2.1 Steady state with fixed prices

We follow Blanchard’s (1980) treatment (see p. 9) and consider the simple case of fixed prices for illustrative and motivational purposes. Hence, there is no actual and no expected inflation; nominal and real rates are identical, and the system simplifies. In that steady state, we have from equation (12), $\dot{E}=0$ :

(15) \begin{equation} r=i={i}^\ast -\theta \end{equation}

The domestic real interest rate, which is equal to the domestic nominal interest rate, depends on the world’s nominal interest rate and risk premia. Substituting (15) into (12) yields

(16) \begin{equation}{i}^\ast -\theta =cy-h\left (m-p\right ) \end{equation}

From the IS curve, equation (11), with $\dot{y}=0$ , we have

(17) \begin{equation} y=b^{-1} \biggl [aq+g+NX\biggl ({y}^\ast, e\frac{P}{{P}^\ast }\biggr )\biggr ] \end{equation}

Output depends on the stock market $q$ , the balance of trade $NX$ , and fiscal policy $g$ . Note that since $NX$ depends on the real exchange rate $E=e P/{P}^\ast$ , so too does output.

In the steady state, the stock market is given by introducing equations (15) into equation (14) and imposing $\dot{q}=0$ :

(18) \begin{equation} q=\cfrac{\aleph \left (V\right )\left [\pi \left (K\right )- A\tau _{_k}\right ]+w\left (V\right )h\left (\mathcal{R}\right )\beth _k(k,A)}{{i}^\ast -\theta } \end{equation}

In the steady state, accordingly, the stock market equals the ratio of profit to the steady-state real interest rate. The Green Tobin’s $q$ depends on variables associated with the development of green technologies and cleaning efforts such as the stock of knowledge $V$ , pollution abatement efforts $A$ , and R&D expenditures $\mathcal{R}$ . This is what makes the Green Tobin’s q, green.

By substituting (16) into (18), we write the Green Tobin’s $q$ as

(19) \begin{equation} q=\frac{\aleph \left (V\right )\left [\pi \left (K\right )- A\tau _{_k}\right ]+w\left (V\right )h\left (\mathcal{R}\right )\beth _k(k,A)}{cy-h\left (m-p\right )} \end{equation}

Overall, thus, through equation (11), $r$ increases with transactions demand for money, while through equation (8), profits grow in line with output. Since we are dealing with an open economy model, and output depends on the real exchange rate, $E$ , then $E$ also impacts the stock market.

As we have two unknowns $y$ and $q$ for two equations (17) and (19), solving them yields the steady-state equilibrium values of $y$ and $q$ :Footnote ⁷

(20) \begin{equation} y_{ss}=\cfrac{bh\left (m-p\right )+\sqrt{\left [bh\left (m-p\right )\right ]^2+4abc(\aleph \left (V\right )\left [\pi \left (K\right )- A\tau _{_k}\right ]+w\left (V\right )h\left (\mathcal{R}\right )\beth _k(k,A)) }}{2bc} \end{equation}

(21) \begin{equation} q_{ss} =\frac{\aleph \left (V\right )\left [\pi \left (K\right )- A\tau _{_k}\right ]+w\left (V\right )h\left (\mathcal{R}\right )\beth _k(k,A)} {{\begin{array}{l}\left (2b\right )^{-1} \big[bh\left (m-p\right ) \\+\sqrt{\left [bh\left (m-p\right )\right ]^2+4abc\bigr (\aleph \left (V\right )\left [\pi \left (K\right )- A\tau _{_k}\right ]+w\left (V\right )h\left (\mathcal{R}\right )\beth _k(k,A)\bigr )} \big]-h\left (m-p\right )\end{array}}} \end{equation}

Thus, the equilibrium values of real output and the Green Tobin’s $q$ are a function of monetary balances (and thus directly changes in monetary policy) and profits $\varphi ({\cdot})$ and the determinants of profits.

