THE GUATEMALAN CONSTRUCTION CHARACTERIZATON OF THE PERCEIVED RISK BY MANAGERS OF SUFFERING WORK ACCIDENTS

. The construction industry is considered one of the highest risk production sectors, even more so in developing countries such as Guatemala. A characterization has been carried out on the perception of Guatemalan construction company managers regarding, the risk of accidents exist for the different activities they perform. The characterization has been carried out on a representative sample of the business population via a questionnaire. A preliminary data analysis was performed followed by a Descriptive and a Multiple Correspondence Analysis. Companies are characterized as “medium-size” companies, with an average of 81.1 construction workers per year and average annual turnover of 1.29 million euros. 4 clusters of construction activities occur with similar accident weightings. Companies in Cluster 1 are associated to the variables grouped with a Low risk weighting, with a medium to high number of on-site workers and with a turnover of more than 100,000 euros. In contrast, those in Cluster 2 are associated with the variables grouped as having a Medium risk weighting, with a low number of on-site workers and a turnover of less than 100,000 euros. The companies in Cluster 3 are only associated with High risk weighted. And those of Cluster 4 with Not applicable risk-weighting variables.


Introduction
The construction industry is considered one of the highest risk production sectors given that accident rate stays continuously high over time (Hassan et al., 2007;Pérez-Alonso et al., 2012;Rodríguez-Garzón & López-Alonso, 2013;Sousa et al., 2014;Carrillo-Castrillo et al., 2017;Hernández-Arriaza et al., 2018;Liang et al., 2018;Yuan et al., 2018;Fargnoli & Lombardi, 2019;Peng & Chan, 2019;Ayhan et al., 2020;Mihić, 2020). To prevent accidents, it is necessary to carry out exhaustive maintenance, regular safety inspections, continuous safety training and develop an accident investigation plan (Goh et al., 2016). In addition, there are companies that not only possess such controls, but also highlight them to have greater rel-evance amongst companies that work with transparency and wish to obtain greater legitimacy (Berry et al., 2009). However, this is not always the case, the reason being an increase in the work cost (Yilmaz & Kanit, 2018) -this results in significant economic and social impact, in many cases, even on-the-job training of staff is avoided (Cheng et al., 2010). Moreover, older workers are taken on, which in turn, means they are more likely to suffer accidents with a higher proportion of fatalities (Chiang et al., 2018). There are studies that link a company's accident rate to its own economic performance (Forteza et al., 2017). Thus, a significant positive linear relationship was obtained between the site risk and the accident rate, as well as a significant quadratic relationship (an inverted U-shape) between the accident rate and the economic performance of the company.
The determinant factors for this high accident situation can be found in the peculiarities of the construction sector (Cheng et al., 2010). Indeed, the combination of aspects -such as frequent employee turnover, outsourcing and strict contract deadlines, outdoor work in all weather conditions, the lack of highly skilled workers, the use of obsolete work equipment and frequent changes of job sites as well as the presence of different companies working on the same job site simultaneously -all contribute to the high accident and injury rate (Carrillo-Castrillo et al., 2017).
To control the high accident rate caused by the interaction of these multiple factors, adequate safety management is necessary along with a causal accident analysis (Hernández-Arriaza et al., 2018;Wong et al., 2018). Likewise, hazard identification in the construction industry is subject to a larger number of variables and unknowns than in other manufacturing industries making the hazard identification process more difficult and resulting in many injuries and fatalities (Mihić, 2020).
According to Solís-Carcaño and Franco-Poot (2014), psychological science defines perception as a cognitive process, inferential and constructive in character, by which a subject can generate an internal representation of what happens externally, based on information collected by the senses along with memory information. Several studies have shown that, in general, the perceptions of management and workers differ so it is interesting to understand the perceptions of a construction company's management safety practices both from the managers' own standpoint and from that of the workers (Wei et al., 2002;Reese & Eidson, 2006). For the latter, the work of Gillen and collaborators (Gillen et al., 2004) stands out, in which workers identified the management's commitment to safety, their concern for workers, the consistency between spoken messages and practice, professionalism and communication skills as key qualities in successful managers. Mohamed (2002) also indicates that the safety environment on construction sites is closely related to the workers' perceptions and beliefs regarding the safety issue.
Accidents occur due to three main causes (Abdelhamid & Everett, 2000): not identifying unsafe conditions before starting a job, continuing to work on-site even though someone has identified unsafe conditions, and deciding to act in an unsafe way regardless of the workplace conditions. The latter two causes are clear examples of perceived reality being a strong component that can influence workers' behaviour (Solís-Carcaño & Franco-Poot, 2014). In many cases, it is believed that the construction personnel themselves must possess their own innate risk perception and that this should be enough to avoid accidents; however, nothing is further from the truth because, at the very least, the training variable significantly alters the risk (Rodríguez-Garzón et al., 2014 and the worker's age variable varies risk perception between younger and older workers (Peng & Chan, 2019;Idrees et al., 2017), among other variables.
On the other hand, more attention needs to be paid to construction tasks and activities (Rodríguez-Garzón & López-Alonso, 2013), given that task-level studies account for only 2.28% of all health and safety research in the construction industry (Zhou et al., 2015). There is a lack of workplace risk-exposure measurements done in the field because most research tends to be epidemiological and focused primarily on accidents (Swuste et al., 2015). Thus, in a study with construction workers in Spain (Rodríguez-Garzón & López-Alonso, 2013), it was shown that the job or construction task performed by the worker, presented statistically significant differences with respect to what they thought about the risk knowledge of their work that those responsible for the company have.
Guatemala is no exception -according to hospital registers of IGSS accidents (2016), amongst the manufacturing-construction-service sectors in 2016, 28.15% of accidents corresponded to construction; while in 2017, it accounted for 24.75%. As in any developing country, the risk of health loss is 10 to 20-times higher than in industrialized nations (Dong, 2005;Tadesse & Israel, 2016). Construction workers in Guatemala are six times more likely to die on the job than their counterparts in Switzerland (International Labour Organization [ILO], 2001), so strategic priority should be given to their safety (ILO, 2008), especially because the annual construction growth in Guatemala was 2.7% in 2017 and 3.1% in 2018 (Cámara Guatemalteca de la Construcción, 2018). There is a register of 7,720 construction companies (Directorio Nacional Estadístico de Empresas [DNEE], 2015) comprising micro, small, medium and large-scale companies, more than 50% of which are concentrated in the Guatemala department although the majority of their operations are reported as being in other regions of the country.
For this reason, Hernández-Arriaza et al. (2018) characterized occupational risk prevention in Guatemala's construction industry, correlating organizational parameters and the size of the company with parameters for health and safety prevention and management activities both in the company and on site. But the perception of accident risk was not characterized in the different construction tasks or activities carried out in the Guatemalan construction industry according to their managers. However, as Wei et al. (2002) indicate, it is interesting to know the perceptions from safety practices of the management in the construction company.
For all these reasons, the objective of this work is to characterize the perception of accident risk that managers of the Guatemalan construction industry have in the different construction tasks or activities that they carry out in their construction activity.

Research design and data analysis
To characterize the perception of Guatemalan construction company managers regarding the accidents their workers might suffer, and which risks exist that cause accidents in the different construction activities carried out, the following study variables have been adopted: 5 general company variables (annual company turnover (C), the number of office workers (D), the number of on-site workers (E), the number of work crews per year (F), the number of years working on site (G)) and 64 accident risk weighting variables (see Tables 1 and 2). In Tables 1 and 2, all the variables are shown and the nomenclature for the variables and their categories are explained.
Field data acquisition was carried out by sampling Guatemalan construction companies using a questionnaire designed for the purpose. Once the data were obtained, a preliminary data analysis was performed, along with a descriptive analysis of the studied variables, determining the frequencies (%) for qualitative and mean values, and the standard deviation (s.d.) for quantitative values. In addition, a multiple correspondence analysis was carried out with the 64 accident-risk weighting variables for the different construction activities and the 5 general company variables, allowing us to determine clusters of construction activities with similar weightings in which accidents occur.
Finally, to assess the possibility of bias produced by the non-response of the companies to which the questionnaire was sent and did not answer, an analysis of simple variance (ANOVA) was carried out for the 3 most significant quantitative variables of the study (annual company turnover, number of office workers, number of onsite workers), for the top 20 and the last 20 respondents. These 3 variables were considered the most significant of the 5 general variables (quantitative) of the company studied, because they presented higher correlations with the 64 qualitative variables. All data analysis was performed with the SPSS Statistics V17.0 program.

Company census
Guatemalan construction companies as a whole were considered in the population study, of which there were 7,720 as of December 2015, according to data from Guatemala's National Statistical Institute (DNEE, 2015). However, in order to have reference data for them, data from Guatemala's Ministry of Communications, Infrastructure and Housing (MICIVI) was taken as a census sample; this showed 1,954 companies as of December 2015.

Technique and sampling questionnaire
To carry out the sampling, a simple random sampling technique was used and a specific questionnaire was de-signed to collect the information. This was structured into 2 sections (or groups of variables): 1. General company data: 5 variables (see Section 1.1). 2. The weighting for suffering an accident in 64 different construction activities: 64 variables (see Tables  1 and 2). The weightings for suffering an accident are: 1. Low; 2. Medium; 3. High and 4. None or Not applicable. All the variables and their nomenclature are shown in Tables 1 and 2.

Sample size, sampling plan and data collection system
The sample size, sampling plan and data collection system is the same as that performed by Hernández-Arriaza et al. (2018). Three hundred and fifty surveys were sent to Guatemalan construction companies, of which 100 were completed and returned with significant responses for analysis. The sampling phase took place from March 2016 to December 2017.

General company characteristics
The companies studied show an average annual turnover (C) of 1.29 million euros (s.d. 6.29), an average number of office workers (D) of 7.3 (s.d. 15.06), an average number of on-site workers (E) of 81.1 (s.d. 75.22), an average number of work crews each year (F) of 6.8 (s.d. 5.55) and the companies involves in on-site work (G) for an average of 15.0 years (s.d. 7.53).

Guatemalan construction company managers' perception regarding the risk of their workers suffering an accident
As Tables 1 and 2 show the weighting frequencies of suffering an accident for each of the 64 different construction activities that might be performed in the Guatemalan construction industry.
As for the 3 construction activities that stand out as having a High-risk probability, these are: high voltage electrical connections and networks (R51) (45.1%), using cranes and loading booms (R53) (38.4%) and using manual, electric or combustion lifts (R51) (27.8%).

Multiple correspondence analysis
This epigraph shows the results of the multiple correspondence analysis performed on the 64 risk-weighting variables and the 5 general company variables (Tables 3  and 4). A two-dimensional model has been obtained summarizing the information for all the analysed variables. For interpretation purposes, the model generates the following parameters (the results of which are presented below): Measures of discrimination, quantifications and object (company) scores.
As already indicated, the model obtained has two significant dimensions and provides adequate reliability since the first dimension explains 39.8% of the variance with a Cronbach's α coefficient of 0.978 and eigenvalue of 27.491 while the second explains 29.7% of the variance with a Cronbach α coefficient of 0.965 and a eigenvalue of 20.505. For the factorial model as a whole, the average for the explained variance is 34.8%, the mean coefficient of Cronbach's α is 0.972 and the average eigenvalue is 23.998.  Tables 3 and 4 show the discrimination measures for each variable with respect to each of the two dimensions and the mean. If these values are represented in an orthogonal axis system, a discrimination measure figure is obtained for the variables in the model, as shown in Figure 1, where you can observe the correlations of the variables. It should be noted that there is greater correlation between variables if the angle formed by the lines that join the origin of the coordinates with each of the variables is small for each pair of variables analysed; conversely, if the angle is very large between two variables, this indicates that a good correlation does not exist between them.
In addition, similar discrimination measures for a variable in the two dimensions reflect difficulties in assigning it to a given dimension. Ideally, a variable has a high value in a single dimension and a low value in the other, as with variables R7, R8, R29, R60, R61, and R62, which correlated more with dimension 1, and therefore this dimension better discriminates against the categories for these variables; whereas variables R17, R34, R37, R41, R43, R57, and R59 are more correlated with dimension 2, so this dimension better discriminates against the categories for these variables.
It should be noted that the 5 general company variables are not significant in the mathematical model obtained because the discriminations in both dimensions are very small and close to the origin of the coordinates, and none clearly correlate to either of the 2 dimensions. However, Figure  1 shows which of the 64 risk-weighting variables correlate with each of the general company variables. Hence, the company's annual turnover (C) variable and number of years that the company has been working on sites (G) variable correlate with the risk-weighting variables found in the circular sector ranging from R17 to R21, while the number of work crews for the company per year (F) variable aligns with variable R21 and all the others align with the R21 variable. Likewise, the number of company workers (E) variable correlates with the risk-weighting variables found in the circular sector from R21 to R27. Conversely, the number of company office workers (D) variable does not clearly correlate with any risk-weighting variable since it is located practically at the coordinate origin.
In addition, the model obtained allows us to identify the categories for each variable that most discriminates the objects (companies) thus, the quantifications of the variables are obtained and represented on a factorial plane (Figure 2). Category quantifications are the average scores for objects in the same category (Hernández-Arriaza et al., 2018). Figure 2 shows the positions of the four possible categories (1. Low, 2. Medium, 3. High and 4. None or Not applicable) for the risk-weighting variables that have measure of discrimination averages greater than 0.5; this makes the figure clearer because, if the 64 weighting variables were included, the values would overlap with each other and the representation would be confusing.
Using the Factorial Plane representation (Figure 2), one can observe the correlations (or correspondences) of the variable categories. One can see how the 4 riskweighting variable categories, together with the 5 general company variable categories are grouped into 4 clusters.
Finally, the obtained model allows the objects (companies) to be represented on a factorial plane using each of their scores in each of the two dimensions (Figure 3). Based on the scores of the objects obtained for each company in the mathematical model developed (represented in Figure 3), one can see how the sample companies are grouped into 4 clusters with homogeneous company characteristics, and are also associated with the 4 clusters of variable categories described above in Figure 2. Table 5 describes the characteristics of these 4 clusters identified in Figures 2 and 3.

Response rate and sampling bias
In accordance with the sampling plan, 350 surveys were conducted, of which 100 valid survey responses were received, making the effective response rate 28.57%, which is slightly less than the 38.0% obtained by Chen and Mohamed (2009) in Hong Kong for a sampling study of construction contracting companies, and well below the 61.11% obtained by Idrees et al. (2017) for a study looking at workplace safety perception on 5 construction projects in Pakistan. Table 6 shows the results of the analysis of variance carried out to verify the non-existence of non-response  bias in the sampling phase. It is observed that the p-value for the three variables analyzed is greater than 0.05, so there are no statistically significant differences between the 3 groups of respondents.

General company characteristics
Guatemalan construction companies have an average annual turnover (

Descriptive analysis
When analysing the results for the risk weighing of workers suffering an accident in the different construction activities or tasks (64 variables) carried out by Guatemalan construction companies, it should be noted that, of the 64 activities studied, in 35 (54.7%) the risk weighting was indicated as being Low; in 20 (31.3%) it was Medium; in 3 (4.7%) it was High and in 6 (9.4%) it was None or Not applicable. This significant percentage of companies that have a Low or Medium risk weighting for suffering on-site construction accidents clearly indicates the minimal company management and training on accident risks given to workers, as shown by Rodríguez-Garzón et al. (2014, who stated that, when workers receive low levels of training, they have less risk perception; in contrast, when there is adequate, high-level training, workers have greater risk perception. This is the case for construction compa-nies in Guatemala -42.4% of them do provide training whereas 53.6% indicate that they only provide it sometimes (Hernández-Arriaza et al., 2018), which does not ensure it occurs and is the cause of safety shortcomings (Tam et al., 2004). Also, Solís-Carcaño and Franco-Poot (2014) looking at medium, small and micro construction companies in Mexico (which account for 97% of the sector overall) indicated that these companies do not provide the minimum safety or training conditions, nor do they establish risk prevention programmes; they added that the argument often made to justify this situation is that such companies face the permanent challenge of simply trying to survive. Furthermore, they state that the perception of a particular reality in the workplace is a strong component that can influence worker behaviour. Likewise, in a study with construction workers in Spain (Rodríguez-Garzón & López-Alonso, 2013), it was concluded that the workers who carry out the structures (structural workers) have more training that the group of masonry workers; additionally, the greater the training received by the worker, the lower the trust in the health and safety head of the company. In addition, as Hinze (1997) pointed out, the perception of worker safety and individual safety behaviour is affected by demographic factors such as gender, age, experience, education, employment level, marital status and the number of dependent family members. In this sense, Idrees et al. (2017), in a study of construction workers in Pakistan, found that older workers are more aware of safety while young workers have more accidents; moreover, workload and job satisfaction are significantly dominant factors in the safety perception of older workers, while organizational relationships, mental stress and job security are dominant factors for younger workers. In addition, several authors highlighted the need to analyse the particular characteristics of the tasks or work activities carried out in construction (Hassan et al., 2007;Mitropoulos et al., 2009;Rodríguez-Garzón & López-Alonso, 2013;Di Pasquale et al., 2015;Sadeghi et al., 2015;Fargnoli et al., 2018). Mitropoulos et al. (2009) justified this need since regulatory approaches do not adequately consider the characteristics of work processes. Hassan et al. (2007) in a comparative construction safety study on large and small-scale projects revealed that large-scale projects showed a high and consistent level of safety whereas small ones had low and varied levels of safety. The determinants for these differences in safety levels were organizational commitment, factors influencing workmate communication, worker-related factors, personal-role and supervisor-role factors, safety barriers and safety behav- iour factors as well as management commitment at all levels in accordance with the management structure and the behaviour in the taking of risk factors. Indeed, improving human safety depends on reducing the risks that workers are exposed to when carrying out specific work activities taking human factors into consideration (Fargnoli et al., 2018). This means that more attention should be paid to analysing human behaviour (Di Pasquale et al., 2015), and the factors that can affect it, such as stress, training, experience, the relationship with management, etc. (Sadeghi et al., 2015). Therefore, it is important to know the perception of risk for the different construction tasks or activities in Guatemalan companies. Thus, the Medium risk (not High) perception result for scaffolding and digging tasks is consistent with the results obtained by Hassan et al. (2007) on building workers in Malaysia; since they obtained safety scores for these high tasks both for large and small-scale projects. However, this was not the case for construction workers in Mexico (Solís & Arcudia, 2013), in which one of the main hazards was the use of ladders and scaffolding. As for the 3 construction activities that stand out as having a High-risk probability, are consistent with a study on construction workers in Mexico (Solís & Arcudia, 2013), in which the main hazards identified were: the collapse of structural elements, the use of ladders and scaffolding, electric shocks and working in confined spaces. In a study with construction workers in Spain (Rodríguez-Garzón & López-Alonso, 2013), it was observed that there were statistically significant differences between bricklayers and structural workers, with respect to what the workers thought about the risk knowledge of their work that those responsible for the company have (important risk perception factor), being higher in bricklayers.
Finally, the 6 construction activities highlighted because they are indicated as having "None or Not applicable" risk probability, are hardly carried out by Guatemalan construction companies when compared to the other activities; neither is gas pipeline laying carried out by smallscale Malaysian building companies (Hassan et al., 2007).

