Abstract
Objectives. We analyzed demographic and social network variables associated with the timing of ratification of the Framework Convention on Tobacco Control (FCTC).
Methods. We compiled a 2-mode data set that recorded country participation in FCTC negotiations, as well as the number of individuals per country per year who joined an online tobacco control network. We used logistic regression analysis of these 2 data sets along with geographic location to determine whether exposure to prior FCTC adoptions was associated with a country's likelihood of adoption.
Results. In the logistic regression analysis, higher income and more nongovernmental organizations (NGOs) involved in the Framework Convention Alliance (a network dedicated to the FCTC) were associated with being among the earliest adopters (for income, adjusted odds ratio [AOR] = 2.41; 95% confidence interval [CI] = 1.55; for NGOs, AOR = 1.66; 95% CI = 1.26, 2.17) or among early adopters (for income, AOR = 1.42; 95% CI = 1.09, 1.84; for NGOs, AOR = 1.23; 95% CI = 1.03, 1.45). Network exposure and event history analysis showed that in addition to income, the likelihood of adoption increased with increasing affiliation exposure to FCTC adopters through GLOBALink (an online network facilitating communication between tobacco control advocates).
Conclusions. Public health programs should include a plan for creating opportunities for network interaction; otherwise, adoption and diffusion will be delayed and the investments in public health policy greatly diminished.
Although the risks of tobacco smoking have been known for decades, the pandemic of tobacco use continues. There are now an estimated 1.3 billion smokers worldwide, along with millions more who use various oral tobacco products.1 Tobacco is the leading cause of preventable death worldwide, resulting in about 6 million deaths per year.2 Despite great progress in tobacco control, primarily in North America and Western Europe, the number of tobacco-attributable deaths is projected to grow substantially during this century, especially in low- and middle-income countries.
In 1999, in recognition of the shift and growth in tobacco consumption and the potential for an enormous future burden of death and disease, the World Health Organization (WHO) member states initiated formal negotiations on an international treaty aimed at reducing this global threat. The Intergovernmental Negotiating Body (INB), which was charged with negotiating the text of the treaty, held 6 formal negotiating sessions in Geneva between 2000 and 2003. Over 170 states sent at least 1 delegate to 1 of the INB sessions. Scientific experts and representatives of advocacy networks also attended the negotiations, where they held seminars on technical aspects of the convention and distributed information to delegates. GLOBALink, an online network internationally recognized for facilitating communication between tobacco control advocates, was one such network.3 In addition to the INB sessions, countless regional negotiating sessions and technical conferences were convened during the period.
In May 2003, the 56th World Health Assembly unanimously adopted the WHO Framework Convention on Tobacco Control (FCTC).4 The key provisions include a comprehensive ban on tobacco advertising, promotion, and sponsorship; a ban on misleading descriptors intended to convince smokers that certain products are safer than “standard” cigarettes (for example, the term lights in Marlboro Lights); and a mandate to place rotating warnings that cover at least 30% of tobacco packaging. The FCTC also encourages countries to implement smoke-free workplace laws, address tobacco smuggling, and increase tobacco taxes. The FCTC entered into force on February 27, 2005, 90 days after the 40th member state ratified the treaty. Further ratifications or its legal equivalent (acceptance or approval) continued over the next 4 years. As of May 2009, 168 countries were party to the treaty.
The institutionalization of tobacco control within the WHO and the subsequent ratification of the FCTC by nearly all WHO member states provides an opportunity to analyze the system and network dynamics that facilitate global tobacco control diffusion. Diffusion refers to the process by which an innovation is communicated through certain channels over time among members of a social system.5 The premise, which has been confirmed by empirical research, is that new ideas and practices spread through interpersonal contacts largely consisting of interpersonal communication.5,6 Other researchers have investigated specifically how social networks provide the channels through which new ideas and practices, such as the FCTC, are spread.5–7
Given studies of diffusion in other contexts, we hypothesized that the global diffusion of the FCTC has been partly driven by interpersonal communication and networking developed throughout the negotiation of the FCTC and facilitated through existing global tobacco control networks. In other words, we hypothesized that the extent of a country's participation in the FCTC negotiations and its citizens’ involvement in international tobacco control networks would be associated with early or late FCTC ratification.
