The Role of Digital Technology in Climate Technology Innovation^†

A. Data

I use the USPTO’s patent database to ascertain the levels of development in digital technology and climate technology for each OECD member country.5 The USPTO patent database provides detailed information on ultimately granted patent applications filed by individuals or firms worldwide, including abstracts, lists of previous patents cited, technology classes, lists of inventors, and names and addresses of owners and inventors. The USPTO patent database used for the analysis here contains a set of ultimately granted patent applications from 1976 to 2021. As is well known in previous studies that also use the USPTO patent database, it takes from one year to even as long as ten years, three years on average, for filed patent applications to be granted (see Figure 1-A). Thus, the patent granted year is too far away from the year the innovation occurred. Although the application year may not be the exact year the innovation occurred as well owing to the time required to prepare the patent application documents, this gap should be narrower. Thus, following the previous studies, I use the application year as a proxy for the year the technological innovation occurred.

Also, because the USPTO patent database contains only ultimately granted patent applications and given that the granting process excessively long, the number of patent applications filed falls rapidly after 2016, as shown in Figure 1-B. For example, most of the patent applications filed with USPTO in 2021 were under review as of 2021. Thus, the number of ultimately granted patent applications among the 2021 cohort is very low. Therefore, despite the possibility that the actual number of patent applications filed as well as the quality of the applications in 2021 may equal those factors for 2016 and thus that the number of ultimately granted patent applications counted in 2030 is identical in both years, the number of ultimately granted 2021 patent applications should be very low if counted in 2021. In other words, because the patent application examination process is long, the number of patents created by firms, hence in the economy in 2021, becomes observationally low compared to that in 2016 in the 2021 version of the USPTO patent database. To correct for such bias coming from this type of examination lag, I use patent applications filed up to 2016 in the subsequent analyses. Furthermore, I use the patent applications filed by entities in the 38 OECD member countries in the upcoming analyses, as these should be comparable in terms of the quantity, quality, and composition of the technologies.

FIGURE 1.

USPTO PATENT APPLICATIONS

To classify the technology of each patent, I use USPTO’s Cooperative Patent Classification (CPC) scheme in the following analyses, which USPTO constantly updates to maintain time consistency throughout all of the years the patent data are available. There are nine codes in the 1-digit CPC section, 130 in the 3-digit CPC class, and 670 in the 4-digit CPC subclass. Table 1 shows the number of patent applications and the number of CPC codes for the top ten OECD member countries in terms of the total number of patent applications in 2016. Table A1 in the Appendix shows the same information for all 38 OECD member countries. For more detailed explanations and analyses of the USPTO patent database, I refer to Hall et al. (2001).

TABLE 1

TOP 10 PATENTING OECD MEMBER COUNTRIES

B. Climate Technology and Digital Technology Definitions

As briefly explained previously, USPTO assigns one or more CPC codes to each patent to classify the technologies each patent contains. Among these CPC codes, Y02 is the code USPTO additionally assigns to patents to track the developments in technologies related to mitigation and adaptation to climate change. This code was initially developed jointly by the European Patent Office (EPO), the United Nations (UN), and the International Centre on Trade and Sustainable Development (ICTSD) and used thereafter. To develop an automated way to assign Y02 to relevant patents, climate technology experts from the three institutions first applied textual analysis to all available descriptions in the patent database, such as abstracts and claims, to select patents broadly related to climate technology. The experts then manually inspected and removed false matches using additional information, such as technology classifications, to finalize the assignment. This routine is developed as an algorithm that can automatically assign Y02 to new patents under the supervision of patent experts, and this algorithm has been maintained and updated constantly. See Veefkind et al. (2012) for a more detailed explanation. Thus, in this paper, I classify patents assigned to Y02 as patents for climate technology. For example, Y02C under Y02 is for technologies to capture, store, sequestrate, or dispose of GHG, and especially Y02D is for technologies to lower the power consumption of ICT products by, for instance, low-power computing.

Among CPC classes, G06 pertains to technologies for computing, calculating, and counting. Under G06, six subclasses, G06F, G06K, G06N, G06Q, G06T, and G06V, are codes assigned to technologies related to digital transformation technologies. Thus, I classify patents assigned to these six subclasses as patents for digital technology. For example, one of the CPC groups under G06N, G06N 20/00, is for machine-learning-related technologies, and one of the CPC groups under G06F, G06F 1/32, is for technologies for lowering device power consumption levels by processing digital data.

