Appendix B: The Global Risks Perception Survey 2014 and Methodology
As discussed in Part 1, the Global Risk 2015 methodology was reviewed last year for this edition. A number of workshops, interviews and discussions were held with experts and with the Advisory Board, taking into account lessons from past editions as well as developments in the global risks landscape. The concept of trends was introduced in both the survey and the report and some of the global risks analysed in the past were relabelled and classified under a different category or better defined as trends.
The Global Risks Perception Survey (GRPS or survey) was adjusted accordingly to capture the main aspects of both risks and trends and to assess their interconnectedness and impact on societies. The following section describes the survey and methodology in greater detail.
The Global Risks Perception Survey
The Global Risks Perception Survey, discussed in Part 1, is the main instrument for assessing global risks and trends in this report. The survey was conducted between mid-July and the end of September 2014 among the World Economic Forum’s multistakeholder communities of leaders from business, government, academia and non-governmental and international organizations.1
Raw responses were cleaned to improve overall data quality and completeness. All questionnaires with a completion rate of below 50% were dropped, reducing the number of available responses from 1,120 to 896. In 12 cases, the respondent did not provide any information about gender, but it was possible to infer this information from the other records provided (first and last names). Similarly, 101 respondents did not indicate the region in which they are based and were manually assigned to one on the basis of their country of residence. Among the respondents, 43% completed the survey last year. Details of the sample composition are reported in Figure B.1.
The graph below shows the profile of the 896 survey respondents. To capture the voice of youth, the survey also targeted the World Economic Forum’s community of Global Shapers.2 Those under 30 years of age accounted for approximately one-fifth of respondents.
The Global Risks Landscape 2015 (Figure 1)
Respondents were asked to assess the likelihood and global impact of each of the 28 risks. For each risk, they were asked, “How likely is this risk to occur globally within the next 10 years?” and “What is the estimated impact globally if this risk were to materialize? (Impact is to be interpreted in a broad sense beyond just economic consequences)”. The possible answers ranged from 1 (“very unlikely” and “low” impact, respectively) to 7 (“very likely” and “high” impact, respectively). Respondents were given the possibility to choose a “Don’t know” option if they felt unable to provide an informed answer. For each risk, partial responses, i.e. those assessing only the likelihood or only the impact, were dropped. A simple average for both likelihood and impact for each of the 28 global risks was calculated on this basis. Formally, for any given risk i, its likelihood and impact, denoted as likelihoodi and impacti, respectively are:
where likelihoodi,n and impacti,n are respectively the likelihood and impact assigned by respondent n to risk i and measured on a scale from 1 to 7. Ni is the number of respondents for risk i who assessed both the likelihood and impact of that risk.
To draw the global risks interconnections map presented in Part 1, survey respondents were asked to answer the following question: “In your view, which are the most strongly connected risks? Please select at least three pairs and up to six pairs from the 28 risks below.”
In addition, respondents were asked to identify three to six pairs of strongly connected trends and risks, disregarding directions of causality. This question read: “In your view, which trends are the most strongly connected with risks? Please select at least three pairs and up to six pairs from the 13 trends below (you can select the same trend and risk more than once).” The information thus obtained was used to construct the risks-trends interconnections map included in the inside cover flaps of the report.
In both cases, a tally was made of the number of times each pair was cited. To obtain normalized connection weights, this value was divided by the count of the most frequently cited pair and, in addition, this ratio was square-rooted to dampen the long-tail effect (i.e. a few very strong links, and many weak ones) for display and presentation purposes. Of the 378 possible pairs of risks, 122, or 32%, were not cited. Similarly, of the possible 364 trend-risk combinations, 116, or 32%, were not cited. Formally, the intensity of the interconnection between risks i and j (or between trend i and risk j), denoted as interconnectionij, corresponds to:
where N is the number of respondents. Variable pairij,n is 1 when respondent n selected the pair of risks i and j as part of his/her selection. Otherwise, it is 0. The value of the interconnection determines the thickness and brightness of each connecting line in the graph, with the most frequently cited pair having the thickest and brightest line.
In the global risks and risks-trends interconnections maps (Figures 2 and 3), the area of each node (corresponding to a risk or a trend) is scaled according to the number of times the corresponding risk or trend was cited as a part of the connection pair.
The placement of the nodes was computed using ForceAtlas2, a force-directed network layout algorithm implemented in Gephi software, which minimizes edge lengths and edge crossings by running a physical particle simulation.3
The Risks of Highest Concern (Figure 1.1)
Although the report generally focuses on a time horizon of 10 years, respondents were asked to identify the risks of highest concern within two different time frames: 18 months and 10 years. To identify the top 10 global risks of highest concern, described in Part 1, respondents answered the following question: “In this survey, we are looking at risks within the next 10 years. For this question only, please select the five global risks that you believe to be of most concern within the next 18 months and 10 years, respectively.” For any given risk i from the list of 28 risks, the share of total respondents (N = 896) that declared being concerned about that risk was derived as follows:
with ci,n equal to 1 if respondent N selected risk i as a risk of concern and 0 otherwise. The risks with the 10 highest shares were selected as the risks of most concern.
Progress and Preparedness (Figure 1.7 and Figure 3.1)
Survey respondents were asked to identify up to three risks for which they felt most progress to address them has been made over the past 10 years. Similarly, they were asked to select up to three risks which they believed their region was least prepared for.4
For any given risk i from the list of 28 risks, the share of total respondents (N = 896) who think that most progress has been made in addressing that risk was calculated as follows:
with pi,n equal to 1 if respondent N selected risk i, and 0 otherwise.
Similarly, the share of total respondents who thought that their region is least prepared for risk i was obtained:
with ri,n equal to 1 if respondent N selected risk i, and 0 otherwise.