The system (10)–(14), linearized around the steady-state equilibrium of conditions (20) and (21), yields the following Jacobian:

\begin{align*} J=\left[\begin {array}{l@{\quad}l} \frac {\partial \dot {y}}{\partial y}& \frac {\partial \dot {y}}{\partial q}\\[3pt] \frac {\partial \dot {q}}{\partial y} & \frac {\partial \dot {q}}{\partial q} \end{array}\right]_{y_{ss},q_{ss}}=\left [\begin {array}{c@{\quad}c} -b\sigma & a\sigma \\[3pt] cq_{ss}& cy_{ss}-h(m-p) \end{array}\right] \end{align*}

As long as the domestic real interest rate is positive in the steady state,Footnote ⁸ then $cy_{ss}\gt h(m-p)$ , and the determinant of the Jacobian is negative, and the steady-state equilibrium is a saddle point.

2.2 Discussion

The model presented above, though simple in motivation, becomes quite involved in exposition and would also clearly require a careful parametrization. However, the key mechanisms are readily apparent. When seen through equations (20) and (21), the steady-state equilibrium output and stock market valuation depend critically on changes in monetary policy and changes in green technology development (i.e., new patents) as well as open-economy variables such as the world interest rate ( $i^\star$ ), risk premia ( $\theta$ ), the trade balance ( $NX$ ), on the real exchange rate ( $e$ ),Footnote ⁹ and finally also on monetary and (implicitly) fiscal policies.

Accordingly, in our later Section 4, we assess how changes in US monetary policy and US patent counts affect international stock market valuations. We do so using an estimated global econometric model. Before coming to those results, we describe the underlying methodology that we use.

3.1 Data

Table 1 presents the variables, definitions, and sources. We collected most variables from the GVAR database of Mohaddes and Raissi (Reference Mohaddes and Raissi2020). Consistent within the small model of the previous section, we use quarterly real GDP ( $y$ ), changes in the consumer price index ( $cpi$ ), the real exchange rate ( $e$ ), the short-term and long-term interest rates ( $r$ and $l$ , respectively), and the real stock market value ( $q$ ).

Table 1.

Variables and sources

Notes: Data are in natural logs. The short and long rates are computed as $ln(1+R^x_{it})/4$ , where $R^x, x=\{r,l\}$ is the nominal rate of interest per annum, in percent.

We use the proportion of patents on environmental technologies to total patents to represent green patents (labeled $green$ ). We collected this data from the Organization for Economic Cooperation and Development (OECD) Green Growth Indicators.Footnote ¹⁰ Because the green patent series is annual, we used the Denton procedure to change to a quarterly frequency. Green patents represent our proxy for green innovation. A green patent denotes the efforts of firms to create new green innovations to make their production and organization more efficient and compatible with clean energy. There is also the consideration that green patents can represent a cost for non-holders of these patents, particularly firms with poor financial resources. However, in the long term, when the monopoly granted by the patent expires, green innovation becomes available to all firms. In the appendix, Table A.1 presents the full descriptive statistics of the variables. Specifically, we collected quarterly data for the G7 (Canada, France, Germany, Italy, Japan, the UK, and the USA) from 1980q1 to 2019q4.

Moreover, Keller (Reference Keller2001) examined the diffusion of technology among these seven most industrialized economies (see also Growiec, Reference Growiec2012). That study explored three potential channels of technology diffusion: international trade, foreign direct investment, and communication. The estimates indicated that international trade is one of the main drivers to explain technology spillover effects. Our econometric model exploits this finding by employing bilateral trade to connect the economies and transmit the US green patent shock.Footnote ¹¹ Note, we use the economy-wide $q$ measure since (a) there is no aggregate and internationally comparable green-specific measures and (b) our interest is precisely in looking at total stock market performance within and across countries reflecting green and brown industries.