Figure 2 shows digital and climate technology development trends worldwide, where the degree of development is measured as the share of patent applications for digital and/or climate technology from the total number of patent applications filed each year. 6 Because each patent is assigned to one or more CPC, there are cases in which a patent is assigned to both digital technology and climate technology. In such cases, I include the patents in both pools of patents for digital and climate technologies when counting the number of patents in the corresponding pools. Then, I additionally define climate + digital technology — technology used for both climate change mitigation and digital transformation — and compute a trend for the development of such technology separately. The share of patents for climate technology among the patents registered with USPTO stagnated in the 1980s and 1990s and then showed a steady increase in the 2000s. This was followed by another period of stagnation after 2010. In contrast, the share of patents for digital technology showed a steady increase in all periods. Also, the share of patents for climate + digital technology, such as G06F 1/32, showed a steady increase in all periods as well.

FIGURE 2.

TRENDS FOR DIGITAL AND CLIMATE TECHNOLOGIES

C. Different Revenue Uses from a Carbon Tax

To estimate the effect of digital technology on climate technology and vice versa, I define a rapid increase (spike) in the number of innovative technological improvements as a technology shock for each technology and compute as follows. Just as research papers, patents are required to cite all of the previous patents their technological improvements are based on or related to. Thus, previous studies using the patent database use the number of forward citations received as a measure of the quality or degree of improvement (innovation) each patent contains. Thus, I define patents with the number of forward citations received above the 99^th percentile of the forward citation distribution for all patents worldwide as patents related to innovative technological improvements.

In this paper, I use the technology-year mean-adjusted values for the number of forward citations received, where the technology is defined at CPC4-level. As is well documented in Hall et al. (2001), both the number of patents created each year and the corresponding trend vary across different technologies. Thus, the number of patents which could potentially cite a specific patent may differ for each technology and year. Also, patents that were created in early periods have the potential mechanically to receive more citations compared to recent patents. In an extreme case, patents created today should have received zero citations. Due to these reasons, for example, a patent related to internal combustion engines created in 1990 and that received 2,000 forward citations could have a lower degree of innovation than an AI-related patent created in 2021 that received 50 forward citations. Thus, it is necessary to adjust for these biases in the number of forward citations to use this measure correctly to assess the quality of patents or the degree of innovation each patent contains. To do this, I compute the average number of forward citations received by patents in each technology-year and divide the number of forward citations each patent in each technology-year received by this average number accordingly.7

Then, I define the share of the number of patents for innovative technological improvements from all patents as a measure of the degree of innovative technological improvement for each technology. Here, I use the technology-year mean-adjusted number of forward citations received when counting the number of patents to take the quality of innovation into account. Finally, I compute the DHS (Davis, Haltiwanger, and Schuh, 1998) growth rate of this degree of innovative technological improvement for each technology-year, defining the year when the growth rate is above the 75^th percentile of the growth rate distribution for each technology as the year the technology shock (spike) occurred for that technology. According to this methodology, the years the digital technology shock occurred within the regression sample period (1983 to 2016, a total of 34 years) are 1983, 1987, 1989, 1991, 1995, 1999, 2001, 2003, 2004, and 2010 (a total of ten years). For 2010, such digital technology shocks include a total of 495 versatile patents, such as “Electronic Device with Text Error Correction Based on Voice Recognition Data” (Apple, US8719014), “System and Method for Calculating the Thermal Mass of a Building” (Ecofactor Inc., US8131497), Digital Mapping System (Google LLC, US7894984), and “Controlling Power Consumption of a Mobile Device Based on Gesture Recognition” (Qualcomm Inc., US9086875). We could think of other ways to measure a digital technology shock, such as finding an exogenous shock to digital technology, including government policy changes, and using them as instrument variables. However, I was unable to find such exogenous variations.

