We present three sets of empirical results: First, we show that delegation sizes in IPCC meetings increased between AR5 and AR6, on average. Second, by focusing on online negotiations in AR6, we assess the extent to which delegation size matters for countries’ engagement levels during approval sessions, where we find mixed results. Third, we provide evidence that differences in time zones did mute the positive effect of larger delegations on engagement levels in WGII and WGIII.

4.1 Changes in delegation sizes from AR5 to AR6

We begin by demonstrating that country delegations increased in size across all three Working Groups for virtual meetings. Figure 1 shows average increases when we pool our data across all three WGs (left panel) and for each WG separately (other three panels). The solid line documents that, across the board, delegations were, on average, larger by about two delegates in virtual approval sessions in AR6 compared to in-person approvals in AR5. This increase is most pronounced in WGs II and III and consistent with evidence that WGII in AR6 “had the highest number of delegates ever registered for an approval session” (ENB 2022, 22).

Average delegation size in AR5 and AR6. Solid lines show changes in average delegation size from in-person IPCC approval sessions in AR5 to virtual IPCC approval sessions in AR6. Dotted lines show changes in delegation size for Synthesis Report approval sessions in AR5 and AR6, both of which were conducted face-to-face. The left panel shows results for pooled data across all three Working Groups; the other three panels show results separately for each Working Group

Average changes in delegation sizes do, however, mask important variation at the country level. Out of a total of 156 countries which sent delegates to either the AR5 or AR6 Working Group approval sessions, roughly 6 out of 10 increased their average delegation size (93 countries), while one third, or 52 countries, reduced it; 7% held delegation sizes constant (11 countries). The increase in countries’ delegation sizes—averaged across WGs for each of the ARs—were much larger (\(+\)3.6 delegates on average) than reductions in those countries that sent fewer delegates (−0.8 delegates on average). Aside from Japan, whose delegation size decreased substantially from an artificially high baseline of 53 delegates in WGII in AR5, which was hosted in Yokohama, most countries that reduced their delegation sizes did so by less than one delegate on average.

Compared to attendance in AR5, 31 countries no longer had a presence in Working Group sessions of AR6, whereas 25 countries attended AR6 Working Group sessions but were absent in the approval of AR5, including many small island states like Antigua and Barbuda (1.33 delegates), Samoa (2.33 delegates), St. Kitts and Nevis (5.33 delegates), and Vanuatu (8 delegates). The three largest delegations came from the United States (25.3 delegates, \(+\)19.3 from AR5), Canada (23 delegates, \(+\)19.7 from AR5), and South Korea (21.3 delegates, \(+\)10.3 from AR5). Some countries, such as Turkey (10.7 delegates, up from 0.3 delegates in AR5), Malaysia (15 delegates, up from 1 delegate in AR5), and Argentina (13.7 delegates, up from 1 delegate in AR5) increased their delegations more than ten-fold for the approval of the AR6. Figure SI1 in Appendix C visualizes these changes for all countries and all approval sessions.

As discussed in the Research Design section above, the increases in delegation sizes may not relate to the virtual format of the approval sessions, but may instead reflect the greater salience of climate change in domestic politics (Colgan et al. 2021; Bayer and Genovese 2020). To caution against concerns that the identified empirical patterns are purely a result of broader societal and political trends, Fig. 1 also plots, as dotted lines, average sizes of negotiating delegations in Synthesis Report approval sessions in AR5 and AR6. Relying on the fact that Synthesis Reports were negotiated face-to-face in both assessment rounds, we find that delegation size has grown only minimally between the two ARs’ Synthesis Report approvals, which suggests a modest increase in delegation size for reasons that are plausibly unrelated to the virtual format. We also note that the average delegation size remains remarkably similar across all four AR5 approval sessions, which minimizes concerns that the approval of the Synthesis Report draws systematically larger or smaller delegations than the WG approvals. We are, hence, confident that the increase in member governments’ delegation sizes results largely from changes in meeting formats rather than the increased salience of climate politics. If the latter was indeed the case, we would expect much larger average delegations in the approval of the Synthesis Report of AR6.

We quantify the effect of virtual negotiations on countries’ average delegation size in a linear regression model using the difference-in-differences estimator. Table 1 summarizes the results when we pool data across Working Groups (Model 1) and when estimating the models separately for each Working Group (Models 2−4). The four models correspond to the four panels in Fig. 1 above. The coefficient estimate in the top row (AR\(\times \)WGI-III) shows the effect of virtual negotiations on delegation size as an increase of between 1.2−2.0 delegates on average. This effect is strongest for WGII (Model 3) and WGIII (Model 4), and all estimates are statistically distinguishable from zero at conventional levels of significance.

