The raw data are secondary data on firefighting assistants, regular firefighters, and the reduction in the amount of property damaged (RAPD). The firefighting assistants are VFFs and OFFs. The regular firefighters are FFOs. These data are results from the operation of firefighting resources from 16 metropolitan and provincial government fire departments in South Korea from 2011 to 2018. In South Korea, there are 17 metropolitan and provincial government. Among them, Sejong Special Self-Governing City was created in 2012. In addition, some of data is has been statistical data since 2012.

Qualitative analysis was performed using academic papers, periodicals of the National Emergency Management Agency in South Korea, related laws in South Korea, and major books. Quantitative analysis was performed using statistical analysis. Data for statistical analysis were from the penal structure, which was obtained from “Fire Administrative Data and Statistics”, a National Fire Agency publication. Data on the population (POP) and area (ARE) of metropolitan and provincial governments were obtained from the National Statistical Office. Statistical analysis was performed using statistical package programs (e.g., Stata 14.0 and Excel). The statistical methods used were descriptive statistical analysis, correlation analysis, and panel regression analysis.

Panel regression analysis was conducted to analyze the impact of independent variables on dependent variables and measured the fixed coefficient model and the varying coefficient model. In the fixed coefficient model, a pooled ordinary least-squares (pooled OLS) model was used. In the varying coefficient model, a fixed effect model (OLS-fe) and a random effect model (OLS-re) were used. Before panel regression analysis, multicollinearity and heteroskedasticity were identified by variable inflation factor (VIF) analysis and the Breusch-Pagan/Cook-Weisberg test, respectively. The result of VIF analysis was a value of 4.38; thus, no multicollinearity was observed. The results of Breusch-Pagan/Cook-Weisberg are 167.56 for the chi-square test (chi2) and 0.0000 for the p value; these findings indicate the existence of heteroskedasticity.

Pooled OLS is a method that estimates data in cross-sectional regression while ignoring that the data are panel structures. The probability effect model considering the panel object characteristics is OLS-re. The Breusch-Pagan Lagrangian multiplier test was used to determine whether it was reasonable to estimate the model, pooled OLS or OLS-re. The results showed that chi2 was 1.62, which is a significance probability greater than 0.01. Thus, we estimated pooled OLS rather than OLS-re. Next, the F-test was used to determine whether it is reasonable to estimate the model, pooled OLS or OLS-fe. The F-test tests the following null hypothesis.

If the null hypothesis is accepted, the constant term in the expression below will be equal to alpha (α) for all panel objects. This eliminates the need to consider the heterogeneity of panel objects (equation (2)).

Thus, if the regression model is simply estimated as OLS, a matching estimate can be obtained for the parameter. After all, if the null hypothesis is correct, the model shall be as shown in equation (3).

Figure 1 presents the research model. Independent variables were POP/FFO, VFF/FFO, and OFF/FFO, and RAPD/No. of fire cases. Control variables were POP, ARE, the specific target for firefighting targets (STFF), and urbanization (URB). Definitions of the variables and operational in this study are shown in Table 1.

The established regression equation based on the research model is equal to equation (4).

In this case, the meaning of RAPD/No. of fire case, POP/FFO, VFF/FFO, OFF/FFO, and CV is FFO, VFF, OFF reduction amount of property damaged/No. of fire cases, population/FFOs, VFFs/FFOs, OFF/FFOs, and the control variable, respectively. The meaning of α, β₁-β₃ , β₄ and ε_{i t} is constant, coefficients of independent variables, coefficients of control variables (population, area, STFF, and urbanization), and the probabilistic error term, respectively.

POP/FFO, VFF/FFO, and OFF/FFO were used as independent variables. RAPD/No. of fire cases was used as the dependent variable. POP, ARE, STFF, and URB were used as control variables.

This paper focused on the relationship between firefighting assistants and fire administration. Therefore, VFF/FFO and OFF/FFO were used as independent variables, who are firefighting assistants per firefighting-related law. POP/FFO was also used as an independent variable to relatively compare results in the statistical analysis.

RAPD/No. of fire cases was used as the dependent variable and is one of the products of fire administration. In addition, RAPD/No. of fire cases is directly related to the operation of firefighting assistants in fire departments. For example, when a fire breaks out, the volunteer fire brigade is dispatched to extinguish the fire and pursues that endeavor until a regular fire brigade of fire officials arrives. Alternately, when FFOs are extinguishing fires in the fire-causing area, VFFs are dispatched to the later unit to prevent the fire site from developing into a large-scale disaster. OFFs are also dispatched to the fire site to assist when the number of firefighters is insufficient.

POP, ARE, STFF, and URB were used as control variables. POP, ARE, and STFF were used as major variables for the distribution of firefighting resources in South Korea(13). The variables can be influenced by the operation of firefighting personnel. Additionally, the rule of distribution of firefighting resources in South Korea is based on the expenditure needs theory and the model of the fire standardized index(29-31). In the model of the fire standardized index, major variables, for example, POP and ARE, were used. In addition, research on the distribution of firefighting resources has reported that, for example, URB and the location of the fire department are major variables of fire service demands. Because of these reasons, URB was used as control variable.