A. Model Specifications

To estimate the effect of digital technology (a subset of G06 defined previously) on climate technology (Y02), I estimate the following regression model:

The dependent variable log(npat_cjt) that represents the developments in technology j is the number of patent applications for each technology ( j ) in each country ( c ) in year t. As explained previously, I use the technology-year mean-adjusted number of forward citations received when counting the number of patent applications to take into account the quality of innovation, where CPC4 is used for the technology classification. The independent variable is a dummy variable equal to one if technology j belongs to the climate technology Y02, and is a dummy variable equal to one if we observe a digital technology shock in year t-s. Thus, the coefficient in front of the interaction term , β₁ estimates the additional effect of digital technology shock s years ago on climate technology compared to all other technologies. Because it takes time for the USPTO examiners to evaluate the patent applications submitted to the USPTO, to protect their rights, and to make the information available to the public, s should be more than one year. Also, because other firms need time to learn the available information and apply it to their technological developments and then spend more time preparing and submitting patent applications to the USPTO, s should be more than two years. Thus, I use s = 3 in the baseline regression analysis. To test the robustness of the results, I also use various values of s and report the results in Table B1 in the Appendix. The coefficient estimate for the effect of a digital technology shock in the same year on climate technology, however, is small and statistically insignificant.

δ_ct is a country-year fixed effect to control for country-level transitory components that could affect technological development trends and the patent application submission difference across countries. Given that is absorbed by δ_ct , it is not included as an independent variable separately. δ_ct is a country-technology fixed effect to control for the difference in the compositions of technological developments across countries. Because is absorbed by δ_cj, it is not included as an independent variable separately. α is a constant term. As we need to use the three-year prior technology shock in this regression, I limit the analysis sample period to the years 1983 to 2016 (34 years). There are 38 countries with 666 CPC4 in this regression sample, and the total number of observations is 241,402. The mean and standard deviation of the dependent variable log(npat_cjt) are 1.97, and 1.78, respectively. The number of years in which a digital technology shock amounts to ten, as explained previously.

B. Baseline Result

Table 2 shows the baseline regression results for no fixed effects, the country-year fixed effect only, the country-technology fixed effect only, and both the country-year and the country-technology fixed effects included. As shown in the figure, the estimates for β₁ are positive and statistically significant for all combinations of fixed effects. The results for the main specification in column four show that a digital technology shock additionally increases the number (technology-year-adjusted number of forward citations received) of climate-technology-related patents by 1.24 (exp(0.213)), compared to all other technologies. As the average number of climate-technology-related patents each year is 7.17 (exp(1.97)), this result shows that a digital technology shock increases the number of climate-technology-related patents by around 17.3% on average. This is an important result showing that, in addition to the previous findings that the digital transformation process helps mitigate climate change by reducing energy use in various ways, the development of digital technology additionally helps mitigate climate change by promoting the development of climate technology, which is very important in the long run. Thus, it is clear that government policies related to the development of digital technology should have an important effect on the development of climate technology.

TABLE 2

EFFECT OF A DIGITAL TECHNOLOGY SHOCK ON CLIMATE TECHNOLOGY

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

C. Technology Shock Measure Validation and Robustness Test

To test whether the technology shock measure used for the previous analysis captures a simple trend or spurious relationship between digital technology and climate technology, first I regress the current ( t ) climate technology on the future ( t + s ) digital technology shock measure. Because I need to compute the future technology shock, I limit the regression sample period so that it ranges from 1983 to 2013. The first column of Table 3 shows the result for s = 3. As shown in the table, the coefficient is small and statistically insignificant. Thus, there is a low possibility that the current digital technology shock measure simply captures spurious relationships. Table B2 in the Appendix shows the results for s = 1 and s = 2 . We can observe that the coefficient for s = 1 is statistically significant at 10%. This may stem from the possible release of information for the developed technology before patent application submission, or the large sample size (221,532 observations). However, this result requires further analysis. For s = 2 , the result is statistically insignificant.

The second column in Table 3 shows the estimate of the effect of a digital technology shock on all technologies. This is computed to test whether we obtain the baseline results not because the digital technology shock measure indeed captures its effect on climate technology but because it captures its effect on overall technology or because there is merely a spurious relationship. However, the resulting coefficient estimate is small and statistically insignificant, implying that the possibility of such concerns actually coming to be is low.

The third column in Table 3 shows the estimate of the effect of a climate technology shock on digital technology. This reverse causality estimation allows us to test whether the baseline results are driven by a simple correlation or a common trend between digital technology and climate technology and not by the effect of digital technology. This test is important as there is no strong reason as to why climate technology can directly affect digital technology. As shown, the result is statistically insignificant. Thus, the baseline result is likely not capturing a simple correlation.