Lending further credibility to our empirical strategy, we find that delegation sizes for Synthesis Report approval sessions in AR5 and AR6 were not statistically significantly different (as indicated by the AR6 estimates). Similarly, delegations in Working Group and Synthesis Report approval sessions in AR5 were roughly of the same size (as indicated by the WGI-III estimate). This strengthens claims that the observed increase in delegation size indeed stems from virtual session formats because changes in delegation size between AR5 and AR6 only occurred for exactly those approval sessions that happened online (i.e., WGI−III approvals), but not for the ones that were conducted face-to-face (i.e., Synthesis Report approvals).

4.2 Delegation sizes and engagement levels in AR6

Building on the above finding that delegation sizes increased in AR6, we now examine whether larger delegations translate into greater engagement levels in IPCC negotiations. We provide a first answer to this question by describing the relationship between countries’ delegation sizes and delegations’ engagement levels, as approximated by country mentions in ENB reports. Figure 2 plots the distributions for both variables for the 30 largest countries by delegation size. For each of the three Working Groups in AR6, the bar plots to the right show a country’s delegation size; the “\(\times \)” marks delegation sizes in AR5 for comparison. Colors denote gender breakdowns, where female and male delegates are shown in red and blue, while gray indicates delegates whose gender we could not assign based on information about their first names. Bar plots to the left show countries’ engagement levels.

Confirming what we described in the previous section, delegations in AR6 were substantially larger compared to AR5 for almost all countries in the figure. With the exception of WGII delegations of Japan (53 delegates in AR5, 17 delegates in AR6) and Saudi Arabia (9 delegates in AR5, 7 delegates in AR6) and WGIII delegations of Germany (22 delegates in AR5, 17 delegates in AR6) and China (16 delegates in AR5, 15 delegates in AR6), AR6 delegations became larger for all top-30 countries.^{Footnote 4} Among our 30 largest countries, delegations in WGI had an average size of 9.6 delegates, which was significantly smaller than delegations in WGII (12.8 delegates, \(p<0.025\)) and WGIII (13.8 delegates, \(p<0.006\)). Many countries therefore seem to have used the online setting as a way to increase their presence at IPCC negotiations.

Delegation sizes and engagement levels for 30 largest countries (by delegation size) in AR6 for WGI−III. Colors indicate gender breakdowns for each of the delegations, where we distinguish between female delegates (red), male delegates (blue), and those for whom we could not assign gender based on first name information (gray); “\(\times \)” marks delegation size in AR5 for comparison. Note: Japan nominated 53 delegates for WGII in AR5 (as it hosted the approval session in Yokohama). We omit this count in the plot for ease of visualization

However, delegation size does not directly translate into engagement levels as measured by country mentions in the ENB reports. While the 30 largest countries account for 77% (537 of 695 mentions in WGI), 74% (616 of 831 mentions in WGII), and 80% (1,145 of 1,424 mentions in WGIII) of mentions compared to all other countries, and hence account for a vast majority of total interventions, considerable variation exists within this diverse set of states. As indicated by the gray histograms, countries at the top of the list in Fig. 2 do not necessarily engage more in the approval sessions. This pattern is robust across Working Groups.

India, Saudi Arabia, and the United States are consistently the countries with the largest number of interventions, accounting for about one third of total mentions in each of the Working Groups (32%, 221 total mentions in WGI; 31%, 254 mentions in WGII; 35%, 500 mentions in WGIII). While mentions do not tell us anything about the direction or success of the interventions, these data indicate that these countries are actively involved in shaping SPM text. Other countries that engaged heavily were Germany (50 mentions) and the UK (36 mentions) in WGI, Norway (46 mentions) and France (36 mentions) in WGII, and Germany (95 mentions) and Norway (91 mentions) in WGIII. At the same time, this also means that several countries with sizable delegations remained largely silent, such as, for instance, Argentina, Indonesia, Malaysia, Turkey, and Vanuatu. Similar to a conclusion reached for IPCC plenary sessions (Hughes 2022, 2023) and authorship contributions (Hughes and Paterson 2017), these results identify a core group of IPCC member countries actively involved in approving the key findings of the AR6 Working Group reports. We also notice that the number of interventions in Working Group III on mitigation options was about twice the number of interventions in the other two Working Groups on the physical science basis and climate impacts and adaptation.