However, we also expected the predictability of these social network variables to be impacted to some extent by the structural and demographic aspects of states (e.g., location, population, income level, degree of political freedom, tobacco prevalence, and tobacco production). For example, a country with a high smoking prevalence may perceive tobacco control as more important and ratify sooner than a country with low prevalence. Conversely, a tobacco-producing and exporting country may view tobacco control as a threat to its financial success and resist ratification. Here we analyzed the structural, demographic, and social network variables that led individual countries to ratify the FCTC when they did and made a first attempt at specifying the driving forces behind global tobacco control diffusion.
The timing of the adoption of the FCTC was measured as the day the country ratified the FCTC recoded to the month (coding adoption of the FCTC monthly made the data manageable). The ratification dates are available on the WHO's Tobacco Free Initiative Web site.8 In addition, 2 dummy variables were created to indicate the earliest adopters (first 30 ratified, 15.5%) and the early adopters (first 95 ratified, 49.2%). The decision to divide the countries by the first 16% and first 50% was based on past diffusion research that categorized adopters on the basis of time of adoption. According to Rogers, early adopters are the first 16% and early majority are the first half.5 Several structural characteristics of the countries were obtained, including population,9 gross national income,9 degree of political freedom,10 tobacco production (in tons),11 and current male and female smoking prevalence.12
Country participation in the negotiations was measured by the number of delegates sent to each INB session published in the official list of participants. These are publicly available on the WHO Governing Bodies Web site.13 Increases in interaction between domestic and international advocates for tobacco control were measured through the number of new GLOBALink members per country per year between 1993 and 2006 and the number of nongovernmental organizations (NGOs) per country participating in the Framework Convention Alliance, a network dedicated to the FCTC in 2009. GLOBALink staff provided the GLOBALink annual membership database to H. L. W. in 2007. The specific dates for individual GLOBALink memberships were not available and were undoubtedly scattered throughout the year. Consequently, memberships were updated as of December for each year. Framework Convention Alliance membership was recorded from the Framework Convention Alliance Web site in February 2009. Missing data were recoded to the mean. The number of countries for which we imputed missing values was 19 for gross national income, 70 for tobacco production, 31 for current male smoking prevalence, and 27 for current female smoking prevalence. Missing data analyses showed that the countries with any missing data (95 of 193) did not have differential adoption dates (odds ratio [OR] = 1.00; P = .73), but had a lower level of log-scaled population (OR = 0.83; P < .001), had a lower level of log-scaled tobacco production (OR = 0.76; P < .05), and had a lower level of NGOs participating in the Framework Convention Alliance (OR = 0.67; P < .001).
To determine the factors associated with ratification of the FCTC, 4 regression models were estimated: (1) an ordinary least squared regression on time of ratification (reversed), (2) a logistic regression on being an earliest ratifier (first 15.5%), (3) a logistic regression on being an early ratifier (first 49.2%), and (4) an event-history model using logistic regression for the likelihood of adoption at each month. In the 3 logistic regressions (models 2–4), the numeric value representing the number of delegates participating in each INB per country and new members in GLOBALink per country per year were included.