Moreover, Figure 1 shows the share of green patents to total patents over 2000–2019 for the G7. We see that France, Italy, and the UK almost doubled the share of green patents since 2000, reflecting a concerted effort to increase the production of green technologies. Germany shows the highest propensity to develop green patents, while Italy and the USA are among the weakest. Notably, the upward trajectory of green patenting activities stalled after 2008. This likely reflected the effect of the Great Recession and perhaps some reduction of appetite for green investment given some assessment of the performance of such funds (e.g., Master and Temple-West, Reference Masters and Temple-West2023; Shifflett, Reference Shifflett2023).

Figure 1.

Environmental technology patents (% of total patents, G7). Notes: This figure shows for the G7 countries the percentage of total patents registered that are labeled environmental technology patents by the OECD methodology, over 2000 to 2019. Source: “Green Growth Indicators,” the OECD.

3.2 Methodology

We follow Pesaran et al. (Reference Pesaran, Schuermann and Weiner2004) to describe the GVAR framework. Equation (22) presents a $VARX(1,1)$ with one lag for the vector of domestic variables and one lag for the vector of (exogenous) foreign variables. The left side of equation (22) shows the vector of domestic variables of the region $i$ at time $t$ , $x_{it}$ . On the right side, $a_{i0}$ is the constant of the region $i$ ; $a_{i1}$ is the parameter of the trend term; $x_{i,t-1}$ is the vector of domestic variables lagged one period; $x_{it}^{\ast }$ and $x_{i,t-1}^{\ast }$ are, respectively, the vector of current and lagged foreign variables; and $\varepsilon _{it}$ is the vector of idiosyncratic errors:

(22) \begin{equation} x_{it}=a_{i0}+a_{i1}t+\mathrm{\Phi }_i x_{i,t-1}+\mathrm{\Lambda }_{i0}x_{it}^{\ast }+\mathrm{\Lambda }_{i1}x_{i,t-1}^{\ast }+\varepsilon _{it} \end{equation}

We connect each region using foreign variables:

(23) \begin{equation} x^{\ast }_{it}= \sum _{j=0}^N w_{ij}x_{jt} \end{equation}

The vector of foreign variables of region $i$ , $x_{it}^{\ast }$ , uses bilateral trade between regions $i$ and $j$ , $w_{ij}$ , and the variables of the region $j$ , $x_{jt}$ . Foreign variables essentially capture the vulnerability of different regions to external shocks, spillovers, and international fluctuations. Most GVAR studies use bilateral trade to connect regions (Dees et al., Reference Dees, Mauro, Pesaran and Smith2007; Attílio et al., Reference Attílio, Faria and Rodrigues2023). We use bilateral trade and bilateral financial flows in the empirical section to test the robustness of the results (see below).

We create the vectors $z_{it}=(x_{it},x_{it}^{\ast })^{\prime}$ and $x_t=(x_{0t}^{\prime},x_{1t}^{\prime},x_{2t}^{\prime},x_{3t}^{\prime},\dots, x_{Nt}^{\prime})^{\prime}$ and the identity $z_{it}\equiv W_i x_t$ and use them to rewrite equation (22). In equation (24), $z_{it}$ is a vector with domestic and foreign variables, $x_t$ is a vector with all domestic variables of the model, and $W_i$ is the link matrix, with some zero elements and shares of bilateral trade.

(24) \begin{equation} A_iW_ix_t=a_{i0}+a_{i1}t+B_iW_ix_{t-1}+\varepsilon _{it} \end{equation}

where $A_i=[I_{k_i},- \Lambda _{i0}]$ , $I$ is the identity matrix, and $B_i=[\Phi _i,\Lambda _{i1}]$ . We can stack equation (24) as follows:

