries, even though services innovations more often take a non-technological form. A similar finding was reported by Tether (2005) in several developed European countries. With these exceptions, analysis of innovation behavior in the services sector is limited, and as it regards the interaction with energy, there is a dearth of evidence. It is our intention to fill this gap in the literature by including both manufacturing and services firms in our analysis, covering firms across thirty-six industries in total. With many developing economies also currently being encouraged to undertake service sector liberalization, we believe the additional coverage on service sector firms will provide either alternative insights or further robustness to results obtained in our core analysis.

3. Theoretical Framework

We develop a theoretical framework that characterizes a non-oil price setting, oil producer that is engaging in a stylized “resource-for-capital” trade arrangement. In essence, we create a broad macro-environment for which the firms—which are our main research focus—are operating in, based on an extended optimal extraction and drilling problem in the tradition of Pindyck (1978). After establishing the macro-environment, we deviate from the general equilibrium set up in the standard macroeconomic modeling frameworks and focus on explaining analytically the effects of rising energy demand on firm level innovation choice.

With the theoretical intuition of competitive pressures resulting from rising energy demand shocks set in place, we consider separately the influence on innovation in energy exporting and importing countries. The firms we study are the “everyday” enterprises in the developing host economy, that are price takers.

In this setting, the energy price effect is transmitted via macroeconomic mechanisms associated with the national energy sector, which then spillover to the “everyday” small and medium sized enterprises. This theoretical approach is similar to Chuku, Lang and Lim (2023) but we generalize the mechanisms to highlight that in general, demand shocks that raise energy prices will have an impact on the competition for energy and consequently, will impact firm level activity globally.

3.1. Macro-environment

We characterize the macro-environment that the firms are operating in using a Pindyck (1978) style oil extraction and drilling framework, where the construction of the physical capital (equipment and machinery) used in the operation is funded by a “resource-for-capital” trade arrangement that is like Chuku, Lang and Lim (2023), and consistent with anecdotal evidence in Lin and Wang (2017) and Johnston (2019). Specifically, drilling activities involve the use of physical capital financed through a special purpose vehicle (SPV) funded from a foreign country (herein, “foreign donor”), where the equation of motion is

K_{t + 1}^{C} = (a_{t})^{κ} {(g_{t}^{C})}^{(1 - κ)} + (1 - δ) K_{t}^{C},

(1)

where $a_{t}$ $(g_{t}^{C})$ is the contribution made by the foreign donor (home economy), $κ$ is the share of the donor, and $δ$ is the depreciation rate. Due to the “resource-for-capital” arrangement, the home share, $g_{t}^{C}$ , is funded by the donor too (assumed to be the same, $g_{t}^{C} = a_{t}$ ), and in exchange the donor receives some fraction of imported oil. The SPV rents out the physical capital to the oil operator and receives gross returns, $(1 + r_{t}^{C}) K_{t}^{C}$ in each period, which implies that the revenue share of the donor and home government is $κ (1 + r_{t}^{C}) K_{t}^{C}$ and $(1 - κ) (1 + r_{t}^{C}) K_{t}^{C}$ respectively.

Given these, the aid funds’ flow constraint of the donor is given by:

Γ_{t} + ς P_{t}^{O} X_{t} + κ (1 + r_{t}^{C}) K_{t}^{C} = 2 a_{t},

(2)

where $Γ_{t}$ is the funds set aside by donor government for aid finance, $X_{t}$ and $P_{t}^{O}$ are oil extracted and its price in period $t$ , and $ς \in (0, 1)$ is the fraction of extracted oil claimed by the foreign donor. Rearranging terms in (2), we have

K_{t}^{C} = \frac{(2 a_{t} - Γ_{t})}{κ (1 + r_{t}^{C})} - \frac{ς P_{t}^{O} X_{t}}{κ (1 + r_{t}^{C})} .

(3)

Following the geological constraint-restricted extraction function of Okullo, Reynès and Hofkes (2015), period-specific oil extraction, $X_{t}$ , is given by

X_{t} = α_{0} {(Φ_{t})}^{α_{1}}, where α_{0} > 0, α_{1} \in (0, 1],

(4)

and would reduce existing reserves, $Φ_{t}$ , whereas drilling activities, $ω (K_{t}^{C})$ , add to reserves. The equation of motion for oil is therefore

Φ_{t + 1} = Φ_{t} - X_{t} + ω (K_{t}^{C}), where

(5)

ω (K_{t}^{C}) = ω_{0} {(K_{t}^{C})}^{ω_{1}},where ω_{0} > 0, ω_{1} \in (0, 1) .

(6)

The energy operator chooses sequences of donor funds{a_t}0∙ (from choice of quantity of equipment, [3]), and next-period oil reserves ${Φ_{t + 1}}_{0}^{\infty}$ , to maximize

\max_{{a_{t + s}, Φ_{t + s + 1}}_{s = 0}^{\infty}} E_{t} \sum_{s = 0}^{\infty} β^{s} V_{t + s}^{O} = E_{t} \sum_{t = 0}^{\infty} β^{s} [(1 - ϱ) P_{t + s}^{O} X_{t + s} - (1 + r_{t + s}^{C}) K_{t + s}^{C}],

(7)

where $E_{t}$ is the expectation operator, $β \in (0, 1)$ the subjective discount rate, and unit marginal cost of extraction is assumed to be a $ϱ \in (0, 1)$ fraction of oil revenue, subject to (3)–(6) and the terminal condition, $\lim_{T - > \infty} β^{T} E_{t} (λ_{t + T} - \frac{\partial V_{t + T}^{O}}{\partial Φ_{t + T + 1}}) Φ_{t + T + 1} = 0$ , where $λ_{t}$ is shadow price of the oil operator. As shown in the Supplemental Appendix, the optimal oil extraction in period- $t$ as a share of total oil reserves is given by

\begin{matrix} \frac{X_{t}}{Φ_{t}} = {\begin{matrix} \frac{β P_{t}^{O} [(1 - ϱ) α_{1} + κ^{- 1} ς]}{{(ω_{0} ω_{1})}^{- 1} {(K_{t - 1}^{C})}^{1 - ϖ_{1}}} \\ + {(\frac{K_{t}^{C}}{K_{t - 1}^{C}})}^{1 - ϖ_{1}} \frac{(1 + r_{t}^{C})}{(1 + r_{t - 1}^{C})} β [\frac{ω_{0} ω_{1} ς P_{t}^{O}}{κ (1 + r_{t}^{C}) {(K_{t}^{C})}^{1 - ϖ_{1}}} + α_{1}] \end{matrix}}^{- 1} \\ \times [{(\frac{K_{t}^{C}}{K_{t - 1}^{C}})}^{1 - ϖ_{1}} \frac{(1 + r_{t}^{C})}{(1 + r_{t - 1}^{C})} β - 1] . \end{matrix}

(8)

In terms of the rental rate, $(1 + r_{t}^{C})$ , given that (1) and $g_{t}^{C} = a_{t}$ implies that $K_{t + 1}^{C} = a_{t} + (1 - δ) K_{t}^{C}$ , in standard fashion of a capital good producer, the SPV chooses the sequences of {KC_t}∙t=0, taking the rental rate $(r_{t}^{C})$ , world risk-free rate ( $r^{W}$ ; theoretically the opportunity cost of the donor funds), existing stock as given, to maximize the lifetime discounted value of rental income:

\begin{matrix} {K_{t + s + 1}^{C}}_{s = 0}^{\infty} = \arg \max \sum_{s = 0}^{\infty} E_{t} [Λ^{s} μ_{t + s} Π_{t + s + 1}^{C}], where Λ \in (0, 1),and \\ Π_{t + 1}^{C} = (1 + r_{t}^{C}) K_{t}^{C} - (1 + r^{W}) a_{t}, \end{matrix}

which, as shown in Supplemental Appendix A, yields

(1 + r_{t + 1}^{C}) = (1 + r^{W}) [\frac{μ_{t}}{μ_{t + 1}} \frac{1}{Λ} - (1 - δ)],

(9)

where µ _t is the shadow price derived from the non-oil economy’s problem.