Multiple correspondence analysis
After performing the multiple correspondence analysis on the studied variables, a two-dimensional mathematical model was obtained, summarising the information from all the analysed variables and presenting good reliability.
From Table 6 and Figures 2 and 3, and despite there being no clear trend for the 5 general company variables, (which present very small discriminations close to the origin of the coordinates), it can be concluded that companies in Cluster 1 are associated to the variables grouped with a Low risk weighting, with a medium to low number of work years, a medium to high number of on-site workers and a medium to low number of office worker with a turnover of more than 100,000 euros. In contrast, those in Cluster 2 are associated with the variables grouped as having a Medium risk weighting, with a high number of work years, a low number of on-site workers, a mediumto-high number of office workers and a turnover of less than 100,000 euros. The companies in Cluster 3 are only clearly associated with the variables grouped as High risk weighted, and the number of company work years of between 31 and 40 years. Finally, the companies in Cluster 4 are associated with most of the "None or Not applicable" risk-weighting variables, in which no general company variable category is clearly associated. In summary, Clusters 1 and 2 (Low and Medium risk weighting, respectively) are where the largest percentage of analysed companies is concentrated, and this Low risk weighting is a clear indication of minimal accident-risk management and training for company workers, as shown by Rodríguez-Garzón et al. (2014 furthermore, these companies assume that the worker should know what to do himself/herself to avoid suffering on-site accidents (Aboagye-Nimo et al., 2015).

Conclusions
A characterization has been carried out on the perception of Guatemalan construction company managers regarding the risk of accidents that their workers might suffer and what accident risks may exist in the different activities performed.
The 3 construction activities carried out by companies that most stand out as having a High-risk probability are: High-voltage power connections and power grids (45.1%), the use of cranes and loading booms (38.4%) and the use of manual, electric or combustion lifts (27.8%).
Using the Multiple Correspondence Analysis technique, it has been concluded that companies are grouped into 4 clusters that have similar characteristics in terms of the accident-risk weighting they present in the particular construction activities performed.
Companies in Cluster 1 are associated to the variables grouped with a Low risk weighting, with a medium to low number of work years, a medium to high number of onsite workers and a medium to low number of office worker with a turnover of more than 100,000 euros. In contrast, those in Cluster 2 are associated with the variables grouped as having a Medium risk weighting, with a high number of work years, a low number of on-site workers, a medium-to -high number of office workers and a turnover of less than 100,000 euros. The companies in Cluster 3 are only clearly associated with the variables grouped as High risk weighted, and the number of company work years of between 31 and 40 years. Finally, the companies in Cluster 4 are associated with most of the "None or Not applicable" risk-weighting variables, in which no general company variable category is clearly associated.
Companies in the Guatemalan construction sector should improve the occupational risk-prevention training of both their managers and workers in order to the accident-risk weighting of the construction activities they carry out were more grounded.
It should be noted that these conclusions obtained from the results of the study are conditioned by the following limitations: -The study was carried out by surveying the managers of the construction companies in Guatemala and not the workers directly, therefore the data obtained from risk perception, are given from the perspective of the managers, which is interesting as several authors indicate (Wei et al., 2002;Reese & Eidson, 2006). -In addition, to develop the present study, we considered a sample made up of 5.1% of companies in the current Pre-qualified Companies Census from the Guatemalan Ministry of Communications, Infrastructure and Housing (MICIVI), as of December 2015; therefore, it is an estimate based on the companies in the sample -for other companies, it might be different. Finally, the following two future research topics are proposed that are related to the present study: -It would be necessary to propose a new work in which the risk perception measured directly on the workers was analyzed through some of the methods that exist based on psychometric scaling, and thus contrast the results with those obtained from the perspective of the managers. -Study the economic cost for construction companies in Guatemala of everything related to the prevention of occupational hazards and correlate it with the accidents suffered by these companies.