(25) \begin{equation} Gx_t=a_0+a_1t+Hx_{t-1}+\varepsilon _t \end{equation}

where

\begin{align*} a_0=\begin{pmatrix} a_{00} \\ a_{10} \\ \dots \\ a_{N0} \\ \end{pmatrix}, a_1=\begin{pmatrix} a_{01} \\ a_{11} \\ \dots \\ a_{N1} \\ \end{pmatrix}, \varepsilon _t=\begin{pmatrix} \varepsilon _{0t} \\ \varepsilon _{1t} \\ \dots \\ \varepsilon _{Nt} \\ \end{pmatrix}, G=\begin{pmatrix} A_0 W_{0} \\ A_1 W_{1} \\ \dots \\ A_N W_{N} \\ \end{pmatrix}, H=\begin{pmatrix} B_0 W_{0}\\ B_1 W_{1} \\ \dots \\ B_N W_{N} \\ \end{pmatrix} \end{align*}

We obtain the GVAR by using the inverse matrix of $G$ in equation (25):

(26) \begin{equation} x_t=G^{-1}a_0+ G^{-1}a_1 t + G^{-1}H x_{t-1} + G^{-1}\varepsilon _t \end{equation}

Equations (27) and (27.1) below present the vectors of domestic and foreign variables. Following Dees et al. (Reference Dees, Mauro, Pesaran and Smith2007), we do not include the exchange rate in the US model. However, we only include the exchange rate as a foreign variable in the US model. These equations show that the international stock market ( $q^{\ast }$ ) and international credit markets $(r^{\ast })$ affect the domestic dynamics of the economies.

(27) \begin{align} x_{it}&=(y_{it},q_{it},{cpi}_{it},e_{it},r_{it},l_{it},{green}_{it}) \nonumber \\ x_{it}^{\ast }&=(q_{it}^{\ast },r_{it}^{\ast }) \quad \text{for all regions, except the US} \end{align}

(27.1) \begin{align} x_{it}& = (y_{it},q_{it},{cpi}_{it},e_{it},r_{it},l_{it},{green}_{it})\nonumber\\ x_{it}^{\ast }&=(e^{\ast }_{it}) \quad \,\text{for the US} \end{align}

Equations (27) illustrate one relevant difference with our econometric approach compared to VAR/VEC/SVAR models (for an overview of the VAR methodology, see Kilian and Lütkepohl, Reference Kilian and Lütkepohl2018). In these models, there are no foreign variables built as in equation (23) (i.e., the vectors $x_{it}^\ast$ ). In the GVAR, on the other hand, we use these vectors to connect economies and model linkages of economic integration. In equation (27), the foreign variables, the stock market and the interest rate ( $q_{it}^\ast, r_{it}^\ast$ ), are exogenous to domestic economies, implying restrictions on individual economies (we test this hypothesis using the weak-exogeneity test, see Tables A.4 and A.5 in Section A). In other words, when we analyze IRFs, these variables are treated as given for economies; economies adjust to shocks according to international financial markets ( $q_{it}^\ast, r_{it}^\ast$ ).

3.3 Impulse response analysis

In standard VAR models, the IRF analysis adopts the orthogonalized impulse responses, which use orthogonal shocks instead of the original shock (or residual associated with a particular equation); see, for example, Kilian and Lütkepohl (Reference Kilian and Lütkepohl2018).

In the GVAR, we use a generalized form of the IRF (GIRF). This uses the original shock. We assume that this original shock follows a multivariate normal distribution. GIRFs allow contemporaneous dependence between the error terms (the elements of the idiosyncratic vector) between regions $i$ and $j$ , where the covariance between the errors of region $i$ and region $j$ tends to be zero as the number of observations goes to infinity. Under conditions of stability, smallness, and weak dependence, the GIRFs consider that the covariance between the foreign variables and the error term tends to be zero. Finally, given that we use the GVAR in the error correction form, the shocks imposed are permanent in nature and thus lead to permanent level differences in the response variables.

A variant of the GIRF is the structural GIRF (SGIRF), which imposes a particular causal ordering on the shock processes. For a more detailed description of the underlying methodology for both types of IRFs, see Section B.