Having deducted the priority share claimed by the donor, $ς P_{t}^{O} X_{t}$ , the remainder of the oil extracted is exported to the rest of the developing world ( $σ \in (0, 1)$ share) and then supplied to the domestic firms. Let $X_{D t}$ denotes the total supply of oil to the domestic firms, the domestic oil market identity is given by

(1 - σ) (1 - ς) P_{t}^{O} X_{t} = (1 + τ_{t}) P_{t}^{O} X_{D t},

(10)

where $| τ_{t} | < 1$ is a tax (if positive; subsidy if negative) rate imposed at the domestic oil price level by the government.

Finally, the home government, who saves by purchasing net foreign assets, $b_{t}$ , at a world risk-free rate, $1 + r^{W}$ , acts as a benevolent social planner to maximize the discounted sum of the aggregate utility function of domestic agents (largely to save space), which depends only on consumption, $C_{t}$ , as in:

\begin{array}{l} \max_{{C_{t + s}, b_{t + s + 1}}_{s = 0}^{\infty}} E_{t} \sum_{s = 0}^{\infty} Λ^{s} \ln C_{t + s},subject to \\ τ_{t} P_{t}^{O} X_{D t} + (1 - κ) (1 + r_{t}^{C}) K_{t}^{C} + w_{t} N_{t} = C_{t} - (1 + r^{W}) b_{t} + b_{t + 1}, \end{array}

(11)

where $w_{t} N_{t}$ is the wage income accrued by domestic agents’ labor supply to firms, assumed to be fully observed and considered by the home government.

As shown in the Supplemental Appendix, let µ _t be the shadow price of (11), we have:

\frac{μ_{t}}{μ_{t + 1}} = Λ (1 + r^{W}) = \frac{C_{t + 1}}{C_{t}} .

(12)

3.2. Firms and Innovation Choice

There is a continuum number of firms indexed by $j \in (0, 1)$ operating in the domestic macro-environment described, each being a monopolistically competitive seller that solves a two-step optimization problem (competitive price taker in input markets, price setter in output market). Specifically, each firm first observes the demand for its product variety and therefore the setting of profit maximizing price, and then it chooses quantities of inputs that are required to produce this demand to minimize its unit production cost, taking unit input prices as given. Finally, after this two-step mechanism, a firm only decides on its innovation decision based on the expected profit it will receive.

This type of two-step optimization approach is commonly used in the macroeconomic literature to model the firm’s problem in a developing small open economy setting, especially when the focus is on examining “macro-to-micro” spillover effect from the broader macroeconomic sectors. Examples focusing on spillover from the environment and the national energy sector to firms include Annicchiarico and Dio (2017) and Lim and Zhang (2023) respectively.

Collectively, goods market equilibrium requires that total consumption, $C_{t}$ , in each period $t$ is met by the supply of a composite final good, $Y_{t}$ , which consists of baskets of innovative varieties $(M_{t}^{I})$ and baseline varieties $(M_{t}^{B})$ , as in Brambilla (2009) and Brambilla, Hale and Long (2009). For simplicity, we assume the entry and exit in each period exactly cancels out and each firm produces only one variety depending one’s innovation choice and outcome, therefore $1 = M_{t}^{B} + M_{t}^{I}$ . Specifically,

Y_{t} = {\int_{0}^{M_{t}^{B}} {[Y_{j t}^{B}]}^{(ϑ - 1) / ϑ} d j + \int_{0}^{M_{t}^{I}} {[Y_{j t}^{I}]}^{(ϑ - 1) / ϑ} d j}^{ϑ / (ϑ - 1)},

(13)

where $ϑ > 1$ is elasticity of substitution, $Y_{j t}^{B}$ $(Y_{j t}^{I})$ is quantity of baseline (innovative) product by firm $j$ . As shown in Supplemental Appendix A, the implied demand function for each firm $j$ is then $Y_{j t}^{ι} = (P_{j t}^{ι})^{- ϑ} P_{t}^{ϑ} Y_{t} ι = B, I$ , where the aggregate price index is defined by $P_{t} = {[\int_{0}^{M_{t}^{B}} {(P_{j t}^{B})}^{1 - ϑ} d j + \int_{0}^{M_{t}^{I}} {(P_{j t}^{I})}^{1 - ϑ} d j]}^{[\frac{- 1}{ϑ - 1}]} = 1$ (normalized, as shown above).

In the beginning of each period $t$ , each firm $j$ learns the technology of both type of production,

Y_{j t}^{ι} = A_{j t}^{ι} {(x_{D j t}^{ι})}^{γ} {(N_{j t}^{ι})}^{1 - γ}, γ \in (0, 1), ι = B, I,

(14)

where $x_{D j t}^{ι}$ $(N_{j t}^{ι})$ is the quantity of oil (labor) inputs employed by firm $j$ , and total factor productivity (TFP) faced by a firm $j$ would differ between type, as follow:

A_{j t}^{ι} = {\begin{array}{l} A_{j 0} \\ A_{j 0} {(K_{t}^{C})}^{ξ} \end{array} \begin{array}{l} if ι = B \\ if ι = I \end{array},

(15)

where $A_{j t}^{i}$ , is subject to a firm-specific term, $A_{j 0} > 0$ , drawn from a distribution $G (A_{j 0} | \bar{A})$ , where Ā. is the mean of $A_{j 0}$ , and an Arrow-Romer externality that is type specific, where a firm engaging in innovation would experience a further productivity boost in the form of an Arrow-Romer-Marshall type of externality, at a strength, $ξ \geq 0$ , that is positively dependent on the capital stock built in the oil sector. We believe equation (15) is an improved approach in modeling the productivity spillover from linkages with the national oil sector than the approach used in studies such as Lim and Morris (2022), as it not only accounts for the distinction between innovators and non-innovators [such spillover-by-proximity is usually applicable only to the former, as in Audretsch and Feldman (1996, 2004), but also reflects the well-documented infrastructure spillover effects in the literature (e.g., Duggal, Saltzman and Klein 1999; Romp and De Haan 2005)].

For a more realistic characterization of firm heterogeneity in terms of cost structure, we assume that each firm j’s actual realized fixed, $F_{j t}^{A c t u a l, ι}$ , and unit marginal costs, $m c_{j t}^{A c t u a l, ι}$ , consists of both deterministic and stochastic components. Specifically, while prior to realization a firm $j$ would have decided on their optimal costing, pricing, and innovation choice based on the expected or more precisely, deterministic component explained by the economic mechanisms below, to mirror empirical reality the observed realized costs differ across firms due to the stochastic components being randomly drawn from distributions, $ϒ_{F} (F_{j t}^{A c t u a l, ι} | {\bar{F}}_{j t})$ and $ϒ_{m c} (m c_{j t}^{A c t u a l, ι} | m c_{j t}^{ι})$ , where ${\bar{F}}_{0} > 0$ is a deterministic mean value of fixed cost for all firms and $m c_{j t}^{ι}$ is obtained from firm $j$ ’s cost minimization problem,

\min_{x_{D j t}^{ι}, N_{j t}^{ι}} m c_{j t}^{ι} = w_{t} N_{j t}^{ι} + (1 + τ_{t}) P_{t}^{O} x_{D j t}^{ι}, ι = B, I,

where $m c_{j t}^{ι}$ is the unit marginal cost, taking real wage $(w_{t})$ , the tax/subsidy rate $(τ_{t})$ , and oil price as given, subject to (14), therefore yielding

m c_{j t}^{ι} = \frac{Θ_{γ}}{A_{j t}^{ι}} w_{t}^{γ} {[(1 + τ_{t}) P_{t}^{O}]}^{1 - γ}, s = B, I, where Θ_{γ} = \frac{γ^{- γ}}{{(1 - γ)}^{1 - γ}} .

(16)

Based on this, with monopolistically competitive pricing for its own product, a firm $j$ ’s price is set at a mark-up,

P_{j t}^{ι} = (\frac{ϑ}{ϑ - 1}) m c_{j t}^{ι}, ι = B, I .

