*3.1. Data Collection*

A quantitative research approach was adopted to test the developed hypotheses. For the data collection, an online questionnaire-based survey was chosen, primarily because it could reach many respondents (Abdel-Kader and Luther 2008), in addition to enabling the respondent to answer how and where he/she wants, without feeling the pressure to respond immediately (Gillham 2008). The questionnaire also allowed us to collect the data needed to test the developed hypotheses in the previous section (Gillham 2008). Furthermore, with a questionnaire survey, it is possible to collect sufficient data to allow the generalization of results to the population being analyzed, which allows advances in contingency research that comes from the ascertainment of general patterns (Chenhall 2003).

As this instrument is not without limitations, to minimize them, the recommendations of Hill and Hill (2008) were followed. The questions were organized into sections to create a common thread between them and a pre-test was conducted. We sent to experts (i.e., three academics and five practitioners) a draft of the questionnaire for review and recommendation. The purposes of the pre-test were threefold: (i) to assess the adequacy of our questionnaire design, (ii) to verify whether the items captured the relevant dimensions of our variables, and (iii) to verify whether practitioners found the questionnaire items understandable and plausible. The feedback led to minor changes in the structure of the questionnaire and the wording of some individual items.

To implement the survey, we chose to send it by electronic mail, as in previous studies (e.g., Cadez and Guilding 2008; Cescon et al. 2019; Holm et al. 2016; Hyvönen 2007; Latan et al. 2018; Lau and Sholihin 2005; Massicotte and Henri 2021; Monteiro et al. 2022). A free software program available on the internet was used for this purpose. It does not require programming skills and is inexpensive to use (Fleming and Bowden 2009; Ganassali 2008). Compared with conventional questionnaire-based surveys (i.e., mail, telephone, and on-site), online questionnaire-based surveys have two main advantages: (i) low-cost administration, enabling large sample sizes, and (ii) speed and accuracy of data collection (Fleming and Bowden 2009). Moreover, Loomis and Paterson (2018) stress that there are no significant differences between the two data collection modes regarding response rate, item nonresponse, and nature of the data, which validates this approach.

#### *3.2. Sample and Procedure*

After finalizing the questionnaire, the target population of the study was selected. For this, we asked the National Statistics Institute (NSI) a list of the top 500 manufacturing companies, according to turnover, operating in Portugal. The choice of the larger manufacturing companies is justified by the fact that several studies state that larger companies have greater information needs for planning, control, and decision-making purposes (Abdel-Kader and Luther 2008; Alves 2010; Chenhall 2003; Chenhall and Langfield-Smith 1998a; Haldma and Lääts 2002). In addition, most of the studies reviewed had as targets the larger manufacturing companies (e.g., Baines and Langfield-Smith 2003; Cescon et al. 2019; Hoque and James 2000; Oyewo 2021; Visedsun and Terdpaopong 2021). Therefore, to be able to compare our results with those of these studies, it was considered important that the target population had similar characteristics.

Since we needed email addresses to send the questionnaire and collect the data, we updated the information provided by the NSI through a telephone call made to all companies. This led to the exclusion of nine companies1. After that, the questionnaire was sent to the person in charge of the financial department of 491 large manufacturing companies operating in Portugal that are the subjects of this study. Three months after the first mailing, a second mailing of questionnaires was carried out. Aware that the number of contacts, the persistence of the researcher, and the personalization of those contacts significantly affect the response rate to surveys (Ganassali 2008), along with the third mailing of questionnaires, we started making phone calls inviting financial managers to participate in the study. At the end of this process, 119 questionnaires were received. As there was a need to exclude five questionnaires because they were incorrectly completed (Gillham 2008), 114 usable responses were considered, which corresponds to a response rate of 23%. This response rate is above or in line with those of previous studies on management accounting (Baines and Langfield-Smith 2003; Cescon et al. 2019; Holm et al. 2016; Latan et al. 2018; Monteiro et al. 2022; Visedsun and Terdpaopong 2021).

Regarding the characteristics of the firms surveyed (turnover and number of employees) it was verified that 49% of the firms had turnover exceeding EUR 55 M and 57% had more than 250 employees (Table 1). Regarding the respondents, it appears that 50% were in an administrative/financial position.


