Assessment of Public Flood Risk Perception and Influencing Factors: An Example of Jiaozuo City, China

Abstract

There are many studies showing that public flood risk perception may promote people’s motivation to reduce flood risk and enhance their coping behavior, thus providing useful insights for flood risk management. The purpose of this article is to estimate residents’ flood risk perception in Jiaozuo City and to identify the influencing factors. A questionnaire survey method was used to collect data and a composite index was constructed to measure public risk perception. Each respondent’s grade of flood risk perception was calculated using the relationship between the standard deviation (SD) and the mean value (MV) of flood risk perception index (RPI) scores. Moreover, the hypotheses concerning different groups were tested using an independent sample T-test and one-way ANOVA (analysis of variance), and the group differences in flood risk perception on each observed dependent variable were explored using post hoc tests. The flood risk perception of the total respondents was divided into three levels based on the SD and MV of RPI scores: low (68.4%), moderate (13.7%), and high (17.9%). Respondents with low education, low income, less flood experiences, and who have married, lived in rural areas or near rivers/reservoirs had a higher flood risk perception than others, and respondents who lived in flood storage areas had a lower risk perception. Moreover, the ability to mitigate floods and the trust in flood-control projects were negatively related to the flood risk perception.

1. Introduction

Flooding is one of the most frequent and damaging natural disasters that has the potential to inflict major losses and damages on human life, property, infrastructure, the environment, and public services [1]. The frequency and severity of floods are expected to increase because of global climate change [2]. Damaging floods frequently occurred in China. The report of the Ministry of Water Resources of China showed that the flood disasters had caused financial damage amounting to nearly USD 763.78 billion and more than 61 thousand deaths from 1990 to 2020. Flood risk is the outcome of the simultaneous occurrence of hazard and vulnerability [3]. The hazard analysis focuses on the intensity and frequency of floods, which could be increased due to global climate change and enhanced human activity in flood-prone areas [4,5]. The vulnerability mainly refers to the characteristics of the potential damage and the local resilience in flood-prone areas, which is related to the lack of proper mitigation strategies and human development in exposed areas [3,6]. For example, the study by Bubeck et al., found that the improved preparedness could reduce the substantial damage caused by flood disasters [7]. Furthermore, people’s behaviors towards flood disasters were influenced by their risk perception attitude [8,9]. Therefore, understanding the public flood risk perception is conducive to formulating and implementing flood risk management policies [10,11]. Several studies have focused on flood risk perception. For example, Salukele et al. [12] identified that flood risk perception had a positive impact on coping strategies. Houston et al. [13] pointed out that a higher risk perception could help people to implement correct and timely flood mitigation actions. Liu and Jiao [14] found a positive correlation between flood risk perception and flood coping behaviors. In addition to floods, there are also some other natural hazards. Lindell and Perry [15] confirmed that people’s risk perception had an impact on their adjustment strategies to earthquake hazards. Peacock [16] revealed the correlation between hurricane risk perception and wind protection measures adopted by homeowners in Florida. However, some research provided different results. For example, some studies found that people with a higher risk perception also may not take appropriate preparedness actions [17,18,19].

The study of risk perception started around 1940 [20]; White published his groundbreaking thesis on human adjustment to floods and found that previous flood experience could directly influence people’s behaviors. This work opened the way for risk studying on human dimensions in the field of multi-hazards [21]. The key paper to study public risk perception was carried out by Starr in the 1960s [22]. He pointed out that the risk acceptability was not only related to the evaluation of the profit of the risk, but should consider people’s preferences, and proposed that society could find a “basic optimal” balance between the risk and the benefits of any activity through a trial-and-error method. This method laid the theoretical foundation for the studies of risk perception. Risk perception can be defined as the combination of “perceived probability” and “perceived consequences” of a given event [18]. Similarly, previous works have identified many factors affecting people’s flood risk perception. These factors included knowledge, previous experience, and their values and attitudes towards specific hazards [23,24]. The three main components of risk perception are awareness, worry, and preparedness [25,26]. Proper risk perception assessment is important for policymakers to understand people’s attitude towards risk mitigation. There are two well-known approaches for measuring risk perception [27]. One is the psychometric approach introduced by Fischhoff and Slovic [28,29], who attempted to quantify individual risk perception and attitudes using survey questionnaires. The other is the cultural theory approach developed by Douglas and Wildavsky [30], who considered the problem from the perspective of sociology and anthropology and put individual risk perception more in the context of culture, organization, relationship group, or lifestyle.

It is necessary to know about the relationship between risk perception and its influencing factors when studying public risk perception. There were many studies that mentioned it. For example, some scholars found that the higher the education of respondents, the lower the risk perception, because respondents with a higher education could easily get disaster-related information [31,32]. Wang et al. [8] carried out a survey in 16 sub-districts of Jingdezhen in China and found that the public risk perception was strongly related to the selected influencing factors. Ardaya et al. [33] revealed that flood experiences and demographic characteristics had the greatest impacts on risk perception. Huang et al. [11] presented a case study of flood disasters in Shenzhen city and found that the flood risk knowledge, education, and flood risk attitudes were strongly related to flood risk perception. Lechowska [34] reviewed 50 literatures on flood risk perception and compiled a preliminary list of factors that may affect flood risk perception, which were summarized into six categories, namely cognition, behavior, socioeconomic and demographics, geography, information, and cultural background. Based on the above analysis, the method of measuring risk perception and the factors influencing risk perceptions were identified and investigated.

