Development of a temporally harmonized asset index: evidence from across 50 years of follow up of a birth cohort in Guatemala

Table 5 Sensitivity analysis assessing associations of the harmonized index with alternate methodologies for creating the index, INCAP Oriente Longitudinal Study 1967–2018

Correlation with cross-sectional asset index (S1)
Study wave	Households	Rank correlation
1967	547	0.95
1975	755	0.95
1987	617	0.91
2002	820	0.97
2015–16	1075	0.97
2017–18	1145	0.96
2015–16	773	0.97
Urban 2015–16	302	0.90
Rural 2017–18	816	0.95
Urban 2017–18	329	0.92
Addition of newer assets to harmonized index^a (S2)
2002	820	0.91
2015–16	1075	0.91
2017–18	1145	0.92
Alternative specification of correlation matrix and factor extraction method (S4)
PCA with Polychoric correlation	4959	0.96
Exploratory Factor Analysis^b with Pearson correlation	4959	1.00
Exploratory Factor Analysis^b with Polychoric correlation	4959	0.96
Multiple Correspondence Analysis^c	4959	1.00
Alternative categorization of housing characteristics into high, medium and low (S5)
PCA with Polychoric correlation	4959	0.96
Exploratory Factor Analysis^b with Polychoric correlation	4959	0.92

^a Video player, sound system, computer, telephone, washing machine, and improved sewage system were included and imputed as 0 for those waves (1987 and before) during which it was not collected; ^b EFA with varimax rotation and 1 factor extracted using residual minimization; ^c MCA was calculated with crowding > 0.75 set to 1 and otherwise 0; All rank correlations are significant (p < 0.001). Details of the categorization schema for assets and housing characteristics are provided in Supplementary Table 1. Polychoric correlation was calculated using psych package (v1.8.10)

ISSN: 1471-2288