**Table 1.** Characteristics of sample and respondents.

To examine possible differences between respondents and non-responding companies concerning turnover and number of employees, like Drury and Tayles (2006) and Guilding et al. (2000) we used the nonparametric test of Mann-Whitney. Regarding the turnover, some differences were found in central tendency. Through the analysis of the median, it was found that firms with higher turnover were also the ones that answered the questionnaire.

#### *3.3. Variables Measurement*

Some instruments developed in previous studies were used or adapted to measure the variables needed to test the research hypotheses developed in this study. Besides ensuring a greater reliability and validity of the information gathered, using instruments tested in other studies allows a more accurate comparison of results (Chenhall 2003). In this context, we will describe how the variables were operationalized and the instruments used to measure them.

The environmental uncertainty variable was measured based on an instrument developed by Teo and King (1997) and used later by Newkirk and Lederer (2006). Some adjustments were made in this instrument resulting from the studies of Chenhall (2003) and Löfsten and Lindelöf (2005). We added two items presented by Löfsten and Lindelöf (2005) (the intensity of research and development; and the legal, political, and economic constraints), and an item suggested by Chenhall (2003) (the requirements in terms of social and environmental responsibility) that recognizes social and environmental issues as relevant sources of environmental uncertainty (Latan et al. 2018). Respondents were requested to indicate their perception about the predictability of 15 items using a Likert-type scale ranging from 1 (strongly disagree) to 5 (strongly agree). Rating on these items was averaged to determine the environmental uncertainty index. Prior studies have used a similar approach to measure environmental uncertainty (e.g., Hoque 2005; Oyewo 2022).

In order to measure accounting information relevance (financial and non-financial), we used an instrument based on a set of items identified in the literature (Baines and Langfield-Smith 2003; Chow and Van der Stede 2006; Vaivio 1999). Comprising 25 items, this instrument considers nine items to measure financial information relevance (indicators of costs, results, profitability, and return on investment) and 16 items to measure nonfinancial information relevance (indicators related to processes and operations, employees, and suppliers and customers). Respondents were asked to indicate the relevance attributed to these items of financial and non-financial information to decision-making purposes using a Likert-type scale ranging from 1 (no impact) to 5 (very relevant). The respective scores were additively combined and averaged to derive indexes for financial information relevance and non-financial information relevance. A similar approach was used by Baines and Langfield-Smith (2003).

To measure the organizational performance variable, an instrument previously used by Hoque and James (2000) and modified by Cadez and Guilding (2008) was adopted. This instrument, which was used in several studies (e.g., Abernethy and Lillis 1995; Baines and Langfield-Smith 2003; Chenhall and Langfield-Smith 1998b; Oyewo 2022), consists in asking respondents to assess the organization's performance compared to that of competitors. This instrument considers seven2 financial and non-financial dimensions and uses a Likert-type scale ranging from 1 (much worse than competitors) to 5 (much better than competitors). Rating on these dimensions was averaged to determine the organizational performance index for each company. Prior studies have used a similar approach to measure organizational performance (e.g., Cadez and Guilding 2008; Oyewo 2022).

Statistical analysis of the data collected was performed using the SPSS program, as in previous studies (e.g., Hoque 2005; Monteiro et al. 2021; Monteiro et al. 2022). We carried out univariate, bivariate, and multivariate analyses. In the next section, we proceed to the presentation and discussion of results.

#### **4. Results and Discussion**

The goals of this section are to describe the variables and analyze the association between environmental uncertainty, financial and non-financial information relevance in decision-making, and organizational performance. In addition, we analyze how the interaction between environmental uncertainty and non-financial information relevance for decision-making purposes influences organizational performance. That is, in this section we present a descriptive analysis of the variables and test the hypotheses developed in Section 2.

#### *4.1. Descriptive Analysis*

The reliability and descriptive statistics of the variables under study are presented in Table 2. As shown, and except for environmental uncertainty, the variables were established with all the items considered in the questionnaire. For the creation of the environmental uncertainty variable, 2 of the 15 items were excluded (the shortage of skilled labor and the shortage of materials) since Cronbach's alpha improves with this exclusion. We used Cronbach's alpha to assess the reliability, or internal consistency, and it appears that all the items had a good or very good internal consistency (Marôco 2021; Pestana and Gageiro 2014).