(17)

We model innovation investment (if successful) as resulting in some hybrid product and process innovation outcomes. Nevertheless, the effective cost of innovation is heterogeneous across firms due to all firms having different innovation efficiency, $χ$ (as in Hallak and Sivadasan [2013]), which follows a continuous distribution with density function $f (χ)$ and cumulative distribution function $F (χ)$ , with support $(0, χ_{\max})$ . Specifically, the effective cost of innovation investment faced by a firm $j$ is specified as

ϰ_{j t} (\cdot) = \frac{ϰ_{j 0}}{{(χ)}^{ϰ_{1}}} {(X_{D t})}^{- ϰ_{2}}, where ϰ_{j 0} > 0, ϰ_{1}, ϰ_{2} \geq 0,

(18)

where in addition to firm-specific innovation efficiency, also depends negatively on the size of the domestic oil market, given by:

X_{D t} = (1 - σ) (1 - ς) X_{t} = \int_{0}^{M_{t}^{B}} x_{D j t}^{B} d j + \int_{0}^{M_{t}^{I}} x_{D j t}^{I} d j .

(19)

In other words, we account for the broad economic structure of the economy. Given the focus on economies with oil exports, as well as recent spillover evidence such as Mody (2023), it is reasonable to assume that innovation is pseudo-complementary to the domestic energy market, therefore a negative innovation cost-domestic energy market size nexus. Further, on top of innovative versus baseline product variety choice, a successful innovation endeavor would also result in a proportionate fixed cost saving at $η \in (0, 1)$ . The success of innovation investment is uncertain, with its realization governed by a two-state Markov process, where at a state $ϕ \in (0, 1)$ innovation is successful.

Given these, plus (18) and (19), a firm $j$ opts to invest in innovation if and only if

ϕ E_{t} (π_{j t}^{I}) + (1 - ϕ) E_{t} (π_{j t}^{B}) - ϰ_{j t} (\cdot) \geq E_{t} (π_{j t}^{B}),

which, as shown in Supplemental Appendix A, yields a threshold value of innovation efficiency, $χ_{C t}$ , above which a firm $j$ invests in innovation:

χ_{C t} \geq {[\begin{matrix} Θ_{ϑ} Y_{t} \frac{Θ_{γ}^{1 - ϑ}}{A_{j 0}^{1 - ϑ}} w_{t}^{γ (1 - ϑ)} \frac{{[(1 + τ_{t}) P_{t}^{O}]}^{(1 - γ) (1 - ϑ)}}{{[{(K_{t}^{C})}^{- ξ (1 - ϑ)} - 1]}^{- 1}} \\ + (1 - η) {\bar{F}}_{j t} \end{matrix}]}^{- \frac{1}{ϰ_{1}}} \frac{{(\frac{ϰ_{j 0}}{ϕ})}^{ϰ_{1}^{- 1}}}{{[(1 - σ) (1 - ς) X_{t}]}^{ϰ_{2} ϰ_{1}^{- 1}}} .

(20)

Proposition 1: In a small energy exporting economy—the effect of the share of energy exported $(ς)$ on a domestic firm’s innovation engagement is ambiguous but depends on the size of the efficiency gains it can make to offset the costs of substituting the source of its energy input.

Proof: As shown in Supplemental Appendix A, differentiating (20) with regards to $ς$ , knowing that from (3) $\partial K_{t}^{C} / \partial ς < 0$ , yields

\begin{matrix} \frac{\partial χ_{C t}}{\partial ς} = {\frac{ξ (1 - ϑ)}{ϰ_{1}} \frac{{(\frac{ϰ_{j 0}}{ϕ})}^{ϰ_{1}^{- 1}} \frac{Θ_{γ}^{1 - ϑ}}{A_{j 0}^{1 - ϑ}} w_{t}^{γ (1 - ϑ)} {[(1 + τ_{t}) P_{t}^{O}]}^{(1 - γ) (1 - ϑ)}}{{(Θ_{ϑ} Y_{t})}^{- 1} {[(1 - σ) (1 - ς) X_{t}]}^{ϰ_{2} ϰ_{1}^{- 1}} {(K_{t}^{C})}^{[ξ (1 - ϑ) + 1]}}} \\ \times {[\begin{matrix} Θ_{ϑ} Y_{t} \frac{Θ_{γ}^{1 - ϑ}}{A_{j 0}^{1 - ϑ}} w_{t}^{γ (1 - ϑ)} \frac{{[(1 + τ_{t}) P_{t}^{O}]}^{(1 - γ) (1 - ϑ)}}{{[{(K_{t}^{C})}^{- ξ (1 - ϑ)} - 1]}^{- 1}} \\ + (1 - η) {\bar{F}}_{j t} \end{matrix}]}^{- (\frac{1}{ϰ_{1}} + 1)} \frac{\partial K_{t}^{C}}{\partial ς} \\ + (1 - σ) X_{t} ϰ_{2} ϰ_{1}^{- 1} \frac{χ_{C t}}{{[(1 - σ) (1 - ς) X_{t}]}^{- ϰ_{2} ϰ_{1}^{- 1}}} {[(1 - σ) (1 - ς) X_{t}]}^{- (ϰ_{2} ϰ_{1}^{- 1} + 1)}, \end{matrix}

which is analytically ambiguous. The sign of this comparative static depends on the trade-off between the two expressions. Specifically, the sign of the first expanded expression, which represents marginal gains with respect to Arrow-Romer spillover effect, is negative, which suggests that greater oil exports would lead to lower threshold value, $χ_{C t}$ , therefore greater likelihood for a firm to innovate. In contrast, the sign of the second expression, which represents a marginal cost with respect to the domestic resources foregone that could have otherwise gone into domestic innovation, is positive, which suggests that more exposure to BRI scheme would lead to lower likelihood for a firm to innovate.

3.3. Extension—Rest of Oil-Importing Small Economies

Given the benchmark characterization described by (1) to (20), to understand fully the impact of growing global energy demand on technological innovation of firms in developing economies, we turn our attention to the rest of the developing world, that is, small economies that may neither be oil exporters or recipient of “resource-for-capital” trade but are oil importers. Without a full-blown modeling of such an economy, we evaluate stylistically the impact on a typical firm $i$ ’s business mechanism in such a third oil-importing economy.

Without losing any generality, suppose there is only one such third-party, oil-importing economy. Recall that total oil imported from the oil-exporting economy in the benchmark model is then $σ (1 - ς) P_{t}^{O} X_{t}$ . For a firm $i$ that goes through a similar multi-staged operational decision-making process as described for a typical firm $j$ in the BRI-participating oil-exporting economy (see [13]–[20]), the differences (compared to firm $j$ ) would be as follows: (i) unlike (15), the spillover channel through the Arrow-Romer-Marshall externality would disappear as there is no foreign donor-funded infrastructure capital being built; (ii) spillover from the market size/proximity effect remains, though unlike (18) for oil-exporting economies, we cannot confidently assume that the spillover mechanism to individual firms would be positive, given the context of oil importers. Specifically, the effective cost of innovation investment faced by a firm $i$ in the oil-importing country is specified as

ϰ_{i t} (\cdot) = \frac{ϰ_{i 0}}{{(ψ)}^{ϰ_{I 1}}} {[σ (1 - ς) P_{t}^{O} X_{t}]}^{ϰ_{I 2}}, where ϰ_{i 0} > 0, ϰ_{I 1} \geq 0, ϰ_{I 2} \in R

(21)

where the oil-importing country’s corporate innovation efficiency is characterized by density function $f (ψ)$ and cumulative distribution function $F (ψ)$ instead, with support $(0, ψ_{\max})$ , and $ϰ_{I 2}$ can either be positive, zero, or negative. In the case of the former where smaller domestic oil market leads to lower effective cost of innovation, then there is an induced domestic competition effect that works in contrast to the mechanism described for the oil-exporting economy earlier.

Given such differences, the derivation of a threshold innovation efficiency level, $ψ_{C t}$ , above which a firm $i$ in an oil-importing economy would invest in innovation would then yield:

ψ_{C t} \geq {\frac{ϰ_{i 0}}{ϕ [E_{t} (π_{i t}^{I}) - E_{t} (π_{i t}^{B})]} {[σ (1 - ς) P_{t}^{O} X_{t}]}^{ϰ_{I 2}}}^{ϰ_{I 1}^{- 1}} .