The fourth column of Table 3 shows the result including in the baseline regression specification (1) to estimate the effect of digital technology on climate technology, controlling for the effect of a climate technology shock on digital technology. As shown, the coefficient estimates are identical to those before.

TABLE 3

ROBUSTNESS CHECK FOR THE EFFECT OF DIGITAL TECHNOLOGY SHOCK I

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

Finally, Table 4 shows whether the baseline results hold even after controlling for the effect of past (three years prior) climate technology shocks on climate technology itself. I run this test as the baseline estimates could be biased if climate technology shocks happened to occur simultaneously with digital technology shocks. However, even when controlling for climate technology shocks, the effect of a digital technology shock remains statistically identical to the baseline result. Furthermore, as shown in the fourth column, the effect of a climate technology shock becomes statistically insignificant after controlling for full fixed effects. All of the results above suggest that the baseline regression result more likely identifies the (causal) effect of digital technology on climate technology than otherwise.

TABLE 4

ROBUSTNESS CHECK FOR THE EFFECT OF DIGITAL TECHNOLOGY SHOCK II

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

Additionally, I estimate the effect of a digital technology shock on climate + digital technology, as briefly explained in the previous section. Although the share of patents for this type of technology is quite small (0.6% of the total number of patents), as shown in Figure 2, it could be used as an additional robustness test for the baseline results, removing the indirect effect of digital technology, as it is an aspect of climate technology that directly uses digital technology. In this regression analysis, the dependent variable is the country-year-level logged citation-adjusted number of forward citations for patents pertaining to climate + digital technology, and the independent variable is the yearly-level dummy variable for a digital technology shock. Thus, I include the country-fixed effect only in this regression. These results are reported in Table B3 in the Appendix. The coefficient estimate of 0.304 is similar to the baseline result. However, the statistical significance is low (10%) as the regression sample is at the country-year level, with the number of observations equal to 1,197.

Furthermore, I estimate the effect of a digital technology shock on climate technology after separating climate + digital technology from digital technology and climate technology to remove any potential bias that could arise due to a potentially mechanical relationship between a digital technology shock and some of the climate technologies that are actually climate + digital technologies. As reported in the second column of Table B4 in the Appendix, the coefficient estimate is statistically identical after reassigning the climate + digital technology. For comparison purposes, the first column shows the baseline result.

Finally, Table B5 in the Appendix shows the estimate of the effect of digital technology on climate technology, estimated by directly regressing the (logged) number of patent applications for climate technology on the three-year lagged (logged) number of patent applications for digital technology. The advantage of the baseline model over this regression model is that we can isolate the effect of digital technology on climate technology from other factors that could shift both technologies simultaneously, such as a spurious common trend. Additionally, we can estimate the effect of digital technology specific to climate technology by comparing it to the effects of other technologies. Nonetheless, this exercise can also confirm the robustness of the baseline results. Here, only country and time-fixed effects are separately included in this regression specification, as we only have the country-year variations.

As shown in the first column, the coefficient estimate of 0.333 is statistically identical to the baseline estimate of 0.213 reported in Table 2. The second column in Table B5 estimates an additional effect of the year the digital technology shock occurred on climate technology by testing the interaction between the three-year lagged (logged) number of patent applications for digital technology and the digital technology shock measure used in the baseline model. Unless these years are special with regard to the development of climate technology notwithstanding the fact that these are years digital technology shocks occur, all of the digital-technology-specific effects should be absorbed by the number of patent applications for digital technology. Consistent with this perceptive, we have a statistically insignificant and small coefficient for the interaction term. All of the tests above confirm that the baseline estimate of the effect of digital technology on climate technology is robust to many other specifications.

D. Extension I

To understand how digital technology affects climate technology in more detail, I extend the baseline regression specification and run several additional analyses. In the first extension, I test whether digital technology affects climate technology differentially in Korea compared to other OECD member countries. I run this test as we need to find causes and find ways to improve the relationship between the two technologies if we find that the effect is lower in Korea. To do this, I additionally test the interaction of a dummy variable for Korea () with the baseline specification. As shown in the second row of the first column in Table 5, the result is statistically insignificant for the triple-interaction term. Thus, the relationship between the two technologies is due to their specific characteristics at the technology level, and I could not find evidence that their relationship is different in Korea due to government policies or the level of technological development.