The operational definition of the selection of variables is as follows. An FFO was defined by the “Fire Officers Act” as a public officer(32). A VFF was defined by the “Volunteer Fire Brigade Establishment & Operation Act” as an individual who assists with activities such as fire suppression, rescue, and emergency relief and with fire prevention activities at a disaster site(10). An OFF was defined by the “Act on the Establishment of Obligatory Firefighting Unit” as an individual who assists in activities such as fire suppression, rescue, and emergency relief and with fire prevention activities at a disaster site(11). An RAPD was defined as the amount of economic loss prevented through firefighting activities(33): in other words, the value of property subject to fire minus the amount of property damage caused by fire. If all the buildings burn down despite the firefighters being dispatched to the fire scene, the RAPD is zero.

Table 2 shows the statistical hypothesis of this paper.

Tables 3 and 4 present data of the independent and dependent variables. Figures 2~5 are a three-dimensional graph based on Tables 2~5. Table 6 presents the result of a basic statistical analysis of independent and dependent variables from 2011 to 2018.

Table 3 and Figure 2 display data for POP/FFO, and a decrease from approximately 1244 in 2011 to 945 in 2018 was observed. Table 4 and Figure 3 present data for VFF/FFO, and a decrease from approximately 2.67 in 2011 to 1.97 in 2018 was observed. Table 5 and Figure 4 display data for OFF/FFO, and an increase from approximately 0.007 in 2011 to 0.027 in 2018 was observed. Table 6 and Figure 5 present data for RAPD/No. of fire cases, and an increase from approximately 122,235,000 won/incident in 2011 to 512,285,000 won/incident in 2018 was observed.

In Table 3 and Figure 2, POP/FFO continues to decline: POP/FFO was 1224.22 in 2011 and decreased to 944.84 in 2018. If POP/FFO continues to increase until 2022, it is expected to be similar to that of the United States (911) and Japan (779)(34). This result is interpreted as a result of to be due to the continued recruitment of FFOs owing to increased demand for fire safety and recognition of importance. On the other hand, POP/FFO varies greatly by the type of metropolitan government. The metropolitan city and Gyeonggi province had an average of 1300 residents per fire official for 8 years. The result can be attributed to the use of POP and ARE as major variables of the distribution of firefighting resources. According to Attachment 2 of the “Regulations on the Establishment of Local Firefighting Agencies”(35), the criteria for the establishment of 119 safety centers are at least 50,000 residents or 2 Km² of jurisdiction area. The criteria of small towns are at least 10,000 residents or 20 Km² of jurisdiction area. In small and medium-sized cities and small-scale towns, the size of the population decreases, and the jurisdiction area widens. For this reason, compared with metropolitan cities and Gyeonggi province, the remaining metropolitan governments have a low resident to fire official ratio.

In Table 4 and Figure 3, VFF/FFO continues to decline: VFF/FFO was 2.67 in 2011 and decreased to 1.97 in 2018. The impetus of this finding is the continued recruitment of fire officials due to the aforementioned increased demand for fire safety and recognition of importance. Additionally, the tendency of VFF/FFO is similar to that observed in Table 3 and Figure 2, which differs greatly by the type of metropolitan and provincial governments. Particularly, the average of the metropolitan cities and Gyeonggi province was 1.1 VFF/FFO, and the remaining metropolitan governments had an average of 3.8 VFF/FFO. In other words, compared with metropolitan cities and Gyeonggi province, the remaining metropolitan and provincial governments operated more VFFs. The results are attributed to limitations on the rule of distribution of firefighting resources in South Korea. Additionally, the rule of distribution of firefighting resources in South Korea is based on the expenditure needs theory and the model of the fire standardized index(29-31). On the basis of this theory and model, the optimal deployment scale of FFOs was calculated using population and area variables(36). This theory and model are based on the logic that if the geographical distribution or density of the population is the same, the increase in demand for firefighting administrative services is likely to be affected(29-31). However, the population decreases, and the jurisdiction area widens as the small cities and small towns decrease in size. Thus, population and area variables do not work independently. This result contradicts the logic that demand for fire administration services increases as areas increase under the assumption that the population distribution between regions is the same. In addition, transportation environments can serve as a crucial variable for providing quality firefighting administrative services, which are difficult to explain. For this reason, it can be inferred we infer that compared with metropolitan cities and Gyeonggi province, the remaining metropolitan governments are utilizing more VFFs to provide fire administrative services to citizens on behalf of the insufficient number of FFOs. In addition, it is indirect evidence of an imbalance between firefighting resources and fire safety service levels that depends on the financial conditions of local governments, which had been claimed by various fields(37-39) before they were changed to state fire officials.