(22)

From this, it is straightforward to derive another empirically testable proposition (see Supplemental Appendix A for derivation).

Proposition 2: In a small energy importing economy—the effect of an increase in global energy demand on a domestic firm’s innovation depends on the difference between the magnitude of the competition effect and the energy efficiency effect.

4. Empirical Strategy and Data

4.1. Empirical Strategy

To evaluate the two propositions articulated above (with split sample) we can estimate:

I n n o v_{i j c} = α_{0} + α_{1} E C_{i j c} + α_{2} C o n t r o l s_{i j c} + ε_{i j c},

(23)

where $I n n o v_{i j c}$ is the reported innovation of firm $i$ in industry $j$ of country $c$ , $E C_{i j c}$ is the relative cost of energy to total cost, $C o n t r o l s_{i j c}$ is a battery of firm-characteristic variables, and dummies that capture region-year-income group-(industry)-[country] specific fixed effects, and $ε_{i j c}$ is the random error term.³

This regression is the foundation for testing the induced innovation hypothesis and correlates firm level energy costs with innovation. For the hypothesis to be supported, $α_{1}$ should be positive and significant, highlighting that a rise in the relative cost of energy will spur firms to innovate. We highlighted in the theoretical section that sharp rises in energy costs have different influences on firm level innovation depending on whether they are in energy exporting or energy non-exporting countries. To operationalize this dichotomy, we can test proposition 1 using firms in energy exporting countries only and proposition 2 using firms in countries with no energy exports. The problem with this is that energy costs may drive innovation, but innovation may also drive a firm’s outlay on energy. This bi-causality is one potential source of endogeneity in equation (23). In such cases, $α_{1}$ may understate the impact of rises in relative energy costs on innovation.

As such, to identify the causal effect of rising energy cost on firm level innovation outcomes, we exploit cross-industry-country differences in energy costs to isolate the long-run, quasi-fixed component of an energy shock that drives up demand and increases prices. To do this we focus on the drastic rise in energy demand in China that started in the early 1990s. Empirically we focus on the change in Chinese trade in energy between 1992 and 2007. Autor, Dorn and Hanson (2016) points out that this period is characterized by an export surge driven largely by productivity gains resulting from China’s market opening. This period also avoids an overlap in the time frame of the data from the World Bank Enterprise Surveys (WBES) on innovation (2008–2019), avoiding reverse causality from innovation to energy trade.

Consequently, our instrumental variables approach is as follows. For proposition 1, let $Δ E X^{c - C h i n a}$ equal the change in exports from country $c$ to China between 1992 and 2007 and $R_{i, - j}$ is the relative energy cost share in all industries except the industry with firm $i$ . The product of these two measures provides a predicted value for the energy demand shock experienced by each firm that is isolated from its contemporaneous energy usage. This can be expressed as:

Δ E D S_{i j c} = \sum_{j = 1}^{J} R_{i, - j} \times Δ E X^{c - C h i n a}

(24)

For proposition 2, since the countries in focus do not export energy, we replace $Δ E X^{c - C h i n a}$ with $\frac{Δ M^{c}}{Δ M^{W o r l d - C h i n a}}$ where ∆ $M^{c}$ is the change in country $c^{'} s$ imports of energy and $Δ M^{W o r l d - C h i n a}$ is the change in China’s imports of energy from the rest of the world, again over the period 1992 to 2007. As such, to test proposition 2, our energy demand shock measure becomes:

Δ E D S_{i j c} = \sum_{j = 1}^{J} R_{i, - j} \times Δ \frac{M^{c}}{M^{W o r l d - C h i n a}}

(25)

Our final IV approach therefore is to instrument EC in equation (23) with ∆ $E D S$ so that the key regression is:

I n n o v_{i j c} = β_{0} + β_{1} Δ E D S_{i j c} + β_{2} C o n t r o l s_{i j c} + ϑ_{i j c},

(26)

where $I n n o v_{i j c}$ and $C o n t r o l s_{i j c}$ are as described above and $ϑ_{i j c}$ is the error term.

4.2. Data

Our analysis is based on cross-country firm level survey data from the WBES. In addition to information on innovation, this survey contains data on accounting characteristics such as sales, inputs, labor, stock of capital, investment, costs as well as broad cost-of-doing business indicators, and government ownership. Although data is available for over 177,000 firms in 153 countries, we restrict our attention to the ninety countries in the WBES that are part of the BRI. Of the included BRI countries, twenty-nine BRI exported energy (oil) over our sample period. Our final dataset is a cross section of firms surveyed once between 2011 and 2019. We combine WBES data with energy trade data from the United Nations Commodity Trade Statistics Database (UN Comtrade). We focus on imports under HS codes of Chapter 27: Mineral fuels, mineral oils, and products of their distillation; bituminous substances; mineral waxes. The appeal of this chapter for our analysis is that it contains all the energy inputs such as crude and other petroleum oils, coal, briquettes, liquefied gases, and similar. As such, to the extent that there are competitive pressures in import markets for energy, it should show up in the movements under this chapter in each country.

4.3. Key Variables

All the key variables used in our analyses are described in Table 1. Our dependent variable is innovation. To operationalize this variable, we focus on three questions in WBES which captures innovation output and input. Specifically, WBES asks: (i) Did this establishment introduce onto the market any new or significantly improved products (product innovation = 1 if yes, 0 otherwise); (ii) Has this establishment introduced any new or significantly improved production processes including methods of supplying services and ways of delivering products (process innovation = 1 if yes, 0 otherwise); and (iii) Did this establishment spend on research and development activities, either in-house or contracted with other companies, excluding market research surveys (R&D = 1 if yes, 0 otherwise). About 23.46 percent of the sample are process innovators and 25.61 percent are product innovators.

Table 1.

Definition and Basic Descriptive of Key Variables.

Variable	Definition	Energy exporting countries				Non-energy exporting countries
Variable	Definition	Mean	Std. Dev.	Min	Max	Mean	Std. Dev.	Min	Max
Process	(0/1) Introduced new/significantly improved process in the last three years	0.22	0.41	0	1	0.30	0.46	0	1
Product	(0/1) Introduced new products/services in the last three years	0.23	0.42	0	1	0.33	0.47	0	1
R&D	(0/1) Invested in R&D in the last three years	0.10	0.31	0	1	0.15	0.35	0	1
EC	Firm cost of energy/total operating costs (%)	18.02	22.23	0	100	17.17	22.93	0	100
Labor	Log number of workers	3.17	1.39	0	11	3.16	1.40	0	10
Capital	Log net book value of assets	2.37	5.57	0	28	3.17	6.46	0	33
Age	Age in years	18.08	14.65	1	220	20.05	17.68	1	341
Exporter	(0/1) Exporter of some share of firm output	0.23	0.42	0	1	0.26	0.44	0	1
A2F	(0/1) Purchased fixed asset with a loan in last three years	0.30	0.46	0	1	0.41	0.49	0	1
Website	(0/1) Establishment has its own website	0.53	0.50	0	1	0.48	0.50	0	1
Audit	(0/1) Externally certified financial records annually	0.20	0.40	0	1	0.19	0.39	0	1

Note. The sample size for the variables in energy exporting countries is 36,012 and for non-energy countries it is 39,523.

There is a growing strand of literature that suggests that there is complementarity between different types of innovation (see Lewandowska, SzymuraTyc and Golkebiowski [2016], Doran [2012], and Bianchini, Pellegrino and Tamagni [2018] as examples). These studies point out that while product innovations create a differentiation effect focused on capturing new markets, process innovations are more oriented towards productivity increases and cost reduction. Similarly, Martínez-Ros and Labeaga (2009) argue that the manufacturing of a new product is often possible only if a new technological process is applied and, consequently, firms engaging in process innovation are more likely to engage in product innovation.