The next extension is to estimate the effect of digital technology on smart grid technology (Y04), and this result is reported in the second column of Table 5. The coefficient estimate, however, is statistically insignificant. This may stem from the fact that AI-related technological development, which is expected to have a large impact on smart grid technology, showed major developments after 2016. Thus, this result requires further analysis after compiling a longer dataset. Furthermore, this result may be driven by the fact that we are actually combining various detailed digital technologies and using an aggregated form of digital technology in the analysis. If some of these detailed digital technologies have offsetting effects on smart grid technology, such a result would arise. The results when testing the validity of the above extension analyses are reported in Table B6 in the Appendix.

TABLE 5

EFFECTS OF A DIGITAL TECHNOLOGY SHOCK, KOREA SPECIFICITY TEST, SMART GRID

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

E. Extension II

Because the technology shock measure used in this paper is a dummy variable that assumes zero or one at the annual level, it cannot be a good measure of fine technology classifications because we would not be able to separate different fine technologies’ movements correctly with annual-level zero one values if these fine technologies co-move in a broad sense. With this limitation in mind, I estimate the effects of finely defined digital technologies in the digital technology used in the baseline analysis on climate technology and examine the effect of digital technology in detail.

Table 6 shows the results for Electric Digital Data Processing (G06F), Computing Arrangements Based on Specific Computational Models (Machine Learning-related technologies, G06N), and Image Data Processing or Generation in General (G06T). 8 We see that climate technology is significantly affected by these three technologies, which exist at the heart of digital technologies. These findings not only help us to understand the detailed role of the digital transformation on climate change mitigation as explored in the previous studies but also highlight the additional importance of digital technological developments on climate change mitigation. I will discuss this in detail in the next section, where I analyze the effects of digital technology on finely defined climate technologies. I report the regression results to test the validity of these results using the future technology shock in Table B7 in the Appendix.

TABLE 6

EFFECTS OF DETAILED DIGITAL TECHNOLOGY SHOCKS ON CLIMATE TECHNOLOGY

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

In addition to these aspects, I estimate the effects of finely defined digital technology on smart grid technology. These results are reported in Table 7. In the previous analysis using aggregated digital technology, the estimated coefficient was statistically insignificant. Here, I find statistically significant results at the 10% level for data-processing-related technology and machine-learning-related technology. In fact, it may be possible to obtain statistically more stable (significant) results for these in a few years when a longer dataset becomes available, as machine-learning-related technologies improved rapidly after 2016, and smart grid technology is also relatively new. The robustness test for these analyses using a future technology shock is reported in Table B8 in the Appendix.

TABLE 7

EFFECTS OF DETAILED DIGITAL TECHNOLOGY SHOCKS ON SMART GRID TECHNOLOGY

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

F. Extension III

Finally, I analyze the effects of digital technology on finely defined climate technology. As shown in Tables 8 and 9, digital technology has clear effects on six finely defined climate technologies out of eight: Climate Change Mitigation Technologies Related to Buildings (Y02B); Capture, Storage, Sequestration or Disposal of GHG (Y02C); Climate Change Mitigation Technologies in ICT (Y02D); Reduction of GHG Emissions, Related to Energy Generation, Transmission or Distribution (Y02E); Climate Change Mitigation Technologies in the Production or Processing of Goods (Y02P); and Climate Change Mitigation Technologies Related to Transportation (Y02T).9 Below are some examples that can help us understand what these findings mean and how they are materialized in the real world.

TABLE 8

EFFECTS OF A DIGITAL TECHNOLOGY SHOCK ON DETAILED CLIMATE TECHNOLOGIES I

Note: 1) Statistical significance levels: ***p<0.01, **p<0.05, *p<0.1; 2) Standard errors in parentheses.

