Digitization related services and applications are based on the information and comm unications technology (ICT) ecosystem and encompass almost all areas of society and economic sectors nowadays and exert numerous opposing effects in regard to electricity c onsumption and corresponding CO2 emissions. Our analysis aims to inform policy decis ion makers about the actual climate relevance of the ICT ecosystem by providing sound empirical evidence on the net effect of various ICT core elements based on recent OECD panel data utilizing panel econometric estimation methods that include instrumental variables. When compared with previous empirical contributions, we utilize more comprehensive measures of the ICT ecosystem and explicitly address potential endogeneity co ncerns. In line with all the previous studies using data from developed countries, w e found that the CO2 reducing positive indirect effects outweigh the negative, in other words, CO2 increasing direct and indirect effects on average. Specifically, we found that, in addition to the lowering effect related to the use of basic broadband conn ections, there was another lowering effect albeit smaller related to new fiber based broadband connections. We found that according to our conservative estimates, basic, and fiber based broadband connections induced a substantial reduction of CO2 emissions in the average OECD country amounting to at least 67 Mt CO2 during our period of analysis (2002 2019). This roughly corresponded to the total annual CO2 emissions of an OECD country with the size of Greece. In contrast, other elements of the ICT ecosystem , such as mobile broadband networks or electronic end user devices, showed no significant net impact on CO2 emissions. This result points to potentially opposing and, by, and large, offsetting effects at an aggregate level and/or to the dominant role of th e other macroeconomic, demographic, and institutional control variables in exp laining total CO2 emissions at the country year level. We conclude that undifferentiated climate policy measures imposed on the ICT ecosystem would not do justice to the identif ied heterogeneity, with numerous in part opposing effects, and likely would be accompanied by inefficiencies and market distortions. In view of this heterogeneity, regulatory interventions, if any, should be targeted at reducing particularly resource inten sive digital services with evidentially high CO2 emissions (such as online vid eo streaming or bitcoin mining). Moreover, our findings based on data of developed countries provides evidence for the "pollution haven hypothesis" suggesting that environmentall y intense production of ICT network equipment and end user devices, the extrac tion of rare earth elements and disposal of ICT waste is allocated to some major non OECD member states such as India, China and some East Asian countries (other than OECD member states Japan and South Korea). Whereas in OECD countries the value added from ICT services has been rising, the value added from ICT manufacturing, with particularly high CO2 emissions, has been falling. Although the ICT sector as a whole is growing world wide, the growth of energy intense ICT production and manufacturing differs su bstantially between regions and countries.