To examine if our results are robust to these complementarities, we develop two aggregate measures of innovation (Joint and Separate) that are also binary in nature but reflect the different strategies that firms use to innovate. $J o i n t$ is unity if firms undertake product and process innovation in tandem (0, otherwise) while $S e p a r a t e$ is unity if firms undertake one type of innovation but not the other (0, otherwise). In our sample, 13 percent of the firms were joint innovators and 36 percent undertook product and process innovation separately. Regarding innovation input, 13.67 percent of the sampled firms invest in R&D. These proportions are comparable to the ratios in other recently published innovation studies (see Mohnen and Hall [2013] for a review).

The key independent variable is relative energy cost $(E C)$ . We operationalize this variable as the total annual cost of energy (electricity and fuel) divided by the total operating cost for firm $i$ . The average firm spends 18 percent of total costs on energy in energy exporting countries and 17 percent in non-energy exporting countries. Our analysis posits that this relative cost differs if the firm is operating in an energy exporting country versus a non-energy exporting country. We examine this first in Figure 3, which shows a kernel density plot of relative energy costs for the full sample of firms: one each for firms in oil exporting and non-exporting countries. Firms in oil exporting countries have slightly higher relative energy costs compared to comparator firms in non-exporting countries. The combined Kolmogorov-Smirnov test statistic is 0.2004 with a p-value of 0.062, showing there is a statistical difference between relative energy costs for these two groups of firms.

Figure 3.

Kernel density plot of energy cost.

Control variables used in our regressions include size measured as log of the number of permanent employees (labor) and log net book value of assets (capital), foreign engagement measured as a dummy variable if the firm exports (exporter = 1 if yes, 0 otherwise), the firm’s access to finance measured as dummy variable if the firm used credit from a registered financial institution in the last three years to purchase fixed assets (A2F = 1 if yes, 0 otherwise), use of technology measured as a dummy variable if the firm own a website (website = 1 if yes, 0 otherwise), and a dummy variable if the firm gets its financial records audited by an independent accounting firm (audit = 1 if yes, 0 otherwise). To account for differences in country-industry specialization, we include a battery of country, industry, and time fixed effects and their interactions.

5. Results

Here we present our results after testing the two propositions highlighted above. We first present what we think is a naive approach to assess the relationship between relative energy costs and innovation in Table 2. In this table we run probit regressions consistent with equation (23). We present two panels of results corresponding to two sub-samples of BRI countries. Panel A are results using firms from energy exporting countries only and Panel B are regression results using only firms from countries that have no energy exports. The numbers shown are marginal effects computed at the sample mean with standard errors in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [joint] for column 3 and process or product = 1 [separate] for column 4). Column 5 uses an innovation input measure, R&D. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and audit.

Table 2.

Probit Results of Innovation and Energy Cost.

Dependent variable	(1)Product	(2)Process	(3)Joint	(4)Separate	(5)R&D
Panel A: Energy exporting countries
EC	0.000** (0.000)	0.000 (0.000)	0.000** (0.000)	0.000 (0.000)	−0.000* (0.000)
N $N$	36,012	36,012	36,012	36,012	36,012
Panel B: Non energy exporting countries
EC	0.000** (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
N	39,523	39,523	39,523	39,523	39,523
Controls	✓	✓	✓	✓	✓
Year FE	✓	✓	✓	✓	✓
Industry FE	✓	✓	✓	✓	✓
Region FE	✓	✓	✓	✓	✓
Income FE	✓	✓	✓	✓	✓
Country × Year FE	✓	✓	✓	✓	✓
Industry × Year FE	✓	✓	✓	✓	✓
Country × Industry FE	✓	✓	✓	✓	✓

Note. These are marginal effects computed at the sample mean from different probit regressions. Standard errors are in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [Joint]—for column 3 and product or process = 1 [Separate] for column 4). Column 5 uses an innovation input measure, R&D. Panel A uses firms from energy exporting countries only while Panel B uses firms from non-energy exporting countries. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and audit.

p < 0.1, ** p < 0.05, *** p < 0.01.

From these results, consistent with the induced innovation hypothesis, the coefficient on $E C$ is positive and significant in a few cases but is very small. We consider these results naive because, as we argued above, relative energy cost is an endogenous variable because of potential bi-causality with innovation. Innovation may drive changes in energy costs while energy costs may drive innovation. We present the results in Table 2 here merely to serve as caution that this approach is not sufficient to understand the relationships we seek to analyze.

To address the endogeneity issues, we estimate equation (23) using an instrumental variables approach that instruments relative energy costs with $E D S$ as shown in equations (24) and (25). As such, Table 3 contains our baseline results. These are the second stage results because the regressions in each panel all have the same first stage. The coefficient and standard error for EDS for these first stage regressions are 0.090 (0.008) for Panel A and 0.007 (0.001) for Panel B. The corresponding F statistics are 136.58 and 102.396, respectively. The detailed results for these first stages are presented in the Supplemental Appendix.

Table 3.

Second Stage Results: Innovation and Energy Cost.

Dependent variable	(1)Product	(2)Process	(3)Joint	(4)Separate	(5)R&D
Panel A: Energy exporting countries
EC	0.051*** (0.002)	0.028*** (0.002)	0.037*** (0.002)	0.049*** (0.002)	0.021*** (0.002)
N	36,012	36,012	36,012	36,012	36,012
Panel B: Non energy exporting countries
EC	0.036*** (0.003)	0.003 (0.002)	0.005** (0.002)	0.042*** (0.003)	0.007*** (0.003)
N	39,523	39,523	39,523	39,523	39,523
Control variables	✓	✓	✓	✓	✓
Year FE	✓	✓	✓	✓	✓
Industry FE	✓	✓	✓	✓	✓
Region FE	✓	✓	✓	✓	✓
Income FE	✓	✓	✓	✓	✓
Country × Year FE	✓	✓	✓	✓	✓
Industry × Year FE	✓	✓	✓	✓	✓
Country × Industry FE	✓	✓	✓	✓	✓

Note. These are our baseline results. These are marginal effects computed at the sample mean from different ivprobit regressions. Standard errors are in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [Joint]—for column 3 and product or process = 1 [Separate] for column 4). Column 5 uses an innovation input measure, R&D. Panel A uses firms from energy exporting countries only while Panel B uses firms from non-energy exporting countries. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and audit. The regressions in each panel all have the same first stage where EC is instrumented with EDS from equations (24) and (25). The coefficient and standard error for EDS from these first stage regressions are: Panel A—0.09 (0.008); Panel B—0.007 (0.001). The F statistics for these two sets of results are: Panel A—136.58 and 102.396. The detailed results for these first stages are presented in the Supplemental Appendix.

p < 0.1, ** p < 0.05, *** p < 0.01.

In Table 3, the numbers are marginal effects computed at the sample mean from different probit regressions. Standard errors are in parentheses. Again, we present two panels that differentiate the energy trade of the domestic economy. Panel A uses firms from energy exporting countries only while Panel B uses firms from non-energy exporting countries. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (joint for column 3 and separate for column 4). Column 5 uses an innovation input measure, R&D. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and audit.

These results show that regardless of the measure of innovation, $E C$ is positive and statistically significant in energy exporting countries. Since the numbers in the table are marginal effects, the range of magnitude is from 2.1 to 5.1 percent. This suggests that if we increase the relative energy cost for a firm by 1 percent, we can expect the likelihood of innovation to increase by between 2.1 and 5.1 percent. The impact in non-energy exporting countries (although mostly positive and significant) seems to be driven by product innovation. As shown in Panel B, $E C$ is insignificant in the regression with process innovation only (column 2) but significant in all other cases. The magnitude in these countries is also slightly smaller, ranging from 0.5 to 3.6 percent when significant. This suggests that in countries that do not export energy, a 1 percent increase in the relative cost of energy results in at least a 0.5 percent increase in the likelihood that a firm innovates.

To check for robustness, we adopt three distinct strategies. First, it is widely recognized that, as it regards innovation, the world is highly uneven, both at an international and on an intra-national level (Crowley and McCann 2018). There is also growing evidence that innovation varies across countries based on income. For instance, Morris (2018) points out that innovation in developing economies is likely to be different from innovation in developed countries—more incremental and less radical. Information on the type of innovation (radical or incremental) is not readily available in our dataset and so adopt a next best strategy to examine if there are heterogeneities as it relates to the impact of energy exports on developed and developing countries. Although we include income fixed effects in all our IV estimations that should deal with most of these issues, we cannot be sure that unobserved income effects, peculiar to some countries in this group do not evade these fixed effects.