The 3 network diffusion terms calculated in the event-history model (model 4) were: (1) distance, (2) INB participation, and (3) GLOBALink membership. Reverse distances between country capitals were calculated, and this weight matrix was used to estimate contiguity effects. Affiliation matrices of the INB and GLOBALink data for each country were transposed and postmultiplied to generate adjacency matrices.14–16 Each element in the adjacency matrix represented the number of people from any 2 countries that jointly attended an INB session or enrolled in GLOBALink. These coaffiliation matrices were used to generate vectors of affiliation network exposures that measure the level of exposure to prior ratifiers for each country. These matrices constitute the network within which the network diffusion analyses were conducted.7 The matrix and time of ratification variable were used to construct the 3 network exposure terms that were time varying (as each network partner adopted FCTC, it increased the focal country's exposure). Time-constant variables, the country characteristics, were also included in the analyses.7
The model tested in the event-history model (model 4) is based on country (i) and time (t):15
where yit is the binary indicator of FCTC ratification for country i (i = 1, .., N) at time t, α is the intercept, βj is the parameter estimate for vectors of j (j = 1, … ,k) characteristics (Xji) of country i, and ρl is the parameter estimate for the time-varying network exposure variables [ωil]yt. The ωil network weight matrices are defined as one of the following ωi matrices on the basis of distance (l = 1), INB participation (l = 2), or GLOBALink (l = 3). Note that in this equation, ωil was constant over time for the first 2, but varied over time for GLOBALink. To estimate this equation, the generalized estimating equation was used with an autoregressive correlation structure with single lag and logit link function.17 Network figures were created with the software program Netdraw.18
Characteristics of the countries and the mean and median date of FCTC ratification globally and by region are shown in Table 1. The first country to ratify the FCTC was Norway in June 2003 followed by Fiji in November 2003. FCTC ratification continued to spread (Figure 1) between countries over the next 6 years until ultimately only 33 of 193 countries (16.9%) resisted ratification of the FCTC (including the United States) during the time period. The mean and median dates of ratification for all ratifying countries were March 2006 and November 2005, respectively (approximately 2.5 years after the last INB session). Populous and less populous countries ratified the FCTC at approximately the same rate.
TABLE 1 Characteristics of the Countries and the Mean and Median Date of Ratification of the Framework Convention on Tobacco Control (FCTC)
| Value | |
| Date FCTC ratified | |
| Mean month (mean no. months until adoption globally) | March 2006 (35.4) |
| Median month (median no. months until adoption globally) | November 2005 (31.0) |
| Mean no. of months until adoption (from May 2003) | |
| Africa | 44.0 |
| Americas | 46.8 |
| Eastern Mediterranean | 41.0 |
| Europe | 39.8 |
| Southeast Asia | 26.5 |
| Western Pacific | 25.2 |
| Population size, no. (range) | 29 121 813 (81 000–1 300 000 000) |
| Income, % | |
| Low income | 26.50 |
| Lower middle income | 27.50 |
| Upper middle income | 20.60 |
| High income | 25.40 |
| Democracy, % | |
| Partially free | 30.80 |
| Free | 46.70 |
| Tobacco production, tons (range) | 54 406 (30–2 685 743) |
| Smoking prevalence, % | |
| Total | 23.70 |
| Male | 34.80 |
| Female | 13.10 |
| No. tobacco NGOs, mean (range) | 1.57 (1–28) |
| No. participants in INB, mean (range) | |
| October 18, 2000 | 2.62 (0–13) |
| May 5, 2001 | 2.83 (0–18) |
| November 25, 2001 | 3.09 (0–23) |
| March 23, 2002 | 2.97 (0–26) |
| October 21, 2002 | 3.30 (0–21) |
| February 26, 2003 | 3.58 (0–20) |
| No. new GLOBALink members, mean (range) | |
| 1993 | 0.26 (0–8) |
| 1994 | 0.08 (0–4) |
| 1995 | 0.05 (0–3) |
| 1996 | 0.28 (0–19) |
| 1997 | 1.85 (0–147) |
| 1998 | 1.80 (0–102) |
| 1999 | 1.02 (0–24) |
| 2000 | 3.03 (0–176) |
| 2001 | 2.82 (0–125) |
| 2002 | 3.78 (0–219) |
| 2003 | 4.54 (0–255) |
| 2004 | 3.69 (0–250) |
| 2005 | 4.35 (0–268) |
| 2006 | 4.09 (0–199) |
Note. INB = Intergovernmental Negotiating Body; NGO = nongovernmental organization.
FIGURE 1 Global diffusion of the Framework Convention on Tobacco Control.
Note. The y-axis represents the cumulative percentage of countries that have ratified the Framework Convention on Tobacco Control.