In Table 5 and Figure 4, OFF/FFO increases from 2011 to 2016 and decreases again in 2017 and 2018. Specifically, OFF/FFO was 0.007 in 2011 and 0.026 in 2018. This result can be attributed to the selection of OFFs. From 2011 to 2018, the fixed number of OFFs was 2000, but the number selected was 256 persons in 2011, 650 persons in 2012, and approximately 1150 persons from 2013 to 2018. The obligatory fire brigade is one of the substitute military service systems in South Korea and was introduced to increase the efficiency of firefighting operations and protect the lives and property of individuals by expanding the field activities that are insufficient because of the extent of the demand for fire administration(11). The obligatory fire brigade was established in the wake of a house fire in Hongje-dong, Seoul, in which six FFOs died and three FFOs were injured(37). The “Act on the Establishment of Obligatory Firefighting Unit” was enacted in August 2001 to increase efficiency by expanding the number of firefighters and strengthening the ability to respond to disasters such as fire; in response, 211 OFFs were recruited and began their service with their entry into the Army Training Center on March 29. Additionally, it will be abolished in 2023 after the last recruitment in the first half of 2021.

In Table 6 and Figure 5, RAPD/No. of fire cases continues to increase: from approximately 120 million won in 2011 to approximately 510 million won in 2018. This result is interpreted as economic growth due to social development, and the amount of fire damage caused by large-scale disaster accidents is increasing. Busan in 2014, Gyeonggi province in 2018, and Jeju Island in 2018 show relatively high fire relief, inferred from the large fire in the area in the corresponding year. For example, on August 28, 2018, a fire (property damage was approximately 11 billion won) occurred at Daeyeon Process in Taejeon-dong, Gwangju, and on October 7, 2018, a fire (property damage was approximately 8.8 billion won) occurred at Daehan Oil Pipeline Corporation in Hwajeon-dong, Deokyang-gu, Goyang City.

Table 7 presents the result of correlation analysis between variables. RAPD/No. of fire cases, POP/FFO, VFF/FFO, and OFF/FFO showed no correlation. According to the literature, a Pearson correlation coefficient from 0.8 to 1.0 is very strong positive (+), 0.6 to 0.8 is strong positive (+), and 0.2 to 0.4 is normal positive (+)(40).

Table 8 presents the estimation result of a panel regression model that reflects the heteroskedasticity. Table 9 shows result of research hypothesis. Regarding the result of the F-test in Table 8, if the null hypothesis is accepted, pooled OLS estimation is chosen, and if the null hypothesis is rejected, an OLS-fe model is chosen. The result of the analysis of this study is that the p value of the test statistic is less than 0.01 and that the null hypothesis is rejected at the 1% significance level; thus, we have determined that the fixed effect model is more appropriate than random effect.

The OLS-fe results demonstrated that POP/FFO, VFF/FFO, and OFF/FFO did not significantly affect RAPD/No. of fire cases. Notably, negative regression coefficients are shown for POP/FFO and VFF/FFO, and positive regression coefficients are shown for OFF/FFO. In other words, if POP/FFO and VFF/FFO decrease, RAPD/No. of fire cases increases, and if OFF/FFO increases, RAPD/No. of fire cases increases. The panel analysis results did not demonstrate a significant value; the reason for this finding may be that the demand for firefighting changed from fire control to fire prevention, rescue, and first aid(6). In addition, the reason may be that the ratio of the general volunteer fire brigade, local volunteer fire brigade, and dedicated volunteer fire brigade for extinguishing a fire increased, but the specialized volunteer fire brigade for preparations for large or special disasters did not(13).

In relation to the literature on the distribution of firefighting resources, panel analysis may provide possible evidence to support the opinion that an increase in FFOs has a more positive effect on firefighting mitigation than an increase in VFFs(29-31) and that additional firefighting personnel should be deployed in areas where there is high demand. According to the literature, firefighting policies for VFFs have changed significantly since 2014, when the “Volunteer Fire Brigade Establishment & Operation Act”(41) was enacted. In other words, in 2014, the number of local volunteer fire brigades decreased and the number of specialized volunteer fire brigades increased. The specialized volunteer fire brigade should comprise specialized technical and qualified personnel, and so forth, reflecting regional specialties(42). The specialized volunteer fire brigade was organized to prepare for major disasters that reflect redundancy rather than the purpose of firefighting the initial fire. The trend of an increase in specialized volunteer fire brigades is interpreted as a change in the operation of volunteer fire brigades from the initial firefighting of fire sites to the firefighting assistants of large disasters. For this reason, in terms of RAPD, we infer that the increase in the FFOs has a more positive effect on the reduction of fire damage than the number of individuals in the volunteer fire brigade. In conclusion, it suggests that the number of fire officials corresponding to regular firefighters should be increased to improve the reduction of fire damage and that full-time firefighters should be recruited rather than volunteer firefighters, who are part-time firefighters.