As shown in the Supplemental Appendix, some of the countries in our dataset are high income countries according to the World Bank’s classification. Therefore, we examine the robustness of our results to this concern, we restrict the sample based on average economic income. We drop all firms from countries that the World Bank categorizes as high-income countries and re-estimate the baseline model. These results are presented in Table 4. Like above, Panel A contains estimates for firms in energy exporting countries and Panel B contains estimates for firms in non-exporting countries. All the regressions from this exercise show that $E C$ has a positive and significant effect on the likelihood of innovation, which is consistent with our baseline model. The magnitudes are slightly higher, ranging from 4.3 to 9.2 percent in energy exporting countries and 1.6 to 3.6 percent in non-energy exporting countries.

Table 4.

Second Stage Results: Innovation and Energy Cost for Developing Countries.

Dependent variable	(1)Product	(2)Process	(3)Joint	(4)Separate	(5)R&D
Panel A: Energy exporting countries
EC	0.092*** (0.003)	0.054*** (0.003)	0.073*** (0.003)	0.085*** (0.003)	0.043*** (0.003)
$N$ N	31,241	31,241	31,241	31,241	31,241
Panel B: Non energy exporting countries
EC	0.035*** (0.003)	0.022*** (0.003)	0.021*** (0.003)	0.041*** (0.003)	0.016*** (0.003)
$N$ N	31,938	31,938	31,938	31,938	31,938
Controls	✓	✓	✓	✓	✓
Year FE	✓	✓	✓	✓	✓
Industry FE	✓	✓	✓	✓	✓
Region FE	✓	✓	✓	✓	✓
Income FE	✓	✓	✓	✓	✓
Country × Year FE	✓	✓	✓	✓	✓
Industry × Year FE	✓	✓	✓	✓	✓
Country × Industry FE	✓	✓	✓	✓	✓

Note. These results are firms in developing countries only. These are marginal effects computed at the sample mean from different ivprobit regressions. Standard errors are in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [Joint]—for column 3 and product or process = 1 [Separate] for column 4). Column 5 uses our only innovation input measure, R&D. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and audit. The regressions in each panel all have the same first stage where EC is instrumented with EDS from equations (24) and (25). The coefficient and standard error for EDS from these first stage regressions are: Panel A—0.036 (0.008); Panel B—−0.018 (0.004). The F statistics for these two sets of results are: Panel A—101.995 and 79.197. The detailed results for these first stages are presented in the Supplemental Appendix.

p < 0.1, ** p < 0.05, *** p < 0.01.

Our second robustness check is to check if our results are overly influenced by the subset of firms in our sample that are in the services sector. It is widely reported that firms in the services sector tend to undertake product innovation at a much lower propensity compared to manufacturing firms and focus mainly on process innovation (Alvarez, Bravo-Ortega and Zahler 2015). As such, we remove the firms in the services sector, re-estimate our baseline model and present the results in Table 5. These results are broadly in line with our baseline results and suggest that increases in energy costs result in increased innovation likelihood.

Table 5.

Second Stage Results: Innovation and Energy Cost for Manufacturing Firms.

Dependent variable	(1)Product	(2)Process	(3)Joint	(4)Separate	(5)R&D
Panel A: Energy exporting countries
EC	0.054*** (0.003)	0.025*** (0.002)	0.038*** (0.002)	0.046*** (0.003)	0.013*** (0.003)
$N$ N	19,454	19,454	19,454	19,454	19,454
Panel B: Non energy exporting countries
EC	0.052*** (0.003)	0.018*** (0.003)	0.025*** (0.003)	0.052*** (0.003)	0.016*** (0.003)
$N$ N	19,464	19,464	19,464	19,464	19,464
Controls	✓	✓	✓	✓	✓
Year FE	✓	✓	✓	✓	✓
Industry FE	✓	✓	✓	✓	✓
Region FE	✓	✓	✓	✓	✓
Income FE	✓	✓	✓	✓	✓
Country × Year FE	✓	✓	✓	✓	✓
Industry × Year FE	✓	✓	✓	✓	✓
Country × Industry FE	✓	✓	✓	✓	✓

Note. These results are for manufacturing firms only. These are marginal effects computed at the sample mean from different ivprobit regressions. Standard errors are in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [Joint]—for column 3 and product or process = 1 [Separate] for column 4). Column 5 uses our only innovation input measure, R&D. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and financial accuracy. The regressions in each panel all have the same first stage where EC is instrumented with EDS from equations (24) and (25). The coefficient and standard error for EDS from these first stage regressions are: Panel A—0.099 (0.011); Panel B—0.007 (0.001). The F statistics for these two sets of results are: Panel A—102.391 and 68.316. The detailed results for these first stages are presented in the Supplemental Appendix.

p < 0.1, ** p < 0.05, *** p < 0.01.

A skeptic of our results so far may argue that our empirical approach could be more succinctly summarized into a single equation with an indicator variable used to signal if firm $i$ is in an energy exporting country or not. Moreover, the China shock employed in the existing literature is specific to exports of manufacturing goods from China. Since we want to capture the energy demand inducing effect of this manufacturing exports growth, we follow Autor, Dorn and Hanson (2013, 2016) and use the change in Chinese exports to the nine largest OECD countries—Australia, Denmark, Finland, Germany, Japan, New Zealand, Spain, Switzerland, and the United States of America—as a third robustness check. For this approach, we use the change in industry level goods imports in these countries to compute our $E D S$ measure. So that:

Δ E D S_{i j c} = \sum_{j = 1}^{J} R_{i, - j} \times Δ E X^{C H - O E C D 9}

(27)

where $E X^{C H - O E C D 9}$ is the change in exports from China to the nine OECD countries between 1992 and 2007. With this new measure for the competitive pressure, we re-estimate our baseline regression but as a full sample (combining energy exporting countries and non-energy exporting countries) and insert a control variable (Oil Exporter) to differentiate the two groups of countries. These results are presented in Table 6. Like above, this strategy does not materially change the intuition from our baseline model. These results show that an increase in energy costs leads to an increased chance that firms will innovate. Interestingly the oil exporter dummy is negative and highly significant, confirming our earlier argument that there is a marked difference in the innovation profile for firms in these two broad groups of countries.

Table 6.

Second Stage Results: Innovation and Energy Cost with Alternative Instrument.

Dependent variable	(1)Product	(2)Process	(3)Joint	(4)Separate	(5)R&D
EC	0.080*** (0.002)	0.044*** (0.002)	0.056*** (0.002)	0.076*** (0.002)	0.035*** (0.002)
N	75,535	75,535	75,535	75,535	75,535
Controls	✓	✓	✓	✓	✓
Year FE	✓	✓	✓	✓	✓
Industry FE	✓	✓	✓	✓	✓
Region FE	✓	✓	✓	✓	✓
Income FE	✓	✓	✓	✓	✓
Country × Year FE	✓	✓	✓	✓	✓
Industry × Year FE	✓	✓	✓	✓	✓
Country × Industry FE	✓	✓	✓	✓	✓

Note. These results use the change in Chinese goods exports to the nine largest OECD economies to capture the China shock. These are marginal effects computed at the sample mean from different ivprobit regressions. Standard errors are in parentheses. The dependent variable in columns 1 and 2 are process and product innovation respectively. Column 3 and 4 uses aggregate measures of innovation strategy (process & product = 1 [Joint]—for column 3 and product or process = 1 [Separate] for column 4). Column 5 uses our only innovation input measure, R&D. All regressions control for firm characteristics: labor, capital, age, exporting status, access to finance, technology usage, and financial accuracy. The regressions all have the same first stage. The coefficient and standard error from this first stage regression is −0.0.083 (0.005) and the F statistic is 173.088 as shown in the detailed results in Supplemental Appendix.

p < 0.1, ** p < 0.05, *** p < 0.01.