TABLE 9

EFFECTS OF A DIGITAL TECHNOLOGY SHOCK ON DETAILED CLIMATE TECHNOLOGIES II

In 2016, Google was able to reduce the electricity they used to dissipate the heat produced by their data center servers by 40% and lower the data center’s power usage effectiveness (PUE) by 15% by finding a way to control the building’s HVAC efficiently using DeepMind’s machine learning technology.10 A lower PUE implies higher energy efficiency. Google expects that this technology could lower electricity and water used in power plants and semiconductor manufacturing factories. Google’s effort in this way to apply digital technology newly to climate technology gave birth to several startups with HVAC optimization of buildings as their business focus. One example is BrainBox AI, and such innovations by these startups have reduced GHG emission levels of residences, hotels, airports, and grocery stores by 20~40%. Also, according to BrainBox AI, their customers were able to lower their electricity bills by 25%.11 US969723 is the patent filed by BrainBox AI in February of 2019 to the USPTO, which was granted in January of 2021, and this patent is for a system and methods of optimizing HVAC control in a building or network of buildings. This technology, categorized as G06N (machine learning-related technology), processes HVAC-related historical data, weather forecasts, and occupancy rates through machine learning to find and utilize the optimal HVAC requirements. This new technology started by Google clearly demonstrates the large contribution to Climate Change Mitigation Technologies Related to Buildings (Y02B). Also, as Google claims, this new technology will make an important contribution to the development of the technology categories of Reduction of GHG Emissions, Related to Energy Generation, Transmission or Distribution (Y02E), and Climate Change Mitigation Technologies in the Production or Processing of Goods (Y02P). This type of machine-learning-based technology will contribute to climate change mitigation by helping to improve existing technologies and heralding the birth of new technologies.

Image Data Processing or Generation technology (G06T) helps mitigate climate change in various fields. It helps computers to analyze video data from traffic cameras in real time to control traffic signals and solve the traffic congestion problem, which ultimately lowers GHG emissions. This technology also helps analyze satellite images to find methane gas leaks. The Carbon Mapper Satellite Sensor, developed by joint partners including Planet Labs, NASA, the state government of California, and the University of Arizona, is a good example of using image data-processing technology to find areas with methane gas and carbon dioxide leaks. 12 Owing to such image data-processing technologies, we are able to find problematic areas in real time and mitigate climate change by solving the problems using GHG processing technologies (Y02C). This type of technology will also be used in new areas and will greatly help us to mitigate climate change in the near future. I will discuss examples where data-processing-related technologies, such as technologies for power consumption reductions using data processing, are used for climate technologies, including ICT-related electricity consumption reduction technology, in the next section while discussing policy implications. The analysis results using a future technology shock to test the validity of the results above are reported in Table B9 and B10 in the Appendix. I also analyzed the effects of finely defined digital technologies on finely defined climate technology. I do not include the results from this analysis because they are not stable, which may be due to the possible limitation of the technology shock measure used, as discussed previously, or due to the mismeasurement issue that can arise when dividing data too finely. I plan to re-run this analysis once I find a way to measure a technology shock in a continuous manner.

A. Policies for supporting development in technologies for reducing power consumption by ICTs

As shown in previous studies and by the real-world examples discussed in the previous sections, the digital transformation, especially that of AI technology that uses high-performance computers and data centers, will play an even more significant role in climate change mitigation. Moreover, as illustrated by the previous analysis results in this paper and the results from other papers, it is not easy for us to disagree that digital technology can mitigate climate change by improving and being combined with climate technology. However, it is also a fact that high-performance computers and data centers negatively affect climate change due to their intensive power consumption. As I introduced in the previous sections, many existing studies and policy institutions worry about and discuss this intensive energy use as it pertains to digital technology. Koomey et al. (2011) and the IEA (2022) show that although developments in digital technology improve energy efficiency even more rapidly than performance improvements, there is room for improving energy efficiency even more, and we need to speed this up to accomplish the net-zero goal by 2050. For these reasons and because the pace of the digital transformation and electrification of energies will increase, the importance of technologies for reducing ICT power consumption levels will progressively increase.

Figure 3 shows a specific example where data-processing-related technology (G06F), including technologies for power consumption through data processing, is applied to a technology for reducing power consumption by ICTs. This is a patent about reducing energy consumption by computer processors, which was applied for by Intel in 2011 and granted in 2015. As written in the corresponding abstract at the bottom right, this technology identifies idle processes among the processes executed in the computer processors, combines them, and reduces the power used by these idle processes.

FIGURE 3.

EXAMPLE OF A DATA-PROCESSING-RELATED TECHNOLOGY (G06F) APPLIED TO A TECHNOLOGY FOR REDUCING POWER CONSUMPTION BY ICTS

Technologies that can help reduce the electricity use of ICT products, such as computer-related products, not only greatly help mitigate climate change but also help expedite the digital transformation by lowering the energy cost of firms. Also, as shown in the previous regression analysis results, the development of such digital technology facilitates the development of climate technology. To summarize, technologies for reducing power consumption levels by ICTs represent a solution to the worrying issues of increasing energy consumption and higher GHG emissions levels induced by the digital transformation. Also, through the development of technologies capable of reducing power consumption by ICTs, we can potentially increase the net energy reduction through the digital transformation process while also helping climate technologies to advance. Thus, if we can explicitly support the development of technologies that reduce power consumption by ICTs through climate policies, not only will this also help reduce climate change, but it will have a positive effect on the digital transformation.

