**4. Analysis and Discussion**

First, the dataset was checked for nonsensical answers. For this purpose, two additional statements, 27 and 28, were included to the category "Perceived level of logistics services": "Logistics company/department does not use IT to provide order information" and "Operations of logistics company/department are not environmentally safe". These variables were reverse-coded to be compared with statements DIG7 and SUS22 from Table 3. A closer review of the two pairs of reverse-coded factors revealed no nonsensical answers.

In order to detect the existence of non-response bias, two techniques were used: extrapolation [54] and a comparison of respondents' characteristics known a priori with those of the population [55]. In order to determine the probable direction of bias, the last five returns were compared with the first five returns, assuming that late respondents are most similar to non-respondents because their replies took longest. The answers of the earliest five returns did not differ substantially from those of the latest five returns. Furthermore, respondents' characteristics such as shares of specialization types or shares of the organizational and legal form of ownership did not differ considerably from those of the population. Thus, it is assumed that non-response bias is unlikely to be an issue in the study.

In order to examine which measures can describe and measure LSQ in the developing economy of Ukraine, the factor structure was viewed using Varimax rotation with a Kaiser normalization approach. The initial correlation matrix was singular following the Bartlett's test of sphericity and Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy because mean values of several variables were in the high 4 to 5 range and almost perfectly correlated with each other. Having extracted these variables from the model, the KMO measure (0.6496) was greater than the minimum value of 0.60 normally suggested by Hair et al. [56].

The last run of EFA on the 22 measurement variables identified five factors with eigenvalues above 1. As Table 4 shows, these five factors explained 96.15 % of the variance. Three measurement variables were excluded because all respondents stated the same expectation level (5.0) for logistics service regarding these criteria (DSQ10, DSQ12, and DSQ13). For the first run of the factor analysis, one variable (TIM24—pick-up and delivery on time) does not load highly (>0.3) on any of the identified factors. This variable was excluded from the existing measurement scale. Table 4 shows a rotated orthogonal varimax component matrix which demonstrates how each variable item is loaded on each of the factors.


**Table 4.** General perception of LSQ by agricultural enterprises.

Source: Author's illustration. Note: blanks represent abs loading <3.

Composite reliability (CR) and average variance extracted (AVE) were assessed to check the reliability of the model measurement. The internal reliability of all the observed variables in their measurement of each latent construct was assessed by CR, demonstrating that the observed variables have adequate internal consistency. A CR value of 0.6 or more was recommended by Fornell and Larcker [57] (p. 45). It could be concluded that all factors were based on reliably observed items with CR values in the range of 0.84–0.99, as can be seen in Table 4. Thus, the observed variables are adequate for representing the respective factors. AVE measures the amount of variance in the measured variables. It should be greater than 0.5 [57]. As depicted in Table 4, the AVE was only lower than 0.5 in Factor 3 (0.44). Nevertheless, it was accepted, because its composite reliability is greater than 0.6 and the convergent validity of the construct can be considered as adequate [57] (p. 46).

Factor 1 consists of six variables, namely, REL1 (staff's attitude and behavior to satisfy customers' needs), REL2 (responsiveness to customers' needs and requirements), REL4 (flawless service), DSQ11 (reliability of service (delivery at the proper time)), TIM25 (reliability of service (within a proper transportation time)), and TIM26 (reliability of service (at the promised time)). These last three variables belong to different factors in the predefined model. Basically, these six measures in combination depict the reliability of LSQ. Factor 1 can be renamed "Reliability".

Factor 2 consists of five variables: DIG5 (application of IT and EDI in customer service), DIG6 (application of innovative IT in customer service), DIG7 (availability of order information using IT), DIG8 (shipment tracing using IT), and DSQ9 (availability and condition of equipment and facilities). The first four variables were initially assigned to "Digital transformation". Although the last variable is from "Physical distribution service quality", it also addresses the use of technologies in logistics operations relating to equipment and facilities. Thus, this factor keeps the original heading "Digital transformation".