6. Discussion

Our results suggest that rising relative energy costs induces firms to innovate. This is in line with the predictions from the induced innovation hypothesis which suggests that changes in factor prices can influence firms’ incentives to invest in innovation (see Triguero, Moreno-Mondéjar and Davia [2014] and Newell [2010] as examples). Our findings underscore the need for governments to strike a balance between economic objectives and broader societal goals. Policies that promote innovation and factor price adjustments should be designed in a way that considers social and environmental impacts. For instance, governments can align innovation incentives with sustainability goals to encourage the development and adoption of eco-friendly innovation and incentivize such activities.

As such, there are two dimensions to our results that need to be considered for policy development. First, innovation is risky, firm specific, and a firm’s effectiveness in bringing it to fruition is most times inadequately monitored (Cirera and Muzi 2016). There is a time lag between a firm’s decision to invest in innovation, the associated innovation investment outlays, and the resulting innovation success. This period is fraught with risks for the firm, compounded by challenges related to access to finance (Morris 2018). Therefore, governments should consider implementing policies that incentivize private sector R&D spending, such as tax credits, grants, or public-private partnerships. Moreover, they can provide financial support and create a favorable innovation ecosystem to stimulate firms to invest in R&D activities, leading to the development of new technologies and the accumulation of knowledge. Furthermore, governments can play a crucial role in facilitating the transfer and diffusion of innovation across industries and sectors by implementing policies that promote technology spillover, knowledge exchange, and collaboration between academia, industry, and research institutions.

Second, it is worth noting that rapidly rising energy costs can pose challenges to firms, especially those that are energy-intensive or rely heavily on energy inputs. There is a welfare argument that governments should provide support to these firms to maintain economic stability, preserve jobs, and sustain competitiveness in domestic and international markets. This kind of support can help to mitigate the negative impacts of rising energy prices, allowing firms to adapt and remain viable but our results suggests that careful economic analyses of the impact of these support programs are warranted before governments engage in these stabilization policies. Governments should recognize that policies affecting factor prices, such as taxes, subsidies, or regulations, can have significant implications for the direction and pace of technological change. By carefully evaluating the intended impacts of support programs, governments can encourage the development and adoption of technologies that align with broader societal goals, such as environmental sustainability or productivity improvements.

In addition, although not directly examined in our analysis, to cope with rising relative energy costs, governments may support firms in developing domestic energy resources, diversifying energy supply, and promoting energy independence. This support can contribute to enhanced energy security, reduced price volatility, and a more reliable energy infrastructure.

7. Conclusion

We test the induced innovation hypothesis in this paper by developing a theoretical framework that shows that the relationship between innovation and the relative cost of energy is different in countries depending on whether they export energy or not. We test the propositions from this theoretical framework by focusing on countries that signed a memorandum of understanding with China as part of the BRI. We instrument a firms relative energy cost with a measure of competitive pressure posed by rising energy demand shock in other countries. To operationalize a demand shock that raises energy prices, in our empirical analyses we use an instrumental variables approach that exploits China’s unexpected and drastic rise in energy demand to fuel its goods exports since the early 1990s (the China shock).

Overall, we find that regardless of the measure of innovation (product innovation, process innovation, R&D), increasing energy costs induces firms to innovate. In terms of magnitude, we find that a 1 percent increase in the cost of energy relative to other firm costs increases the likelihood of innovation by between 2.1 and 5.1 percent in energy exporting countries. The impact in non-energy exporting countries is slightly smaller, ranging from 0.5 to 3.6 percent. This suggests that in countries that do not export energy, a 1 percent increase in the relative cost of energy results in at least a 0.5 percent increase in the likelihood that a firm innovates. These results are robust to various methodological variations and data restriction exercises.

Our analysis is conditioned on some data limitations that are worth mentioning. The dataset is an uneven cross section and so firm level heterogeneity may not be completely accounted for. Adopting a similar approach to the one employed in this paper using panel datasets would help to assess the relative generalizability of our findings. A panel dataset could also help with introducing a dynamic element to the assessment and allow for a more targeted approach to control for cyclical effects. We know from the existing literature that there is a lagged effect of investments in innovation on firm level outcomes and so these dynamics would not only allow for an explicit examination of issues related to persistence of innovation but also of the energy demand shocks we study in the paper. Also, there may be some difference in results when a continuous measure of innovation is used in the estimations versus when this measure is binary. For this approach to be meaningful, there is a need for better data on the proportion of firm output attributable to an innovative activity. These are all avenues for future research.

Supplemental Material

sj-docx-1-enj-10.1177_01956574241266969 – Supplemental material for Innovation and Spillover Effects of Energy Demand Shocks in Belt and Road Economies

Supplemental material, sj-docx-1-enj-10.1177_01956574241266969 for Innovation and Spillover Effects of Energy Demand Shocks in Belt and Road Economies by King Yoong Lim and Diego Morris in The Energy Journal

Supplemental Material

sj-pdf-2-enj-10.1177_01956574241266969 – Supplemental material for Innovation and Spillover Effects of Energy Demand Shocks in Belt and Road Economies

Supplemental material, sj-pdf-2-enj-10.1177_01956574241266969 for Innovation and Spillover Effects of Energy Demand Shocks in Belt and Road Economies by King Yoong Lim and Diego Morris in The Energy Journal

Footnotes

Acknowledgements

We thank the participants at the 41st Eurasia Business and Economics Society (EBES) Conference and the 28th Review of Socio-Economic Perspectives (RSEP) International Conference on Economics, Finance & Business. We are also extremely thankful to our respective departmental colleagues for their invaluable feedback and comments on an earlier draft of the paper. We are also grateful to the editor and three anonymous referees for their feedback and suggestions to improve the paper. The views expressed are our own.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online.

1

According to the Organization for Economic Co-operation and Development (OECD) Oslo manual, innovation can be divided into technological innovations in the form of new products and services and non-technological innovations in the form of organizational or marketing changes ().

2

The full list of countries that we include, the year they joined the BRI and the specific income grouping they belong according to the World Bank are included in Supplemental Appendix B. This initiative was started by President Xi Jinping to improve China’s connection with the rest of the world.

3

Economic income group fixed effects and geographic region fixed effects are based on the World Bank’s income and regional grouping of countries. This categorization of income and region is shown in the Supplemental Appendix.

References

Alvarez

Roberto

Bravo-Ortega

Claudio

Zahler

Andrés

. 2015. “Innovation and Productivity in Services: Evidence from Chile.” Emerging Markets Finance and Trade 51 (3): 593–611.

Annicchiarico

Barbara

Dio

Fabio Di

. 2017. “GHG Emissions Control and Monetary Policy.” Environmental & Resource Economics 67 (4): 823–51.

Audretsch

David B.

Feldman

Maryann P.

1996. “R&D Spillovers and the Geography of Innovation and Production.” The American Economic Review 86 (3): 630–40.

Audretsch

David B.

Feldman

Maryann P.

2004. “Knowledge Spillovers and the Geography of Innovation.” In Handbook of Regional and Urban Economics, edited by J. V. Henderson and J.-F. Thisse, Vol. 4, 2713–39. Amsterdam: Elsevier.

Autor

David H.

Dorn

David

Hanson

Gordon H.

2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review 103 (6): 2121–68.

Autor

David H.

Dorn

David

Hanson

Gordon H.

2016. “The China Shock: Learning from Labor-Market Adjustment to Large Changes in Trade.” Annual Review of Economics 8: 205–240.

Badeeb

Ramez Abubakr

Lean

Hooi Hooi

Smyth

Russell

. 2016. “Oil Curse and Finance–Growth Nexus in Malaysia: The Role of Investment.” Energy Economics 57: 154–65.

Baniya

Suprabha

Rocha

Nadia

Ruta

Michele

. 2020. “Trade Effects of the New Silk Road: A Gravity Analysis.” Journal of Development Economics 146: 102467.

Bianchini

Stefano

Pellegrino

Gabriele

Tamagni

Federico

. 2018. “Innovation Complementarities and Firm Growth.” Industrial and Corporate Change 27 (4): 657–76.

10.

Brambilla

Irene

. 2009. “Multinationals, Technology, and the Introduction of Varieties of Goods.” Journal of International Economics 79 (1): 89–101.

