Exploring the construct validity of the social desirability scale of the south african personality inventory

Deacon, Genevieve (2016-12)

Thesis (MCom)--Stellenbosch University, 2016.

Thesis

ENGLISH ABSTRACT: Personality assessments are widely used in South Africa for predicting individuals’ future job performance (Barrick & Mount, 1996; Hough, Eaton, Dunette, Kamp, & McCloy, 1990; Ones & Viswesvaran, 2001). The manner in which an individual responds to a personality questionnaire is typically assumed to be an accurate and true reflection of their personality that can be compared with other individuals’ personality profiles (Van Herk, Poortinga, & Verhallen, 2004; Ziegler, Maccann, & Roberts, 2012). Personality assessments are typically obtained via self-report inventories and they are therefore susceptible to response bias as well as response distortion. It is therefore important to explore the construct of Social Desirability (SD) within the South African Personality Inventory (SAPI) project with specific focus on the factor structure of the Social Desirability scale. This study developed two substantive research hypotheses on the factor structure underlying the Social Desirability scale of the SAPI. The SD scale of the SAPI comprises six positively keyed items and six negatively keyed items. It was hypothesised that the positively keyed items would share variance, simply because they share a design feature essentially unrelated to the construct of interest. The same argument was made with regards to the negatively keyed items. Two method factors (a positively keyed factor and a negatively keyed factor) in addition to the two substantive social desirability factors of interest were therefore hypothesised. This allowed the development of four operational research hypothesis on the nature of the measurement model underlying the SAPI SD scale. Confirmatory factor analysis was used to test the four operational hypotheses. Operational hypothesis 1 was discredited by the results. Operational hypotheses 2 to 4 were all to some degree supported but not unambiguously supported. In the case of all three operational hypotheses several statistical null hypotheses formulated with regards to Lambda and Phi could not be rejected. The implications of these results are discussed in the final chapter. There were a few limitations and recommendations for further research that were posed in this chapter. The current study suggests that it is probably most prudent to abandon further attempts to describe the psychological mechanism that produced the observed inter-item covariance matrix obtained for the current SAPI SD scale. The current study suggests that it probably is more prudent to rather focus on the clear conceptualisation of the SD construct, the development of a new SD scale for the SAPI and the psychometric evaluation of the reliability, construct validity and measurement bias of the new SD scale.

AFRIKAANSE OPSOMMING: Persoonlikheidsmeting word vry algemeen in Suid Africa gebruik om individue se toekomstige werksprestasie te voorspel (Barrick & Mount, 1996; Hough, Eaton, Dunette, Kamp, & McCloy, 1990; Ones & Viswesvaran, 2001). Dit word tipies aanvaar dat die wyse waarop ‘n individu op ‘n persoonlikheidsvraelys reageer ‘n akkurate en ware weerspieëling is van hul persoonlikheid wat vergelyk kan word met die persoonlikheidsprofiele van ander individue (Van Herk, Poortinga, & Verhallen, 2004; Ziegler, Maccann, & Roberts, 2012). Persoonlikheidsmetings word tipies verkry deur middel van selfrapporteervraelyste en sodanige metings is dus vatbaar vir responssydigheid en responsverwringing. Dit is gevolglik belangrik om die konstruk Sosiale Wenslikheid (SW) in die konteks van die Suid-Afrikaanse Persoonlikheidsvraelys (SAPI) projek te ondersoek met spesifieke verwysing na die faktorstruktuur van die Sosiale Wenslikheid skaal. Twee substantiewe navorsingshipoteses oor die faktorstruktuur onderliggend aan die Sosiale wenslikheidskaal van die SAPI is ontwikkel. Die SW skaal van die SAPI bestaan uit ses positiewe gesleutelde items en ses negatief gesleutelde items. Die hipotese is geformuleer dat die postief gesleutelde items variansie sou beel bloot omdat hul ‘n ontwerpkenmerk deel wat wesenlik onafhanklik is van die konstruk van belang. Dieselfde argument is aangevoer ten opsigte van die negatief gesleutelde items. Naas die twee substantiewe sosiale wenslikheidsfaktore is twee metode-faktore (‘n positief gesleutelde faktor en ‘n negatief gesleutelde faktor) gevolglik veronderstel. Dit het die moontlikheid geskep om vier operasionele hipoteses te ontwikkel oor die aard van die metingsmodel onderliggend aan die SAPI SW skaal. Bevestigende faktorontleding is gebruik om die vier operasionele hipoteses te toets. Operasionele hipotese 1 is deur die resultate gediskrediteer. Operasionele hipoteses 2 tot 4 het almal tot ‘n mate steun ontvang maar geen hipotese is ondubbelsinnig gesteun nie. In die geval van al drie hierdie operasionele hipoteses kon verskeie statistiese nulhipoteses gestel met betrekking tot Lambda en Phi nie verwerp word nie. Die implikasies van hierdie bevindinge word in die finale hoofstuk bespreek. ‘n aantal leemtes en aanbevelings vir verdere navorsing word in hierdie hoofstuk bespreek. Die huidige studie stel voor dat die mees raadsame weg om te volg waarskynlik sou wees om verdere pogings te staak om die sielkundige meganisme te beskryf wat aanleiding gegee het tot die waargenome inter-itemkovariansiematrys wat vir die huidige SAPI SW skaal verkry is. Die voorstel van die huidige studie is dat dit waarskynlik meer raadsaam sou wees om te fokus op die duidelike konseptualisering van die SW konstruk, die ontwikkeling van ‘n nuwe SW skaal vir die SAPI en die psigometriese evaluasie van die betroubaarheid, konstrukgeldigheid en metingsydigheid van die nuwe SW skaal. Sleutelwoorde: sosiale wenslikheid, persoonlikheidsassessering, konstrukgeldigheid, SAPI

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/100135
This item appears in the following collections: