Construction and validation of a corruption perception scale at the citizen level

The study described the creation and validation of the Corruption Perception Scale (CPS), which assesses how citizens perceive corruption. In a qualitative step, the instrument was evaluated by experts, followed by a pre-test. In the quantitative step, exploratory and confirmatory factor analysis was performed, totaling a sample of 1,075 cases. Finally, a methodology for the application of CPS was suggested. The final structure of the measure was composed of five dimensions at the individual level (knowledge, behavior, reflexes, control, and attitude), which position the citizen as the protagonist in the analysis of the phenomenon.


INTRODUCTION
Corruption is a pervasive problem, faced by several countries at different times, and which, although its extent may vary from one society to another, it threatens all nations (Mousavi & Pourkiani, 2013). It is defined as the abuse of power entrusted to personal gain (Brown, 2006;Transparency International, 2019), corruption undermines a society's justice, economic stability, and efficiency (Shacklock, Sampford, & Connors, 2006), in addition to placing their democratic and moral values at risk (Lambsdorff, 1998). Among the corrupt practices, the most common refer to the payment of bribes, money laundering, influence peddling (Controladoria Geral da União [CGU], 2009), favouritism, nepotism, illegal political sponsorship, extortion, theft and fraud (Cavalcante, 2018).
The most widely used and globally known indicator of the level of corruption in the public sector is the Corruption Perceptions Index (CPI), published annually since 1995 by the non-governmental organization Transparency International (TI) (Gorsira, Denkers, & Huisman, 2018;Transparency International, 2021;Villarino, 2021). According to Transparency International (2021), the CPI assesses 180 countries and territories and assigns them scores on a scale between 0 and 100, ranging from very corrupt to very transparent, respectively. Thus, in 2020, the best rated countries were Denmark (88 points), New Zealand (88), Finland (88), Singapore and Sweden (85); while Venezuela (15 points), Yemen (15), Syria (14), Somalia (12) and South Sudan (12) stand out negatively in the global context.
Another important index is the World Governance Indicators (WGI), a project of the World Bank Group, which has produced governance indicators for more than 200 countries and territories since 1996, considering six dimensions, including the 'Corruption Control' (CoC). In the ranking that varies from 0 to 100, comparing all countries in the world, in 2019, countries on the African continent such as South Sudan, Equatorial Guinea and Somalia, were in the three worst rankings in the indicator of Corruption Control, with positions less than 1.0. In turn, Finland, New Zealand and Singapore have scores close to 100, indicating adherence to the fight against corruption, based on this indicator. In the view of Villarino (2021), the CoC stands out in the comparison of nations, considering that it provides information on changes over time, for a relevant number of countries, based on methodological refinement and support from an institution renowned as the World Bank.
The use of indices and indicators to measure corruption has helped governments to make policy choices, presenting a scenario of popularity (Perumal, 2021) and, as far as possible, embarrassing corrupt governments (Mungiu-Pippidi & Dadašov, 2016). In the field of research, Malito (2014) emphasizes that there is a high application of these indices in academic production that deals with the impact of corruption in developed and developing countries. However, there are recommendations that they be applied with caution and carefully analysing the purpose of the investigation, considering that such metrics are largely based on the perceptions of experts, lacking both specificity and transparency (Mungiu-Pippidi & Dadašov, 2016).
In addition to this criticism, both metrics (CPI and CoC) are considered to be structured on the aegis of evaluating the administrative structure of the State, making no effort to understand how the common citizen perceives corruption in the face of the influences it is exposed to. This is the proposal of this investigation, which aims to build and validate a scale of perception of corruption in the view of the common citizen and not agents and specialists in government issues. Ko and Samajdar (2010) encourage the need to fill this gap, exploring the sources of perception from a theoretical point of view, from a bottom-up perspective, taking into account the perception of citizens about corruption. The validation of scales of this nature has merit in enabling the comparison of phenomena -in this case the perception of corruption by the populationbetween countries (Overman, Schillemans, & Grimmelikhuijsen, 2020) and regions of the same country, such as Brazil, with continental dimensions.
Studying citizens' perception of corruption is essential, considering that it can impact the population well-being and government actions. Školník (2020) highlights that the (negative) perception of corruption on the part of a citizen leads to the absence of all forms of social participation, such as, for example, electoral participation, acting in councils and municipal meetings and political parties and demonstrations. Neshkova and Kalesnikaite (2019) corroborate, noting that if citizens assess a government as corrupt and dishonest, they become sceptical about political life and, consequently, are less likely to participate in democratic governance.
Bearing in mind the possibility of citizens losing the motivation to participate politically in an environment they consider corrupt, the relevance of this investigation stands out, which enables the identification of ways in which citizens understand about corruption in the country where they live, from different perspectives. Yu, Chen, and Lin (2013) point out that controlling corruption first requires a means of measuring the phenomenon, as only then problems can be correctly diagnosed, and solutions properly evaluated. Thus, according to the authors, for a democratic government to govern effectively, research that assesses the perception of corruption 'by those in the streets' must be taken into account (Yu et al.,p. 57).

DEVELOPMENT OF THE CORRUPTION PERCEPTION SCALE (CPS)
The proposal to build the scale is highlighted, as, in addition to its theoretical innovative character, it reflects an effort to understand how the common citizen recognizes this complex phenomenon (Gorsira et al., 2018). More than 30 years ago, Hilgartner and Bosk (1988) already pointed out that, from a philosophical point of view, the use of subjective perception to measure corruption is justifiable, because public issues are a projection of the collective cognition of society as a whole and not simply a reflection of objective reality. Collins, Uhlenbruck, and Rodriguez (2009) indicate that studies with a subjective perspective are complex to be conducted at the individual level, considering that corruption is difficult to define, observe and measure. To seek to eliminate these barriers, the instrument developed for this research is comprehensive in the analysis of corruption, elaborated from an extensive literature review, incorporating five dimensions of individual level (knowledge, behaviour, reflex, control and attitude) and positioning the citizen as protagonist of the analysis of the phenomenon.
The knowledge dimension reflects 'what the citizen knows about corruption'. Politically aware citizens understand political information differently from those who do not give equal relevance to the topic (Weitz-Shapiro & Winters, 2016). Thus, it is essential that citizens are aware of the meaning of corruption (Lin & Yu, 2014), corrupt practices (Sadek, 2019), combat legislation (Abreu & Gomes, 2021) and that they seek information to update themselves on the theme (Bai, Liu, & Kou, 2014;Yu, Chen, & Lin, 2013), as well as taking it to the heart of the social discussion (Weitz-Shapiro & Winters, 2016). Box 1 lists the CPS' knowledge dimension items, defined based on current literature.

Code Item
Item 1 I know what corruption means.
Item 2 I am interested in looking for information about corruption.
Item 3 In my school/academic background, corruption was/is a debated topic.
Item 4 I discuss corruption in my family/social relationships.
Item 5 I can identify corrupt practices.
Item 6 I know different forms of corruption (e.g., kickback, overpricing and/or other forms).
Item 7 I know the legislation on corruption.
Item 8 I know what are the appropriate punishments for the citizen who commits corrupt acts.
Item 9 I easily notice when corrupt practices occur.
Source: Elaborated by the authors.
The knowledge dimension is also justified, given that the understanding of corruption is reflected in a cultural phenomenon, however, individuals should not be pre-judged in relation to their country of origin (Barr & Serra, 2010). In a practical way, this dimension analyses the relationship between the different forms of obtaining knowledge about corruption and the relevance that such knowledge has under this perception. Thus, we imply that a citizen who is unaware of corrupt practices and acts may have his perception of this phenomenon impaired.
The behaviour dimension represents 'how the citizen behaves towards corrupt acts'. Marquette and Peiffer (2018), when comparing the theory of collective action and the principal-agent theory, conclude that both have very close indications regarding the decision to get involved in corruption, which may be motivated by the citizen's conception that they will not lose their status of beneficiary of something and will not be held responsible for such an act. The relevance of knowing the perception of ordinary citizens regarding their actions towards corrupt acts is reinforced by behavioural analysis. Based on current literature, Box 2 lists the CPS items according to behaviour dimension.

Code Item
Item 10 I have witnessed acts of corruption.
Item 11 I have lived with people who have been accused of the crime of corruption.
Item 12 I have already been invited to practice small acts of corruption to get some benefit in the public sector.
Item 13 I have already been invited to practice small acts of corruption to get some benefit in the private sector.
Item 14 I reported some corrupt act.
Item 15 I have already assisted in the process of investigating corrupt acts.
Source: Elaborated by the authors.
The literature indicates specific characteristics to be understood at the level of individual behaviour, segmenting those who have already witnessed (Gorsira et al., 2018), denounced or investigated corrupt acts (M. Bugarin & T. Bugarin, 2017;Independent Commission Against Corruption [ICAC], 2018), have lived with those accused of the crime of corruption (Asian Barometer Survey [ABS], 2016), or have been invited to commit small acts of corruption to obtain some benefit in the public or private sector (Gorsira et al., 2018).
Practically, the behaviour dimension items measure the perception of corruption among respondents who have already had contact with and witnessed corruptive practices and those who have never experienced them. Some questions assess the active and passive positioning of respondents in the face of corrupt practices, both in the private and public sectors, enabling an even more interesting analysis to understand the perceptions of different profiles of respondents.
In the reflex dimension, it is measured how the citizen perceives the 'consequences of corruption for their life and country'. Neshkova and Kalesnikaite (2019) indicate that the consequences of the perception of corruption can be felt in the citizen's political participation; thus, if they assess their government as corrupt and dishonest, they are less likely to participate in democratic government. At the individual level, it is healthy to understand how much citizens feel affected by the effects of corruption (Neshkova & Kalesnikaite, 2019), the reflexes of corrupt actions for their quality of life (Warren, 2004) and their consequent feeling of deprivation of access to public goods or services (Amundsen, 1999;Leal, 2013;World Bank, 1997). In the social context, this dimension aims to understand how corruption can be harmful to the country's development (Abreu & Gomes, 2021), encourage the waste of public money (Amundsen, 1999), as well as being institutionalized in its culture (ABS, 2016;Lin & Yu, 2014). Box 3 lists the CPS items according to reflex dimension.

Code Item
Item 16 Corruption is widespread in the country.
Item 17 I feel particularly affected by corruption.
Item 18 Corruption harms my quality of life.
Item 19 I believe I have already been deprived of access to some good or service due to corruption.
Item 20 Public services offered to citizens are undermined by corruption.
Item 21 Corruption slows down the country's development.
Item 22 Corruption implies a waste of public money. Reflex dimension items assess whether respondents are particularly affected by corruption and its consequences for citizens and the country. Nearly, the idea of this dimension is to measure whether respondents feel directly affected by corruption. And, having the knowledge and knowing how corruption manifests itself, if citizens can identify the consequences brought by corruption. The analysis of this dimension is justified, as we understand that the reflections of corruption are felt differently between those who know about corruption, both in the theoretical and practical fields, or have already been affected by it, and those who have not experienced the phenomenon. Thus, this dimension becomes even more relevant when analysed together with the knowledge and behaviour dimensions, for example.
The control of corruption is another dimension analysed, indicating 'how the citizen perceives the fight against corruption in the country carried out by the State'. The idea is to measure how corruption becomes visible to the citizen and how evident are the strategies to fight it. It is believed that the more evident these strategies, whether they are legally mediated, by state regulatory bodies or by the media, the easier it is for citizens to identify corrupt actions and acts. In this sense, the study by Weitz-Shapiro and Winters (2016) indicates that more educated individuals are more likely to discern reliable and unreliable information, making better decisions and with less misconduct, in addition to demanding a more ethical posture from authorities.
These results should be encouraging for governments such as Brazil, which have invested in the creation of independent and reputable audit and control units. As long as these agencies are able to maintain their high-quality reputation, their influence can be expected to grow as the population becomes increasingly educated (Weitz-Shapiro & Winters, 2016, p. 71).
Abreu and Gomes (2021) highlight democratic levels related to the functioning of government and political participation significantly impacts the results of the perception of corruption. Thus, as shown in the Box 4 items, the premises investigated in this study are in line with the citizen's assessment of the efficiency of state regulatory bodies and legislation in identifying corrupt acts (Abreu & Gomes, 2021), in the transparency in the disclosure of these acts (Kaufmann, 2003) and efforts to punish and fight corruption (ABS, 2016).

Box 4 Items of control dimension
Code Item

Item 23
Regulatory bodies are efficient in identifying corrupt acts.

Item 24
The legislation facilitates the identification of corrupt acts.

Item 25
The efforts currently made are sufficient to fight corruption.

Item 26
Over the past ten years, the country has made progress in fighting corruption.

Item 27
Media coverage helps to fight corruption.
Item 28 People convicted of corruption are correctly punished by the country.

Item 29
The existence of anonymous reporting channels helps to fight corruption.

Item 30
The country is transparent in disclosing corrupt acts.
Source: Elaborated by the authors.
The control dimension items seek to measure the respondents' perception of the practices carried out to prevent and fight corruption, some of them inspired and questioned by the CPI. The analysis of the dimension falls under the perception of efficiency and sufficiency of the efforts to combat corruption, if the punishments are correct and proportional, and if there is transparency in the actions to combat corrupt practices. It is believed that citizens who are more sceptical about the ability to punish the corrupt would have a different perception than those who are more confident.
Finally, the attitude dimension is defined as 'what the citizen thinks/experiences regarding corrupt acts'. Sadek (2019) points out that perceptions are linked to the individual's level of education and their exposure to information and disclosure of corrupt acts. Judgments at a personal level regarding the (un)ethics of denouncing, practicing or tolerating corrupt acts can be configured as important elements to measure the citizen's trust in democracy. Manzetti and Wilson (2007) argue that corrupt governments can withhold support by distributing benefits to citizens, indicating that corruption can be seen as justifiable by a portion of the population. Thus, these authors reinforce, in countries where political institutions are underdeveloped 457-469 Cad. EBAPE.BR, v. 20, nº 4, Rio de Janeiro, July/Aug. 2022 Construction and validation of a corruption perception scale at the citizen level Kelmara Mendes Vieira Monize Sâmara Visentini Ricardo Teixeira Cunha and weak, corruption can increase the participation of citizens, who seek to profit from these corrupt regimes. Neshkova and Kalesnikaite (2019) highlight that at the local level, where the links between the community and public officials are presumably stronger, corruption has a mobilizing effect, with greater tolerance for corrupt acts. Box 5 lists the CPS items according to attitude dimension.

Box 5 Items of attitude dimension
Code Item Item 31 Every person who becomes aware of any act of corrupt practice has a duty to report it.
Item 32 Corruption is justified by bringing benefits to the population. *

Item 33
Tolerating acts of corruption presupposes an unethical personality.
Item 34 I would feel uncomfortable if I knew of some corrupt act and didn't report it.
Item 35 I would vote for candidates investigated for crimes related to acts of corruption. *

Item 36
Carrying out acts of corruption is unjustifiable.
* Inverted questions. Source: Elaborated by the authors.
The items presented in Box 5 are intended to assess the respondents' tolerance level to corrupt practices, as well as whether they are able to conceive realities in which corruption would be acceptable, endorsing such practices, such as voting for candidates investigated by such practices. When answering the items in this dimension, the citizen reflects on how he perceives, tolerates, and feels when having contact or knowledge of a corrupt act or action.
In this research, perception is defined as a process by which the world is represented by the citizen and whose product constitutes your conscious experience available for reporting (Milner & Goodale, 1995). Thus, the CPS assess how ordinary citizens perceive corruption in their country, considering the five dimensions in Figure 1.
The proposal of this Scale differs from internationally known indices, such as the CPI and the CoC, as it does not have interest in measuring corruption in the country, but rather in evaluating how this phenomenon is recognized in society. It is understood that this instrument is necessary and complements the already existing corruption analysis, considering that it can elucidate situations hitherto not understood with known metrics. For example, countries where corruption is not pursued by public bodies and regulatory agencies, nor widely publicized in the media, will hardly have an acknowledgment by the citizen of how harmful the phenomenon can be to society. It is believed that, without combat and dissemination strategies, the possibilities of recognizing the impacts of corruption are mitigated, which could be proven from the application of the CPS.

CPS CONSTRUCTION AND VALIDATION PROCEDURES
The CPS development process began with a literature review on the subject, which provided theoretical support for the definition of its constructs and the initial set of items that comprise it. The literature review was followed by a qualitative approach for validation and refinement of the items. Next, the quantitative step was taken, which involved two more analyses. The first, with an exploratory nature, aimed at an initial validation of the items and dimensions proposed, and the second, with a confirmatory nature, to advance the validation and construction of the theoretical model of the scale. For each of the steps, different samples were obtained according to Box 6.

Box 6 Steps, samples, and objectives of the scale construction process
Step Sample Objective Qualitative Four experts from different areas of knowledge.
The instrument was evaluated regarding the dimension represented by the item, the degree of relevance and the adequacy of the formulation.

Ten individuals from different socioeconomic and demographic profiles.
Pre-test. Semantic analysis to assess the understanding of the scale by all levels of the target population.
Quantitative 420 individuals from all Brazilian regions.
Validation of items and dimensions through the application of exploratory factor analysis.

individuals from all Brazilian regions.
Model validation through the application of confirmatory factor analysis and construction of the scale application methodology.
Source: Elaborated by the authors.
In the qualitative stage, the instrument was evaluated by four experts. Following the recommendation of DeVellis (2016), experts from different areas of knowledge were selected. To assess the level of agreement between the judges, the coefficient of content validity (CCV) and the Fleiss Kappa (Fleiss, 1971) were applied. Subsequently, a pre-test was carried out to assess the suitability of the instrument for application to the population of interest.
For the quantitative stage, the Brazilian population was considered, which according to the IBGE (Instituto Brasileiro de Geografia e Estatística, 2020) is 211,439,266 people, with a confidence level of 95% and a sampling error of 3%, obtaining a minimum sample of 1,075 individuals. The sample was divided into 420 cases for the exploratory phase and 655 cases for the confirmatory phase. The instrument was applied online, between January and February 2021.
In the first phase of the quantitative step, in order to validate the dimensionality of the scale, exploratory factor analysis was performed with the Factor program, version 10.10.01 (Ferrando & Lorenzo-Seva, 2017). A polychoric correlation matrix was used, with the Robust Diagonally Weighted Least Squares (RDWLS) factor extraction method and Robust Promin rotation (Lorenzo-Seva & Ferrando, 2019). The estimation of the number of factors used the optimal implementation of the parallel analysis (Timmerman & Lorenzo-Seva, 2011). To increase the accuracy of the method, the 95% confidence interval for random values was considered (Crawford et al., 2010). For the removal of items, two parameters were considered: 1) factor loadings less than 0.30; and 2) items with cross loads (difference between factor loadings in two factors less than or equal to 0.10). Thus, all items that met at least one of these criteria were removed. As a complement, a factor replicability analysis was performed, based on the H index (Ferrando & Lorenzo-Seva, 2018), which assesses how well the items represent the latent factors found, with values above 0.80 indicating that the factor structure tends to be replicable across studies. Internal consistency was assessed by calculating Cronbach's Alpha (Cronbach, 1951) and McDonalds Omega (ω) (Mcdonalds, 1999), for which values equal to or greater than 0.7 were considered adequate (Hair, Black, Babin, & Anderson, 2014).
In the second phase, confirmatory factor analysis was applied to check the convergent validity, unidimensionality and discriminant validity of the constructs. The models are estimated with the variance-covariance matrix, estimation by maximum likelihood via direct procedure. The convergent validity was analysed by observing the magnitude and statistical significance of the standardized coefficients, using the following absolute fit indices: chi-square statistics (χ²), Root Mean Square Residual (RMR), Root Mean Square Error of Approximation (RMSEA), Goodness-of-Fit Index (GFI); and by the comparative fit indices: Comparative Fit Index (CFI), Tucker-Lewis Index (TLI).
For the chi-square/degrees of freedom ratio, the recommendations are of values less than three; for CFI, GFI, NFI and TLI, values greater than 0.950 are suggested and the RMR and RMSEA should be below 0.080 and 0.060, respectively (Byrne, 2016;Hair et al., 2014;Hooper, Coughlan, & Mullen, 2008;Kline, 2015). Unidimensionality, on the other hand, is evaluated based on the standardized residuals related to the indicators of each latent variable. Constructs that presented, for a significance level of 5%, standardized residuals below 2.58 are considered unidimensional (Hair et al., 2014). For discriminant validity, the chi-square test of differences was applied, for which differences between the restricted model and the free model greater than 3.84 indicate discriminant validity (Bagozzi, Yi, & Phillips, 1991).
Finally, a methodology for standardizing the application of the CPS was developed. At this stage, the perception of corruption is built from the weighted average of respondents' responses in each of the dimensions. The CPS ranges from one to five, and the closer to five, the greater the citizen's perception of corruption.
The research was approved by the Federal University of Santa Maria Research Ethics Committee (CAAE: 37890820.8.0000. 5346), the interviewees read the Informed Consent Form (ICF) before agreeing to participate. The instrument was completely anonymous and data privacy guaranteed by the confidentiality term. The research database is available to readers by sending an e-mail request to the authors.