There are seven variables loaded highly on Factor 3, namely, COR14 (company's reputation for reliability in the market), COR15 (record of professionalism and consistency in satisfying customers), COR17 (concerned about its ethical image), SUS18 (record of engagement in community activities), SUS19 (performance statement and vision for community responsibility), SUS20 (socially responsible behavior and concerns for human safety), and SUS23 (company offers employees training). Although half the variables belong to different factors in the initial model, they all indicate in one way or another the extent of a company's image as a reliable and professional partner, whether it is ethically responsible, and whether it is a caring employer and "a good company" in its community. The new dimension "Corporate social responsibility" should house these seven variables.

As Table 4 illustrates, Factor 4 includes two variables: SUS21 (logistics operations with minimal environmental pollution) and SUS22 (environmentally safe operations). Both variables were assigned to the dimension "Sustainability" in the initial conceptual model. These variables are the only two focusing on environmental sustainability and should be grouped to a new dimension "Environmental sustainability".

Finally, Factor 5 also encompasses two variables: REL2 (responsiveness to customers' needs and requirements) and COR16 (company's reputation for matching words with actions). In the initial conceptual model, the first variable was related to "Quality/reliability of customer focus", while the second variable belongs to "Corporate image". Their common ground is that they address the logistics company's efforts to understand customers' needs and requirements and satisfy them in a trustable and reliable way. Factor 5 is therefore named "Quality of customer focus".

Thus, the EFA results show a different number of dimensions (five instead of six) and a different allocation of explanatory attributes for LSQ in the developing economy of Ukraine. H1 is not supported. Nevertheless, the EFA shows that LSQ is a construct of 22 explanatory attributes which were partly re-assigned to the following five dimensions: reliability, digital transformation, corporate social responsibility, environmental sustainability, and quality of customer focus.

Finally, in order to test whether there is a substantial difference between the expected and perceived quality of logistics service, the mean score of respondents' answers about the expected level of sustainability-related LSQ attributes was compared with the mean score of respondents' answers about their perceived level of sustainability-related LSQ attributes. For this purpose, seven explanatory variables for the explored dimension "Corporate social responsibility" and two explanatory variables for the explored dimension "Environmental sustainability" were considered from the EFA results gained in the previous step. Mean score differences and their rank in the total list of 22 attributes are shown in Table 5. The lower the difference in mean scores, the higher the respondents' satisfaction. In other words, any positive difference in the mean scores indicates that the expectation is higher than the perceived level of the particular factor for LSQ on the one hand, and that a respective improvement will better match respondents' expectation level on the other hand.


**Table 5.** Difference between expected and perceived level of sustainability-related LSQ factors.

Source: Author's illustration.

The mean score differences of the attributes for corporate social responsibility show a substantial difference between expected and perceived quality. The average expectation level of the respective seven attributes in Table 5 was stated to be either absolutely essential or very important (average mean score from 4.885 to 3.712), but the perceived quality of these social sustainability attributes of LSQ was substantially lower. The social sustainability attributes of LSQ seem to be upgradable and important for satisfaction with LSQ. The top ranked attributes SUS20, SUS19, and SUS17 offer the greatest potential from all 22 attributes to increase satisfaction among agricultural enterprises in Ukraine. On this basis, it may be concluded that the findings provide support for hypothesis H2.

Environmentally safe operations (SUS22) and logistics operations with minimal environmental pollution (SUS21) were stated as absolutely essential (average mean score of 4.519 and 4.462, respectively) and the perceived quality was approximately the same or slightly higher than expected (average mean score of 4.468). Moreover, these two environmentally related attributes of LSQ are at the bottom of the ranked list, offering the smallest potential to further increase the extent of perceived quality. Thus, there is no substantial difference between the expected and perceived environmental-related attributes of logistics service quality. H3 is not supported.