**Table 2.** Reliability and descriptive statistics of variables.


Regarding environmental uncertainty, the perception of respondents indicated that the level of uncertainty was moderate (3.31) and resulted essentially from the intense competition on prices, quality, and product differentiation, and the high level of social and environmental responsibility. Thus, it seems that factors related to intense competition drive environmental uncertainty (Baines and Langfield-Smith 2003; Mia and Clarke 1999). The same was true regarding the requirements in terms of social and environmental responsibility, as advocated by Chenhall (2003) and Latan et al. (2018), which were perceived by respondents as very high. In fact, most Portuguese companies operate in the highly competitive environment of the European Union (Monteiro et al. 2021). In addition, social and environmental issues have been also one of the biggest priorities of the European Union.

Concerning the relevance attributed to financial and non-financial information for decision-making purposes, the results indicate that it was significant (high or very high). The financial indicators considered most relevant were the overall results of the organization, production costs, and sales costs. In turn, the most relevant non-financial indicators were related to customer satisfaction, loyalty, and complaints, all of which are external indicators. Curiously, non-financial indicators related to employees were deemed the least relevant for decision-making.

Respondents' perceptions of organizational performance when compared to competitors' performance are also shown in Table 2. The results obtained reveal that respondents considered their organization's performance to be moderately higher than that of competitors.

#### *4.2. Hypotheses Testing*

In this section we demonstrate how the hypotheses formulated in the second section of the paper were tested, using bivariate and multivariate analyses. An objective of this study was to examine the nature and strength of the association between key variables. This was achieved using correlation analysis, in particular Spearman correlations. Here, when the coefficients are closer to 1 it means a strong positive association between two variables; in contrast, coefficients closer to zero imply weak association. A summary of the results obtained is presented in Table 3.



\*\*, \* Significant at 1 and 5% (two-tailed), respectively.

Our analysis begins by examining the strength of the association between variables using the Spearman correlation coefficient (Table 3). It appears that there was a positive association, although weak, between environmental uncertainty and non-financial information relevance, related to employees and the external environment. These results are statistically significant (*p*-value < 0.05), suggesting that with increasing environmental uncertainty greater emphasis should be placed on non-financial information in decision-making. It should be noted, however, that this association between environmental uncertainty and non-financial information relevance was not evident when it comes to information related to processes and operations. Still, the results allow supporting Hypothesis 1. These results are consistent with those of other studies, conducted in other countries (Boulianne 2007; Chenhall and Morris 1986; Chong and Chong 1997; Hoque 2005; Hoque and James 2000; Lal and Hassel 1998). Therefore, we can state that the relevance attributed to non-financial information for decision-making purposes is greater when environmental uncertainty is higher. In these contexts, managers need sophisticated accounting information to enhance decision-making quality (Latan et al. 2018) and non-financial information, in particular, contributes to decision-making success (Monteiro et al. 2022).

No statistically significant association was evidenced between environmental uncertainty and financial information relevance in decision-making. Regarding the relevance attributed to accounting information, the results (Table 3) indicate a strong and positive association between the relevance attributed to financial information and the relevance attributed to non-financial information in decision-making. Thus, it appears that when greater relevance was attributed to financial information in decision-making it was also attributed to non-financial information, whether related to processes and operations and employees (*p*-value < 0.01), non-financial information connected with customers and suppliers (*p*-value < 0.05), or vice versa. As argued by Chenhall and Langfield-Smith (2007), non-financial information complements financial information but does not replace it, confirming Hypothesis 3. These findings are in line with the results obtained by Lau and Sholihin (2005) and Chow and Van der Stede (2006). Financial information is not sufficient for decision-making purposes (Monteiro et al. 2021). In several situations, both financial and non-financial information is used (Bhimani and Langfield-Smith 2007; Massicotte and Henri 2021). Therefore, we can state that financial information and non-financial information are complementary.