B. Policies for removing factors that could hamper investment incentives for digital technology

Although climate policies are uniformly applied to all areas to accomplish GHG emission goals overall, we can redesign these policies so that we can apply different measures, such as lowering the restrictions or slowing down the policy implementations for areas where we expect to see a rapid reduction of GHG emissions in an innovative way in near future due to, for instance, technological development. By doing so, although we may not be able to meet our short-term goal, we can achieve a larger GHG emission reduction in the mid to long term. As technology-related investments are sensitively affected by the expected returns, regulations in general harm firms’ technological development incentives by lowering the expected return from an investment. Samsung, for example, is said to be facing difficulties in fulfilling its RE100 goal due to an increase in electricity use caused by the production of semiconductors with new technologies, local renewable energy prices, and supply issues. 13 Although participating in the RE100 initiative is thus far voluntary (i.e., although participation is in part due to market pressure, it is not enforced by the government), it is clear that this type of new friction can impact investment returns. This arises because the power supply problem negatively affects semiconductor production and lowers profits by increasing the unit production cost, which in turn lowers expected returns from investments in technological developments. Importantly, however, semiconductors produced using new technologies can do the same tasks using considerably less electricity compared to the existing types, as briefly explained previously. Thus, technological developments in the semiconductor industry can play an important role in significantly improving the net GHG emission reduction effect of the digital transformation.

Continuing the example of semiconductors, firms in this industry always make their mid to long-run roadmaps for technological developments available to the public and attempt to accomplish their innovations accordingly. Thus, it is easy for us to evaluate mid to long-run improvements in the power efficiency and reductions of GHG emissions in this industry compared to those in other industries. By comparing the evaluated expected GHG emissions reduction from using one unit of new semiconductor (where the evaluation takes into account inter-generational technological development spillovers that could arise due to the sequentiality of innovation) with the GHG emissions from producing one unit of new semiconductor, we could adjust the strength of climate policy enforcement in this industry based on the net GHG emissions reduction level. For other digital technologies, we could guide firms to make their technological development roadmaps available to the public so that we could similarly adjust the strength of climate policy enforcement. Thus, we will be able to expedite the GHG reduction speed at the national level if we confirm other cases similar to those in the semiconductor industry and design detailed policies so that we can promote the development of products and technologies related to energy consumption reductions by adjusting the speed of the green transformation (low-carbon transformation) for production facilities. Also, proper government support for producing the products necessary for the digital transformation, such as semiconductors, would greatly help with the digital transformation. However, these relaxations of regulation should be done while traditional environmental regulations, such as those pertaining to wastewater management, are strictly enforced.

Finally, we could consider policies that support the production and development of products that can lower the electricity consumption of consumer electronics, that is, policies that directly support the development and production of low-power consumption products in general. The Korean government currently indirectly supports the development and production of highly energy-efficient consumer electronics by encouraging the demand for such products through an expenditure subsidy program for top-rated energy-efficient products. However, there is no policy of direct support for such products. The reason for considering direct support for the development and production of highly energy-efficient consumer electronics is ultimately to reduce the time and effort required to accomplish carbon neutrality by reducing the level of electricity production through the use of fossil fuels as the total amount of electricity used could be reduced by reducing the electricity used by each product.

Identical to the semiconductor case, we can measure the energy reduction rates of new products (degree of reduction for energy use compared to existing products) by using the level of electricity use for each product, which is a measure currently used for computing the energy efficiency rating. Furthermore, by using this energy reduction rate of new products, we can also compute how much the GHG emissions are reduced due to the development and production of new products. Then, based on this measure, we could fine-tune the timing and degree of climate policies imposed on each firm. Also, for firms planning to use renewable energies, we could provide benefits such as prioritizing renewable energy use or could provide subsidies based on firms’ energy reduction rates of new products. These types of detailed policy support are not possible only through conventional demand-based indirect support. We will be able to expedite the carbon reduction speed at the national level if we can increase the production and demand for products that consume relatively less power.