Measurement effects between modes in mixed-mode survey Consumer confidence

Učinek merjenja med načini anketiranja (angl. measurement effect between modes) v kombiniranih oziroma MM anketah (angl. mixed-mode surveys) otežuje posploševanje anketnih ocen na celotno populacijo in primerljivost anketnih ocen med posameznimi sociodemografskimi skupinami znotraj iste MM ankete, primerljivost med časovno ločenimi izvedbami periodičnih MM anket in primerljivost s podobnimi anketami v drugih državah. Kvantificiranje učinka merjenja med načini anketiranja v MM anketi je ključni korak k ovrednotenju resnosti problema v posamezni MM anketi, oblikovanju strategij za zmanjševanje učinka merjenja med načini in upoštevanju učinka v končnih anketnih ocenah. Cilj pričujočega magistrskega dela je oceniti učinek merjenja med načinoma anketiranja v MM anketi Mnenje potrošnikov, ki jo izvaja Statistični urad Republike Slovenije (SURS) in kjer se za zbiranje podatkov uporablja kombinacija spletnega in telefonskega načina anketiranja. V tem okviru smo zastavili dve hipotezi. Prva se nanaša na sam obstoj učinka merjenja med načinoma anketiranja, druga pa na njegovo smer. Glede na razpoložljive podatke smo uporabili dve metodi, metodo pomožnih spremenljivk (angl. backdoor method) in metodo primerjalnega vzorca (angl. comparative sample method or reference survey approach), ki smo ju implementirali na štirih ciljnih spremenljivkah. Izkazalo se je, da lahko hipotezi potrdimo z obema metodama. Ob tem smo med obema metodama ugotovili tudi določene razlike, ki so pričakovane, saj izhajajo neposredno iz njunih specifičnosti. V zvezi s tem metoda pomožnih spremenljivk sicer kaže nekoliko šibkejše učinke merjenja med načinoma, vendar je zaradi svoje praktičnosti in nizkih stroškov potencialno primernejša, še posebej ob morebitnih izboljšavah.The measurement effect between modes in MM surveys (mixed-mode surveys) may impede the applicability of survey estimates and the comparability of survey estimates between socio-demographic segments within a MM survey, the comparability between different editions of periodic MM surveys and the comparability with similar surveys in other countries. Estimating the extent of the measurement effect between modes in MM surveys is a key step for evaluating the severity of the problem in individual MM survey, designing strategies to minimize the measurement effect between modes and for factoring them in the final survey estimates. The aim of this master’s thesis is to estimate the measurement effects between modes in the MM consumer confidence survey conducted by the Slovenian Statistics Office, in which a combination of web and telephone modes is used to collect data. In this setting, two hypotheses were examined. The first one relates to the presence of a measurement effect between modes and the second one to its direction. Given the nature of data available, we applied the backdoor method and the comparative sample method. Results show that both hypotheses can be confirmed by both methods. As anticipated, we also found differences between the two methods owing directly to their specific characteristics. In this context, the backdoor method shows weaker measurement effect between modes but is potentially more advantageous due to its practicality and low costs, especially if improved