3.4 Specifying trade linkages

Given the importance of the international dimension to our theme, we use a number of different ways of linking the economies together to ensure robustness. We connect economies using three frameworks:

1. 1. First, we use bilateral trade from Mohaddes and Raissi (Reference Mohaddes and Raissi2020). Their variable is the sum of exports and imports. We use the average of bilateral trade in the years 2014–2016.

2. 2. The second addresses the possibility that the impact of green patents on the stock market is mainly through financial markets. Thus, we collected bilateral financial flow from the Coordinated Direct Investment Survey of the International Monetary Fund (IMF). We use the sum of the inflow and outflow of direct investment positions. We use the average of bilateral financial flow in the years 2019–2021.

3. 3. The third framework addresses the concern that bilateral trade structurally changes over time. We incorporate this possibility by using varying bilateral trade in the years 1980–2022. We collected this data from the IMF’s Direction of Trade Statistics.

We shall examine the impact of these alternatives when discussing robustness in the subsequent section.

3.5 Stationarity and exogeneity

In the appendix, Tables A.2 and A.3 present the unit root tests. Most variables are stationary only in the first differences. Table A.4 shows the existence of cointegrating relationships. Consequently, we use the GVAR in the error correction form (see Pesaran et al., Reference Pesaran, Schuermann and Weiner2004, for details). Table A.4 also shows the lags of each $VARX$ . Table A.5 displays the weak exogeneity tests, which evaluate the inclusion of foreign variables in the individual models. Most estimates do not reject the null.

4.1 Monetary policy shocks

Figure 2 shows a 50 basis point initial increase in the federal funds rate and its effect on real output, stock market valuations, and green patents for the US economy. As remarked upon before, we see that even if they have no explicit environmental motivation, changes in monetary policy affect greening incentives and their resulting stock market valuations and so on.Footnote ¹² The panel of figures show the median response and associated $68\%$ confidence intervals as calculated by a bootstrap of $1000$ replications, over a 40-quarter (10-year) horizon.

With the monetary contraction, we see a fall in real output as demand contracts and financial conditions tighten. As interest rates rise, the aggregate stock market index and firms’ valuations also fall. This is consistent with an inverse relationship between asset prices and yields, resulting from financial flows moving out of private equity, venture capital, and other funds to higher-yielding government bonds. Although this negative effect on $q$ is only statistically significant in the very short run, we would nonetheless expect a pro-cyclical comovement of output and stock markets (noting that the effects on output are statistically significant).

Figure 2.

US monetary policy shock, 50 bps: impact on US economy. Notes: This figure shows the percentage response of real output, $q$ , and the green patent number for a temporary 50 basis point increase in the federal funds rate over a 10-year horizon. The gray horizontal dashed line shows the zero point.

Over time, these effects taper off to some degree although green patenting activity seems to undergo a more persistent reduction, monotonically falling to around 5% below baseline in the medium run. This accords with the stylized facts that R&D activity is highly pro-cyclical (e.g., Ouyang, Reference Ouyang2011): in effect, firms in economically constrained times sacrifice such budgets more readily than other production costs (or price margins).Footnote ¹³

4.1.1 Discussion

Accordingly, we can conclude that monetary policy changes can influence innovation activity and the incentives for adapting green technologies given that it impacts demand and profitability and funding sources (and perhaps firms’ risk appetite). Our results however only give a flavor of the issues at play, and further research is needed to understand how monetary policy can affect the green transition.

To illustrate, a superficial conclusion from the above simulation is that whenever monetary policy (conventional or unconventional) is restrictive, green patenting activity will suffer. And indeed, in the European context, Grimm et al. (2022) suggest that quantitative easing (expansionary central bank bond buying programs) had a positive effect on participating firms’ innovation rates.Footnote ¹⁴ However, from another perspective, the long regime of historically low interest rates and quantitative easing policies pursued by G7 central banks after the Great Financial Crisis from 2008 onward may have—consistent with some Schumpeterian endogenous growth theories—have kept "zombie" firms alive longer than would normally be the case, strengthened their market power, and delayed the entrance of new, potentially greener firms (see Aghion et al., Reference Aghion, Farhi and Kharroubi2018; Liu et al., Reference Liu, Mian and Sufi2022).