CONSTRUCTION OF CPS ITEMS
Based on the dimensions and their definitions, developed in the theoretical model, and considering the literature on corruption, we sought to build the items related to each dimension for the scale operationalization. At this stage, the construction techniques recommended by the psychometric literature were considered, such as objectivity, simplicity, clarity, relevance, variety and reliability of the items and the criteria for amplitude and balance of the instrument (Pasquali, 2009).
Boxes 1-5 list the CPS items according to dimensions. All CPS items were constructed considering as response categories the five-point Likert scale (1-strongly disagree, 2-disagree, 3-indifferent, 4-agree, 5-strongly agree).  The mean coefficient of content validity for relevance was 0.906, with item values ranging between 0.829 and 0.996. As for pertinence, the mean CCV was 0.919, with the item values in the range between 0.746 and 0.996. Fleiss' Kappa presented a value of 0.768 (z=22.5; sig<0.001) with values in the dimensions ranging from 0.625 to 0.954. Therefore, the expert assessment pointed to the scale content validity (CCV >0.70) (Pasquali, 2009) and substantial agreement (Kappa>0.6) (Landis & Koch, 1977) among experts.
In the second phase of the qualitative step, a pre-test was carried out to ensure that the items are significant for the target population (Boateng, Neilands, Frongillo, Melgar-Quiñonez, & Young, 2018). The sample of ten citizens was selected for convenience, in order to guarantee the representation of different socioeconomic and demographic profiles. The instruments were applied through interviews in order to assess the face validity of the items. At this stage, respondents reported an adequate understanding of the items, with no need for change being identified.

