The effects of outdated data and outliers on Kenya's 2019 Global Food Security Index score and rank

The availability of updated and complete data is one requirement for building a robust composite indicator (Freudenberg, 2003b). However, outdated data and the presence of outliers in databases challenge the process of building robust composite indicators (Nardo et al., 2005). Outdated data and outliers can occur when using data constructed from surveys or when data is obtained from national or international statistical sources (Giovanni, 2014). Outdated data are barriers to disclosing knowledge, while outliers can act as unintended benchmarks in composite indicators (Leys et al., 2013; Solaro et al., 2017). When not correctly handled, outdated data and outliers may affect findings and lead to biased results in benchmarking exercises (Santeramo, 2017; JRC-EC, 2008). A composite indicator's ability to represent multidimensional concepts is mainly determined by the quality and accuracy of the indicators used in its construction (Nardo et al., 2005). The indicators for constructing a composite index should be specific, measurable, achievable, relevant and time-bound (Santeramo, 2015b). However, the selection of indicators is often affected by the lack of updated data due to missing current data values at the national or global level for specific countries (Thomas et al., 2017). Outdated indicators could be updated or replaced while outliers must be identified and removed statistically, for example, by winsorisation (JRC-EC, 2008; Santeramo, 2015a). Over time, several indicators have been used to measure the concept of food security. Some indicators measure food security determinants, such as the sufficiency of supply. In contrast, other indicators measure food security outcomes such as an individuals' nutritional status or the mortality rate of children under five years of age (Jones et al., 2013; Coates, 2013). Significant variations exist among the food security indicators. Some indicators are used for monitoring or evaluation processes, while others are used in early warning systems (Carletto et al., 2013). Moreover, some indicators only measure a single dimension of food security, such as access or food availability, as isolated contributing factors to food insecurity (Barrett, 2010). However, food security dimensions (access, availability, stability and utilisation) are hierarchal (Barrett, 2010). Food availability is necessary but not sufficient to guarantee access, while access to food (economically, physically or socially) is also necessary but does not guarantee food utilisation (Barrett, 2010). Overall, stability cuts across the access, availability and utilisation dimensions and is essential at all times to ensure food security in a country (Carletto et al., 2013; Thomas et al., 2017). The heterogeneity of the food security indicators raises the need for composite indicators to synthesise the information from different indicators (Santeramo, 2015b). Composite indicators can summarise information from different indicators and give a comprehensive representation of a country's food security status. Composite indicators can also integrate large amounts of data into a summarised unique score, which is essential to rank countries in benchmarking exercises (Freudenberg, 2003b). Moreover, composite indicators are a useful tool in policy making processes and public communication due to their ease of interpretation (Nardo et al., 2005). However, the robustness of a composite indicator can be affected by the subjectivity of methods used in its construction process, such as the weighting methods (JRC-EC, 2008). Some of the methods used when constructing composite indicators are easily manipulated to support desired policies (Freudenberg, 2003b; Mazziotta and Pareto, 2013). Therefore, a composite indicator's construction process must be transparent not to offer misleading information to its users (Freudenberg, 2003b). Research on how to improve the methodologies used in constructing a composite indicator and precise documentation of the steps is necessary to ensure transparency, especially the methods of handling outdated data, outliers, missing data and the weighting methods (Saisana and Saltelli, 2011). The Global Hunger Index (GHI), the Global Food Security Index (GFSI) and the Coping Strategy Index (CSI) are some examples of composite food security indicators (IFPRI, 2019; Pangaribowo et al., 2013; EIU, 2012).