Figure 3.

Green patent shock: impact on US economy. Notes: The figure shows the generalized impulse response of stock market values to a one standard deviation US patent shock. The dark solid line is the median response, and the blue shaded areas represent 68% confidence intervals. The zero line is given by the gray dashed horizontal. The reporting horizon is 10 years.

4.2 Green patent shocks

Now, we consider a one standard deviation shock to the US green variables. Figure 3 presents the impacts of selected US variables, and Figure 4 shows the impact of that same shock on G7 stock markets.

4.2.1 Home shock

We turn first to consider the home (US) effect. The expansion of green patents has some features of a supply shock in that (at least initially) real output rises (reflecting the higher amount of resources available) and aggregate prices fall. Interest rates react very little. As discussed earlier, the literature has reported both positive and negative effects of green patenting on output and financial returns. In our case, we have both effects operating, albeit at different horizons. The behavior of the stock market is—as in the simple model—consistent with that of output: initially, $q$ rises reflecting the improved profitability and future prospects of green firms. But this effect turns out to be short-lived, and aggregate output and stock returns fall after about one year.

Figure 4.

Green patent shock: impact on G7 stock markets values. Notes: The figure shows the generalized impulse response of stock market values to a one standard deviation US patent on G7 stock markets. The dark solid line is the median response, and the blue shaded areas represent 68% confidence intervals. The zero line is given by the gray dashed horizontal. The reporting horizon is 10 years. All y-axis ranges are the same in the panels to facilitate cross-country comparisons.

We might initially view this as a curious result. Endogenous growth theory emphasizes that new patenting and R&D activities are positive drivers of per-capita output. In our case, though, there are a few considerations to bear in mind when assessing this result. First, as implemented here, the impact on stock market valuations from patenting activity is specific to green patents and not total patents: the introduction of green technologies will ultimately improve social outcomes but may involve significant adjustment costs for both green and brown firms, thus affecting private valuations.

Second, like standard VARs, the GVAR model is essentially a reduced-form framework built to capture data co-movements and data coherence, with a relatively loose theoretical underpinning. While this gives us confidence in the empirical reliability of the results, it leaves open their mapping to theory. In that respect, we might do best to consider these GVAR patent shocks as not exclusively an increase in the share of green patents as such but also as a shock embodying greater application of environmental regulations, increased stockholder pressure for ESG measures, more widespread adaption of the available green technologies, and so on. This broader interpretation is arguably more useful in our context since it precisely reflects the multifaceted nature of the ESG agenda.Footnote ¹⁵

Seen in that light, the results become easier to rationalize. We already noted that pollution abatement costs deplete current profits since they shift resources away from the production of goods. Those costs in the small model are assumed to be convex, implying that the more the adjustment the steeper the costs (in terms of production foregone). The ESG agenda will therefore push brown firms to clean up, thus cutting into their profit and stock market evaluations. Moreover, if those firms had higher purely private returns to capital, and the share of such firms are large, then indeed, overall stock returns will be depressed. Analogously, if the private returns to green patenting are low and deferred (e.g., Brynjolfsson et al., Reference Brynjolfsson, Rock and Syverson2021), and the output share of such firms is low, then these positive effects may not be sufficient to boost the overall stock market.Footnote ¹⁶

Ex ante, the overall effect is unclear, although the result that green patenting activity does not lead to a permanently higher level of stock returns in the medium run is robust to the case where the shock is implemented in a structural recursive fashion, further giving us confidence in its empirical validity (see subsection 4.4.1 and Section A).