4.3 Engagement levels and time zone differences

So far, we have shown that large delegations are not synonymous with high engagement levels as captured by country mentions in ENB reporting; and, in fact, in some instances smaller delegations were more actively involved in the approval than larger delegations. This may not come as a surprise, as other factors, like a country’s research capacity, its domestic economic and political constraints and priorities, or its vulnerability to climate impacts may be more important drivers of engagement levels in IPCC approval sessions. Nevertheless, our results document stark differences in levels of country engagement. Clearly, the “grueling” schedule of the meeting (ENB 2022, 23), with longer days as the approval sessions progressed, did not have the same impact on all delegations.

In order to better understand country-level variation, we turn to the role of time zone differences. Despite the IPCC’s efforts to recognize time zone differences in scheduling meeting sessions (IPCC 2022, 2), negotiators and observers alike complained heavily about this particular feature of the online negotiation sessions in AR6. As delegations attended virtual meetings from their own respective time zones, they often experienced negotiations that stretched far beyond standard work hours late into the night and early mornings, disrupting delegates’ personal life and resulting in fatigue and exhaustion (Chasek 2021; Vadrot and Ruiz Rodriguez 2022). Notwithstanding that in-person meetings also run long hours, this problem was especially acute for delegations in time zones that were the most distant from Europe.

To analyze the effect of time zone differences on country participation during the approval, we first convert the day-by-day negotiation schedules for WGII (14−26 February 2022) and WGIII (24 March−4 April 2022) approval sessions from Coordinated Universal Time (UTC) into each delegation’s home time zone.^{Footnote 5} We then compute, on an hour-by-hour basis, whether negotiations took place during or after standard 9am-5pm office hours in a delegation’s home time zone. Aggregating these data up to the country level provides us with a measure that captures the share of negotiation hours that fall outside of each delegation’s normal office hours.

In Fig. 3 we plot this measure for the 30 delegations that were the most exposed to time zone differences. The bar plots show the share of negotiation hours outside of normal work hours. To illustrate, for the delegation of Vanuatu only eight out of a total of 106 negotiation hours in WGII approval sessions took place during normal 9am-5pm work hours. Over the duration of two weeks of virtual negotiations, Vanuatu delegates worked outside of normal hours more than 92% of the time. This is but one example since other countries in the Pacific region, including Australia, South Korea, Japan, the Cook Islands, Samoa, New Zealand, and Kiribati experienced similarly high shares outside core working hours. The same, albeit to a slightly smaller degree, is true for countries in East and Central Asia as well as in Latin America, whose delegations participated in the approval at least half of the time outside of their 9am-5pm work hours.

Share of negotiation hours outside of 9am-5pm office hours in AR6 by country

While larger country delegations are undeniably more likely to engage more actively in negotiations, purely as a result of greater numbers, we also expect that this effect might dissipate when delegations are located in geographies with unfavorable time zones. Indeed, countries furthest away from Europe, which expected negotiations to take place out of core 9am-5pm work hours for most of the time, may purposefully have nominated larger delegations to mitigate these negative effects on engagement levels.

Marginal effect plots of delegation size on engagement levels as a function of the share of negotiation hours that fall outside 9am-5pm office hours. Black lines indicate estimates from a linear regression model with 95% confidence bounds shown in gray for pooled data (left panel), WGII (middle panel) and WGIII (right panel). Dots and vertical whiskers are point estimates and confidence intervals from a non-linear binning estimator (Hainmueller et al. 2019). The histogram along the x-axis shows the distribution of the data

We model this conditional effect with an interaction regression model and show the estimated relationships in Fig. 4 for pool data (left panel) and separately for WGII (middle panel) and WGIII (right panel).^{Footnote 6} Black lines report the effect of an additional delegate on countries’ engagement levels (with associated 95% confidence intervals shown in gray) for different levels of our measure of time zone exposure from linear regression models. Dots and vertical whiskers produce the same effects of interest from a non-linear binning estimator (Hainmueller et al. 2019).

Across the board, we observe that larger delegations are associated with greater engagement levels, so the “strength in numbers” logic finds support in our data in general and for both Working Groups individually. However, this positive relationship attenuates as time zone differences become pronounced. This becomes clear when comparing estimates along the horizontal axis of each of the panels in Fig. 4. Estimates are positive when the shares of the session hours outside of core work hours are small (left end of x-axis), such as for European delegations. However, these effects are statistically no longer different from zero for delegations in time zones with very large shares of hours outside of 9am-5pm work hours (right end of x-axis) as in the case of delegations from the Pacific and Latin American regions. In other words, while increasing delegation size tends to increase delegations’ engagement levels, this is much less the case for delegations that are located in time zones that are greatly different from the time zone that IPCC meetings take place in. These results are correlational because delegations in remote time zones will clearly have expected this effect, but they nonetheless shed important light on the conditional impacts of time zones on countries’ abilities to substantially and meaningfully engage in the virtual approval of the key findings of the AR6.