To compare the relevance given to financial information and non-financial information, both overall and partial (processes and operations, employees and outside), we used the nonparametric Wilcoxon test, which allows comparing measures of central tendency of two variables (Marôco 2021; Pestana and Gageiro 2014). Previous studies have used the Wilcoxon test for similar analyses (e.g., Holm et al. 2016). The results obtained and summarized in Table 4 show that there are statistically significant differences (*p*-value < 0.01) between the relevance attributed to financial information and non-financial information in decision-making. Through the analysis of the measure of central tendency of these

variables it was found that the median of the relevance attributed to financial information (4.17) is above the median of relevance attributed to non-financial information (4.06). However, despite this difference between the relevance attributed to financial information and non-financial information related to processes and operations and employees, no statistically significant differences between the relevance attributed to financial information and non-financial information related to the external environment could be demonstrated. In this context, as verified by Bhimani and Langfield-Smith (2007), there are situations in which financial information is considered more relevant to decision-making than nonfinancial information. In this study, these situations fell into the contexts of decision-making associated with processes and operations, and employees. These results confirm Hypothesis 2 and are consistent with those obtained by other authors (Chow and Van der Stede 2006; Hyvönen 2007). Financial information remains relevant for managers (Hyvönen 2007). Therefore, we can state that managers assign greater relevance to financial information for decision-making than to non-financial information in decisions related to internal processes, operations, and employees.

**Table 4.** Wilcoxon tests: relevance of financial and non-financial information.


To test Hypothesis 4, and thus assess the adjustment between the environmental uncertainty and the relevance assigned to non-financial information, as in previous studies (e.g., Hoque 2005; Hyvönen 2007; Oyewo 2022), we used the linear regression model that allows estimating the direct effects and interaction of independent variables on the dependent variable. Following Afifa and Saleh (2021) and Oyewo (2022), before performing the regression analysis, multicollinearity between the predictor variables considered in each regression model was inspected using correlation analysis. Table 3 shows that all Spearman correlations were at the acceptance level (low correlation levels) and, thus, there was no problem with multicollinearity. Consequently, a regression model was run using the SPSS program. According to Gerdin and Greve (2008, p. 1003) "one frequently used technique for testing the existence of a significant difference in regression coefficients is the moderate regression analysis (MRA)".

MRA has usually the following format:

$$\mathcal{Y} = \mathfrak{a}\_0 + \beta\_1 \mathcal{X}\_1 + \beta\_2 \mathcal{X}\_2 + \beta\_3 \mathcal{X}\_1 \mathcal{X}\_2 + \varepsilon\_1$$

where *Y* represents organizational performance (dependent variable); *X*<sup>1</sup> represents environmental uncertainty (independent variable); *X*<sup>2</sup> represents the non-financial information relevance (moderator); *X*1*X*<sup>2</sup> is the moderating effect that *X*<sup>2</sup> has on the relationship between *X*<sup>1</sup> and *Y*; *α*<sup>0</sup> represents a constant; and *e* is the error variable (Gerdin and Greve 2008).

The results (Table 5) show that there was no effect of environmental uncertainty in the relevance attributed to non-financial information and in the interaction between these variables on the dependent variable related to organizational performance. This is because the linear regression model and the effects (direct and interactive) of independent variables on organizational performance were not statistically significant.


**Table 5.** Results of regression.

In order to verify a possible partial effect of non-financial information and environmental uncertainty on the organizational performance, we used the linear regression for each of the components of non-financial information (processes and operations, employees, and external environment). The results presented in Table 6 confirm that there was no direct effect of the relevance attributed to non-financial information, nor any effect resulting from the interaction with environmental uncertainty on the organizational performance. The models presented and the effects of independent variables on organizational performance are not statistically significant.

**Table 6.** Additional regression analysis (partial effects).


As there was no direct effect from environmental uncertainty and the relevance assigned to non-financial information, not even one resulting from the interaction between them on organizational performance, Hypothesis 4 is not confirmed. In this sense, the results are surprising and contradict the results obtained by Hoque and James (2000), Hoque (2005), and Al-Mawali and Am (2016), who concluded that organizational performance is influenced by the interplay between environmental uncertainty and the relevance assigned to non-financial information for decision-making or the use of customer accounting information. These differences can occur because other aspects are not being considered, which may also influence environmental uncertainty, non-financial information relevance, and organizational performance. Therefore, additional research should be conducted to investigate this association.