The ratifying countries were nearly evenly split among the 4 income categories. Over 77% of the ratifying countries were considered politically free or partially free with an overall smoking prevalence of 23.7% (the smoking rate among men was considerably higher, 34.8%, than that among women, 13.1%). The number of delegates sent to the first INB session (October 2000) ranged from 0 to 13 and averaged 2.62. By the final INB session (February 2003), the number of delegates ranged from 0 to 26 and averaged 3.59. The average number of new GLOBALink members per country varied by year and generally increased, although not monotonically. By the end of 2006, membership in GLOBALink by country ranged from 0 to 1602, with a mean of 31.8. Countries in the Western Pacific and Southeast Asia had the earliest ratification times (25.2 months and 26.5 months, respectively), whereas Africa and the Americas were later (44.0 months and 46.8 months, respectively; Table 1).
The results of the 4 models are presented in Table 2. The initial ordinary least squared regression on reversed time of ratification (to measure innovativeness) indicated that higher income level (B = 0.26; P < .01) and NGO membership in the Framework Convention Alliance (B = 0.27; P < .05) were the only variables associated with earlier adoption of the FCTC. The earliest (first 15.5%) and early (first 49.2%) adoption models using logistic regression analyses showed that income was again associated with being among the earliest adopters (adjusted odds ratio [AOR] = 2.41; 95% confidence interval [CI] = 1.55, 3.74) or among the early adopters (AOR = 1.42; 95% CI = 1.09, 1.84), as well as NGO membership in the Framework Convention Alliance (for earliest adopters, AOR = 1.66; 95% CI = 1.26, 2.17; for early adopters, AOR = 1.23; 95% CI = 1.03, 1.45).
TABLE 2 Regression Results for Adoption of the Framework Convention on Tobacco Control (FCTC)
| OLS Regression, No. or B | Logistic Regression Earliest Adoption, No. or AOR (95% CI) | Logistic Regression Early Adoption, No. or AOR (95% CI) | Event History Analysis Including Exposure, No. or AOR (95% CI) | |
| Total | 193 | 193 | 193 | 6833 |
| Log of population | −0.12 | 0.92 (0.78, 1.08) | 0.93 (0.83, 1.04) | 0.89 (0.79, 1.01) |
| Log of income | 0.26** | 2.41*** (1.55, 3.74) | 1.42** (1.09, 1.84) | 1.17 (0.98, 1.41) |
| Democracy, partially free | −0.05 | 0.81 (0.24, 2.71) | 0.70 (0.30, 1.63) | 1.02 (0.65, 1.60) |
| Democracy, free | −0.08 | 0.78 (0.26, 2.35) | 0.70 (0.31, 1.57) | 0.87 (0.55, 1.37) |
| Tobacco production (in tons) | 0.00 | 0.94 (0.76, 1.17) | 1.04 (0.90, 1.21) | 0.97 (0.89, 1.05) |
| Smoking prevalence, male | 0.14 | 1.02 (0.98, 1.06) | 1.03 (1.00, 1.06) | 1.00 (0.98, 1.02) |
| Smoking prevalence, female | 0.03 | 0.98 (0.93, 1.03) | 1.00 (0.97, 1.04) | 1.01 (0.98, 1.03) |
| No. NGOs in FCA | 0.27* | 1.66*** (1.26, 2.17) | 1.23* (1.03, 1.45) | 1.00 (0.93, 1.08) |
| Region (Africa ref) | ||||
| Americas | 1.09 (0.46, 2.58) | |||
| Eastern Mediterranean | 1.67 (0.48, 5.82) | |||
| Europe | 1.19 (0.40, 3.55) | |||
| Southeast Asia | 4.02* (1.14, 14.13) | |||
| Western Pacific | 4.77*** (2.08, 10.92) | |||
| No. new GLOBALink members | −0.23 | 0.97 (0.95, 1.00) | 1.00 (0.99,1.00) | 1.05 (0.96, 1.15) |
| Exposure based on INB | 2.11 (0.40, 11.06) | |||
| Exposure based on GLOBALink | 2.92* (1.25, 6.78) | |||
| Exposure based on geography | 1.51 (0.33, 6.88) |
Note. AOR = adjusted odds ratio; CI = confidence interval; FCA = Framework Convention Alliance; INB = Intergovernmental Negotiating Body; NGO = nongovernmental organization; OLS = ordinary least squared.