11.

Brambilla

Irene

Hale

Galina

Long

Cheryl

. 2009. “Foreign Direct Investment and the Incentives to Innovate and Imitate.” The Scandinavian Journal of Economics 111 (4): 835–61.

12.

Chuku

Lang

Lin

Lim

King Yoong

. 2023. “Public Debt, Chinese Loans and Optimal Exploration–Extraction in Africa.” Energy Economics 118: 106516.

13.

Cirera

Xavier

Muzi

Silvia

. 2016. “Measuring Firm-Level Innovation Using Short Questionnaires: Evidence from an Experiment.” World Bank Policy Research Working Paper No. 7696. The World Bank Group.

14.

Crabb

Joseph M.

Johnson

Daniel K. N.

2010. “Fueling Innovation: The Impact of Oil Prices and CAFE Standards on Energy-Efficient Automotive Technology.” The Energy Journal 31 (1): 199–216.

15.

Crespi

Gustavo

Tacsir

Ezequiel

Vargas

Fernando

. 2016. “Innovation Dynamics and Productivity: Evidence for Latin America.” In Firm Innovation and Productivity in Latin America and the Caribbean, edited by M.

Grazzi

Pietrobelli

, 37–71. New York, NY: Palgrave Macmillan.

16.

Crowley

Frank

McCann

Philip

. 2018. “Firm Innovation and Productivity in Europe: Evidence from Innovation-Driven and Transition-Driven Economies.” Applied Economics 50 (11): 1203–221.

17.

Doran

Justin

. 2012. “Are Differing Forms of Innovation Complements or Substitutes?” European Journal of Innovation Management 15 (3): 351–71.

18.

Duggal

Vijaya G.

Saltzman

Cynthia

Klein

Lawrence R.

1999. “Infrastructure and Productivity: A Nonlinear Approach.” Journal of Econometrics 92 (1): 47–74.

19.

Duranton

Puga

(2004). “Micro-Foundations of Urban Agglomeration Economies.” In Handbook of Regional and Urban Economics, edited by J. V. Henderson and J.-F. Thisse, Vol. 4, 2063–2117. Amsterdam: Elsevier.

20.

Fang

Zheng

Chen

Yang

. 2017. “Human Capital and Energy in Economic Growth–Evidence from Chinese Provincial Data.” Energy Economics 68: 340–58.

21.

Gauvin

Ludovic

Rebillard

Cyril C.

2018. “Towards Recoupling? Assessing the Global Impact of a Chinese Hard Landing Through Trade and Commodity Price Channels.” The World Economy 41 (12): 3379–415.

22.

Glaeser

Edward L.

Gottlieb

Joshua D.

2009. “The Wealth of Cities: Agglomeration Economies and Spatial Equilibrium in the United States.” Journal of Economic Literature 47 (4): 983–1028.

23.

Hallak

Juan Carlos

Sivadasan

Jagadeesh

. 2013. “Product and Process Productivity: Implications for Quality Choice and Conditional Exporter Premia.” Journal of International Economics 91 (1): 53–67.

24.

Hicks

J. R.

1932. The Theory of Wages. London: Macmillan.

25.

Huang

Yiping

. 2016. “Understanding China’s Belt & Road Initiative: Motivation, Framework and Assessment.” China Economic Review 40: 314–21.

26.

Johnston

Lauren A.

2019. “The Belt and Road Initiative: What Is in It for China?” Asia & the Pacific Policy Studies 6 (1): 40–58.

27.

Johnstone

Nick

Haščič

Ivan

Popp

David

. 2010. “Renewable Energy Policies and Technological Innovation: Evidence Based on Patent Counts.” Environmental and Resource Economics 45 (1): 133–55.

28.

Lewandowska

Małgorzata Stefania

SzymuraTyc

Maja

Golkebiowski

Tomasz

. 2016. “Innovation Complementarity, Cooperation Partners, and New Product Export: Evidence from Poland.” Journal of Business Research 69 (9): 3673–81.

29.

Lim

King Yoong

Morris

Diego

. 2022. “Thresholds in Natural Resource Rents and State Owned Enterprise Profitability: Cross Country Evidence.” Energy Economics 106: 105779.

30.

Lim

King Yoong

Zhang

Shuonan

. 2023.” Optimal Fiscal Management in an Economy with Resource Revenue-Financed Government-Linked Companies.” International Journal of Finance & Economics 28 (2): 2202–225.

31.

Lin

Justin Yifu

Wang

Yan

. 2017. Going Beyond Aid: Development Cooperation for Structural Transformation. Cambridge: Cambridge University Press.

32.

Marshall

Guillermo

Parra

Alvaro

. 2019. “Innovation and Competition: The Role of the Product Market.” International Journal of Industrial Organization 65: 221–47.

33.

Martínez-Ros

Ester

Labeaga

Jose M.

2009. “Product and Process Innovation: Persistence and Complementarities.” European Management Review 6 (1): 64–75.

34.

Mody

Cyrus C. M.

2023. “Spillovers from Oil Firms to US Computing and Semiconductor Manufacturing: Smudging State–Industry Distinctions and Retelling Conventional Narratives.” Enterprise & Society 24 (3): 676–701.

35.

Mohnen

Pierre

Hall

Bronwyn H.

2013. “Innovation and Productivity: An Update.” Eurasian Business Review 3 (1): 47–65.

36.

Morris

Diego M.

2018. “Innovation and Productivity Among Heterogeneous Firms.” Research Policy 47 (10): 1918–32.

37.

Newell

Richard G.

2010. “The Role of Markets and Policies in Delivering Innovation for Climate Change Mitigation.” Oxford Review of Economic Policy 26 (2): 253–69.

38.

Newell

Richard G.

Jaffe

Adam B.

Stavins

Robert N.

1999. “The Induced Innovation Hypothesis and Energy-Saving Technological Change.” The Quarterly Journal of Economics 114 (3): 941–75.

39.

OECD/Eurostat. 2018. Oslo Manual 2018: Guidelines for Collecting, Reporting and Using Data on Innovation, Vol. 4. Paris: OECD Publishing; Luxembourg: Eurostat.

40.

Okullo

Samuel J.

Reynès

Frédéric

Hofkes

Marjan W.

2015. “Modeling Peak Oil and the Geological Constraints on Oil Production.” Resource and Energy Economics 40: 36–56.

41.

Pindyck

Robert S.

1978. “The Optimal Exploration and Production of Nonrenewable Resources.” Journal of Political Economy 86 (5): 841–61.

42.

Popp

David

. 2002. “Induced Innovation and Energy Prices.” American Economic Review 92 (1): 160–80.

43.

Romp

Ward E.

De Haan

Jakob

. 2005. “Public Capital and Economic Growth: A Critical Survey.” EIB Papers 10 (1): 41–70.

44.

Tether

Bruce S.

2005. “Do Services Innovate (Differently)? Insights from the European Innobarometer Survey.” Industry & Innovation 12 (2): 153–84.

45.

Triguero

Angela

Moreno-Mondéjar

Lourdes

Davia

María A.

2014. “The Influence of Energy Prices on Adoption of Clean Technologies and Recycling: Evidence from European SMEs.” Energy Economics 46: 246–57.

46.

Utar

Hale

. 2014. “When the Floodgates Open: “Northern” Firms’ Response to Removal of Trade Quotas on Chinese Goods.” American Economic Journal: Applied Economics 6 (4): 226–50.

47.

Utar

Hale

Torres Ruiz

Luis B.

2013. “International Competition and Industrial Evolution: Evidence from the Impact of Chinese Competition on Mexican Maquiladoras.” Journal of Development Economics 105: 267–87.

48.

Wurlod

Jules-Daniel

Noailly

Joëlle

. 2018. “The Impact of Green Innovation on Energy Intensity: An Empirical Analysis for 14 Industrial Sectors in OECD Countries.” Energy Economics 71: 47–61.

49.

Zahler

Andrés

Iacovone

Leonardo

Mattoo

Aaditya

. 2014. “Trade and Innovation in Services: Evidence from a Developing Economy.” The World Economy 37 (7): 953–79.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.30 MB