A final point to note is that the GVAR, or any empirical model, is estimated on historical data; its parameters are given. In our context, arguably the model captures demand interactions better than supply. If the ESG framework were to become a more pervasive feature of the corporate landscape, it might lead firms to change their production framework (and thus lower their abatement costs), shift more earnestly toward cleaner technologies, and adapt the complementarity between dirty and clean factors of production.Footnote ¹⁷ A critical mass of firms doing so, would potentially paint a more favorable picture of stock market returns under the green transition.

However, the possibility that such a transition implies at least over some horizon negative returns and may require a complementary policy response (e.g., subsidies) is a reasonable conclusion to draw.

4.3 Decomposition of stock market influences

After showing the impulse responses, we now investigate the quantitative importance of green patents on the stock market using the generalized forecast error variance decomposition (GFEVD). This procedure decomposes a given domestic variable into domestic and external determinants. Consequently, we can capture spillover effects. We incorporate this possibility by adding the US green patents in all GFEVD. We normalized each row of the GFEVD table to a sum of 100%.

We explore the spillover effects of US patents on domestic stock markets in Table 2, which illustrates the GFEVD of stock markets for all economies. The table is organized into seven blocks, with each block capturing the influence of domestic variables ( $y, q, r, l, e, cpi$ ) on the domestic stock market. The second part focuses on the impact of green patents on stock markets. In this segment, we analyze the effects of both domestic and US green patents. The US green patent illustrates the spillover effect on domestic stock markets.

Table 2.

GFEVD of stock markets to domestic variables and the US green patents

Notes: This table decomposes the determinants of home stock markets into home variables and US patents over three different quarterly horizons. Each row sums to 100.

We start our analysis in Canada. In the first period, GDP influences the stock market by 2.58%, the interest rates, the exchange rate, and domestic green patents all contribute around 5%, prices by 0.35%, and the stock market itself explains 77% of its own fluctuations. These values denote the extent of the Canadian stock market’s fluctuation attributed to domestic factors. The column pertaining to US green patents reveals that a mere 1.1% of first-period stock market fluctuations result from spillover effects. Over time, however, the influence of green patents grows. In the last period, Canadian green patents affected the stock market by 17.6%, while US green patents contributed to a 35.57% impact. The increasing importance of green patents in the fluctuation of domestic stock markets is not exclusive to Canada. The table suggests a growing influence of green patents over time, including spillover effects from the US green patents.

In the remaining economies, US green patents present an increasing impact on domestic stock markets over time. In the last period, the spillover effect amounted to 21.91% in France, 31% in Germany, 14% in Italy, 24.19% in Japan, and 33.81% in the UK. This suggests the existence of relevant spillover effects of US green patents in the G7 economies. Additionally, similarly to the impact of US green patents, domestic green patents demonstrate relevance in explaining fluctuations in stock markets. The exceptions are France, Japan, and the UK, where domestic green patents displayed lower values (ranging from 0% to 1%).

In short, we found that the US green patents affect international stock markets, provoking their falls. Thus, we provide evidence of persistent spillover effects from green technology innovation on financial markets. The subsequent subsections test our main results by changing the model’s configuration.

4.4 Robustness

4.4.1 Robustness: the structural GIRF

One limitation of the GIRF is that it does not identify shocks in a structural manner. We overcome this limitation by using the structural GIRF (SGIRF) with a given causal ordering. The SGIRF identifies shocks in one economy: here, the USA. In this case, we adopt the order:

\begin{align*}green \rightarrow y \rightarrow cpi \rightarrow r \rightarrow l \rightarrow q\end{align*}

where $\rightarrow$ indicates the presumed direction of causation. In other words, we treat green patents as the most exogenous variable. A shock to green patent affects production, causing changes in prices. The credit markets react to short- and long-term interest rate fluctuations. These movements, in turn, affect the stock markets. The results are shown in Figure A.1 and Figure A.2 in the appendix. As can be seen, the results are essentially invariant to those of the GIRF.