CPS QUANTITATIVE VALIDATION
Next, the step of quantitative validation was carried out, using exploratory factor analysis. The Bartlett test (4,687.7; sig<0.001) and the KMO (0.885) indicated the factorability of the data. Items 3 and 27 were excluded for presenting factor loadings below 0.30 and item 29 due to cross loading, with the final results of the factor analysis presented in Table 2. Since the five dimensions were considered adequate in the exploratory phase, the second stage of the quantitative step sought to analyse the constructs from a confirmatory perspective. Table 3 lists the results of the fit indices of the initial and final models. For constructs in which the initial models were not adequate, the model improvement strategy was adopted, especially with the removal of variables with low factor loadings. In the process of improving the model, the following items were removed, all with factor loadings less than 0.5: Knowledge -items 1, 4, 5, and 6; Behaviour -items 11, 14 and 15; Reflex -items 16, 21 and 22; Control -items 24 and 26; and Attitude -items 32, 35 and 36. After these exclusions, the final models of the five dimensions met all the fit criteria, which can be concluded for their convergent validity. All standardized residuals were below 2.58, also confirming unidimensionality.
Then, to test the discriminant validity of the constructs, the chi-square difference test was applied. Table 4 lists the chi-square values and degrees of freedom for the restricted model and the free model, as well as the chi-square difference. It is observed in Box 6 that for all sets of constructs, the chi-square difference between the restricted and free models is greater than 3.84, confirming the discriminant validity between each pair of constructs. Therefore, all constructs are discriminating among themselves, indicating that they represent different dimensions of the CPS. Thus, after the confirmatory validation step of the measurement model, the scale maintained the five dimensions proposed in the theoretical model, which are measured by a final set of 18 items. Step 3: Obtain the average perceptions for the entire sample. The average perception in each dimension represents the respondent perception in the dimension. So, for example, for the Perception of Knowledge, the following expression is used: (1) where KWp is the Perception of Knowledge for country p; KWj is the perceived knowledge of corruption for respondent j and n is the number of respondents. A similar procedure should be adopted to calculate the perception for the country in the other dimensions.
Step 4: With the average values for each dimension, it is possible to calculate the Corruption Perception Level, which is constructed by the average of perceptions in the five dimensions, mathematically: ( 2) where: CPL p is the Corruption Perception Level of country p; KWp is the Perception of Knowledge of Corruption in country p; BE p is the Perception of Behaviour towards Corruption of country p; RE p is the Perception of Corruption Reflex in country p; CT p is the Perception of Control of Corruption in country p; AT p is the Perception of Attitude towards Corruption in country p.
Step 5: Classification of Corruption Perception Level. From the values obtained in step 4, it is possible to classify the country's Corruption Perception Level, as illustrated in Figure 2.

Figure 2 Classification of the country's corruption perception level
Source: Elaborated by the authors.
In addition to these steps, it is important that the scale user is aware that the CPS was designed to be self-administered and in online forms. The application using interviews will require the adaptation of the instrument. It is also indicated that the term 'country' present in some items is replaced by the name of the country in which the scale will be applied, providing greater identification for the respondent.

FINAL CONSIDERATIONS
Corruption has grown in scale, magnitude, and sophistication of operations as governments around the world seek new approaches and tools to help identify corrupt activities (Bajpay & Myers, 2020). The increase in corruption increases the need to obtain measurement models for its identification and understanding, from different formats and points of view (public agents, managers, institutions and citizens). As for objective measures, over the past few years, much has been made in the construction of corruption indices capable of allowing comparisons between different countries and sectors. However, from a subjective point of view and focusing on the citizen, there is still no consolidated instrument. Thus, this study aimed to create and validate the Corruption Perception Scale (CPS), which aims to assess how the citizen of a given country perceives corruption.
It is understood that objective and subjective corruption measures are necessary and complementary. While the objective measure aims to present a picture of corruption practices in a country, the perception measure assesses to what extent citizens living in that country are able to assess the existence of corruption. In this context, highly corrupt countries, but with a low perception of corruption by the population, may be fertile fields for the proliferation of corruption, since the population will not assume its role as a social agent, which participates and demands ethical actions in management. On the other hand, in countries where the perception of corruption is high, citizens more aware of the existence and reflexes of corruption, can become active agents against corrupt acts and act with social control. In this sense, the main practical contribution of the study is the construction of a tool that allows all interested researchers and governments to evaluate corruption from a citizen's point of view.
In the construction and validation of the CPS, a series of exploratory and confirmatory techniques were adopted that showed that the scale is capable of being considered valid under different criteria. The final structure of the measure is composed of five dimensions, which seek to assess the perception of corruption in a comprehensive way.
The proposed application methodology presents a simple way to assess Corruption Perception, allowing both the general assessment (level of corruption perception) and in each of its dimensions. For institutions and public agents, the CPS can be useful to analyse the advances and challenges of the corruption reduction agenda, considered an important tool for achieving