*P < .05; **P < .01; ***P < .001, by the 2-tailed test.
The event-history model included the network exposure terms and was an event history analysis using the generalized estimating equation in which each case in the data set (N = 6833 [34 × 193] average months until adoption by total countries) represented a country-year up to and including the year of ratification. Robust variance estimates were computed at the country level. Country characteristics represented time-constant variables and exposures were time-varying based on the ratification behavior of the countries’ coaffiliations. The analysis included the 33 never-ratified countries with time of ratification coded as the 72nd month. It is customary in diffusion research to recode the nonadopters as the last time of adoption provided the number of cases is not excessive. We repeated the analysis excluding nonratifying countries with no change in the results.
The results showed that exposure based on geography was positive but not statistically significant (AOR = 1.51; 95% CI = 0.33, 6.88). Exposure based on INB participation was positive but not statistically significant (AOR = 2.11; 95% CI = 0.40, 11.06). Exposure based on increasing GLOBALink membership was positive and statistically significant (AOR = 2.92; 95% CI = 1.25, 6.78). The dummy variables for region indicated that the Western Pacific and Southeast Asia regions had earlier ratification dates (AOR = 4.77; 95% CI = 2.08, 10.92) than did Africa (AOR = 4.02; 95% CI = 1.14, 14.13) and no other regions were statistically significantly different from Africa. The AOR of 2.92 for the GLOBALink effect indicated that a country was nearly 3 times as likely to ratify the FCTC once their exposure to other ratifiers via membership in GLOBALink reached 100%.
To illustrate these diffusion effects, we graphed the network of GLOBALink affiliations for the earliest ratifying countries in Figure 2. This network was densely connected with few isolates (Naru, San Marino, and Seychelles). The core of the network contained countries from all over the globe but also clearly showed considerable coaffiliation on GLOBALink membership, which permitted communication about tobacco control. By contrast, the network of the 33 countries who have not yet ratified the FCTC are shown in Figure 3. In this network, 18 of the countries were isolates, indicating that they had no participants in GLOBALink.
FIGURE 2 Network of the 30 countries that were the earliest to adopt the Framework Convention on Tobacco Control (first 15.5% of all countries), with links indicating the magnitude of comembership on GLOBALink.
Note. Figure created using Netdraw.18 For example, Norway and India had at least 1 member (although not necessarily the same individual) who subscribed to GLOBALink for at least 9 years from 1993 to 2005. Conversely, Jordan and Palau only had 1 member during the same period. There were 5 outlying countries. Thicker lines indicate stronger links on GLOBALink.
FIGURE 3 Network of the 33 countries that did not adopt the Framework Convention on Tobacco Control, with links indicating the magnitude of comembership on GLOBALink.
Note. Figure created using Netdraw.18 There were 18 outlying countries. Thicker lines indicate stronger links on GLOBALink.
These analyses indicate that wealthier countries were more likely to ratify the FCTC earlier than were poorer countries, and ratification dates differed by geographic region. The only other factor that influenced adoption behavior was increasing membership in GLOBALink. Countries were more likely to ratify the FCTC after more of their citizens joined GLOBALink and as numbers of GLOBALink members from ratifying countries grew.
We analyzed the demographic and social network variables that led countries to ratify the FCTC at the time that they did. We compiled a 2-mode data set that recorded country participation in the negotiation of the FCTC. We also recorded the number of individuals by country who joined GLOBALink, an interactive online network created to support tobacco control advocacy. These 2 data sets, along with physical location (latitude and longitude), were used to determine whether exposure to FCTC-related information and prior FCTC adoptions was associated with a country's likelihood of ratification.
The initial diffusion of FCTC ratification was not particularly rapid. The first year consisted of a few scattered countries, none particularly populous, ratifying the FCTC. The evidence shows that contiguity was not a major factor in these early ratifications. Thus, the evidence does not indicate that countries that ratified early served as role models to which other countries could readily point as motivations for FCTC ratification. The initial delay in ratification may have been a result of the fact that ratification required the completion of domestic policy processes. Given competing priorities and legislative schedules, it may simply have been impossible for some countries to ratify any quicker than they did. There was an increase in the rate of adoption in years 2 through 4, after which it leveled off. In retrospect, we can see that the FCTC is among the quickest treaties to enter into force, and has become among the most widely ratified treaties in existence.
Most country attribute variables did not have a significant effect on time of ratification. One exception was the region in which a country was located. Southeast Asia and the Western Pacific ratified the FCTC earlier than did countries in the Americas and Africa. Participation in the INB was not associated with early ratification of the FCTC nor was an increase in the number of delegates sent by a country. Being part of the negotiation process did not accelerate ratification once the FCTC was completed.
Besides income level, network membership—as measured by both NGO participation in the Framework Convention Alliance and increasing membership in GLOBALink—was the only significant predictor of early ratification. Increasing membership in GLOBALink, a network in which information and experiences related to the FCTC ratification were shared, was also associated with ratification once the diffusion process got under way. The graphic comparison of network interactions presented in Figures 2 and 3 clearly illustrates the difference between countries that ratified early and those that ratified late in regard to membership in GLOBALink. The core of the network among nonratifying countries consists of the United States, Switzerland, Argentina, and Indonesia, all of whom may have reasons other than lack of awareness or information about the treaty for postponing ratification of the FCTC. Generally speaking, most countries that have not ratified are outside the flow of information and influence (as characterized by GLOBALink membership) that may otherwise enable them to ratify the FCTC. This is an interesting finding because it points to the critical role that international information sharing may have on domestic policy processes and suggests that investment in such networks may provide a cost-effective method for supporting the spread of international public health norms in other areas.
Our study had several potential weaknesses. Social, political, and cultural characteristics of states are not easy to quantify and there are always challenges in selecting appropriate variables. There is always the possibility that other, and perhaps better, variables could have been used to represent the determinants of ratification. For a number of countries, information on the variables selected was unavailable, including a lack of GLOBALink data after 2006. The analysis also ignored the strength of tobacco control legislation in countries and whether it changed during the period. The goal of this analysis, however, was simply to provide greater information to identify possible determinants of FCTC diffusion. Future work is needed to improve on the empirical model and to strengthen its explanatory power.
The role of system modeling in global tobacco control must also be weighed, as is the case with all large-scale systems modeling efforts, against qualitative information available about the process globally and within specific countries.19 Historical context and political realities within countries cannot adequately be captured in such a macrostudy. In future studies, contextual information could be used to a greater extent to add depth and balance to the analysis; this would provide a more complete picture of the role that the FCTC process and international tobacco control networks have had in diffusing international tobacco control norms to domestic policy.
This study has implications for the future study of the diffusion of innovations and public health. Historically, it has been prohibitively challenging to conduct large-scale studies in which social network and time of innovation adoption data are available. Generally, this is because the time span for diffusion is long and data collection at multiple time points becomes prohibitively expensive.5,20 The public availability of the FCTC documentation and ratification dates, global databases, including the WHO Report on the Global Tobacco Epidemic9 and the Tobacco Atlas,12 and the existence of tobacco control networks, including GLOBALink and the Framework Convention Alliance, provide an incredibly rich source of data for this and future diffusion studies.
The initiation of the FCTC process marked the first time that the WHO member states enacted the organization's power under article 19 of its constitution to negotiate and sign a binding treaty aimed at protecting and promoting public health. It also represented the first time that the member states cooperated worldwide in a collective response to prevent chronic disease. Considerable time and effort have been invested in the negotiation, ratification, and domestic implementation of the FCTC. The evidence presented here suggests that the speed of FCTC ratification was dependent on forums in which invested individuals within past and potential ratifying countries could exchange information, learn about experiences, and gain reassurance about the consequences of action. Future public health programs should thus be accompanied by a plan for creating opportunities for this interaction; otherwise, diffusion of internationally promoted programs and policies will be delayed and the return on investments diminished.
Acknowledgments
Support for this study was provided through an Advancing Scholarship in the Humanities and Social Sciences grant to H.L. Wipfli from the University of Southern California.
Human Participant Protection
No protocol approval was necessary because data were obtained from secondary sources.