Some scholars studied the relationship between risk perception and response behavior to disasters [35,36,37]. Although there is a centuries-old history of floods in China, there are few studies about flood risk perception, especially at the individual level. The aims of this study were to (1) construct a flood risk perception index (RPI) [27,38] to assess public flood risk perception in Jiaozuo City, (2) explore the influencing factors of public flood risk perception, and (3) discuss the suggestions for decreasing flood risks. Many factors influencing public flood risk perception were selected based on the literature [34]. Based on the literature, our study hypothesized that the higher a person’s age (H1), income level (H2), education level (H3), and trust in flood-control projects (H4), the higher the level of the RPI. Moreover, this study also hypothesized that RPI varies by sex (H5), marriage status (H6), risk response ability (H7), flood experience (H8), extent of the effects (H9), and residence characteristics (H10).

2. Data Collected and Analysis Methods

2.1. Study Area

Jiaozuo City, frequently affected by floods, was selected as the study area. It is located in the northwest of Henan Province (Figure 1). The total area of Jiaozuo City is about 4071 km² and contains 10 county-level districts. The total resident population was about 3.52 million at the end of 2020. The terrain in the northern part is higher than that in the south and the altitude ranges from 61 m to 1321 m above mean sea level, with an average elevation of 500 m. Climatically, it lies in the temperate monsoon climate region; the annual precipitation is about 500~700 mm and its annual distribution is extremely irregular (about 80% occurring from June to September). Historical statistics showed that a damaging flood occurred in Jiaozuo City in 1996, which was the worst flood in recent Jiaozuo history. It caused more than 85,000 hectares of fields in our city to be flooded, and 13 roads were washed out by floods. A torrential flood in July 2016 resulted in the productivity of Jiaozuo Wanfang Aluminum company being seriously affected. The most recent flood was in July 2021 and led to USD 310 million in damage; 744 thousand people were affected and around 11,564 houses were destroyed.

2.2. Sample Selection

Five districts (including two urban districts, namely Shanyang and Jiefang, and three county districts, namely Boai, Wuzhi, and Xiuwu) were selected as our study areas for the following reasons: (1) they all have been affected to varying degrees by the floods in July 2021, especially Xiuwu and Boai counties were severely affected; (2) Shanyang and Jiefang districts have the most concentrated population; (3) the Dasha River is the largest flood drainage river in Jiaozuo City, passing through Boai, Xiuwu, and Wuzhi districts. Considering the different characteristics of different districts, for example, (1) the proportion of population aged 0–14 is slightly higher in Xiuwu, Wuzhi, and Boai districts than in Jiaozuo City; (2) the GDP in 2021 of Boai county, Xiuwu county, and Jiefang district are at a very low ranking in Jiaozuo City. Three communities were selected in Shanyang and Jiefang districts, respectively. Questionnaires were distributed in parks, schools, and villages.

There are two representative formulas to determine sample size [39]. One is Cochran’s Formula [40] and the other is Yamane’s Formula [41]. Here, the simplified Yamane’s formula with an error acceptance value of 10% was used to calculate the sample sizes [41,42]. There were about 3.52 million residents at the end of 2020 in this region. According to Equation (1), 100 samples were enough. In fact, 234 sample residents (>100) were surveyed through random sampling in this study.

$$n=\frac{N}{1+N{e}^2}$$

(1)

where n = sample size, N = total number of residents, and e = error acceptance value.

2.3. Data Collection

The data were collected using a random questionnaire survey during the period of 2–9 November 2021. We distributed 260 copies of the questionnaire in parks, schools, and villages, and kept 234 effective copies. Before surveying, the respondents were informed of the purpose of this research, the confidentiality and privacy procedures, how the data could be used, and their rights and responsibilities. All the respondents with sufficient time were voluntary and consent. If the participants encountered any problematic questions, the researchers were responsible for explaining and the participants were given the full right to withdraw from the survey or refuse to participate. To encourage participants to better complete the questionnaire, they would receive a gift such as a pen or a canvas bag after completing the questionnaire. Permission was not required for this survey as it did not relate to participants’ privacy and rights and they knew about the purpose of this survey.

2.4. Questionnaire Design

The questionnaire was designed with the help of experts and used the relevant literature [33,43,44,45]. In order to make participants better understand the contents of the questionnaire, some experts in the field of emergency management were firstly invited to help us to improve the quality of the questionnaire before the formal survey, and their opinions were collected and fully considered. Based on experts’ opinions, the projects that were difficult for the public to understand and were ambiguous were deleted, revised, and supplemented. Firstly, “Have you ever experienced floods” was changed to “How many times have you experienced floods”; this change made the correlation between flood experience and risk perception clearer. Secondly, “whether infrastructure, housing, farmland and communication had been damaged by floods” was changed to “the extent of infrastructure, housing, farmland and communication have been damaged by floods”, so that the relationship between the severity of flooding and risk perception can be further studied. Thirdly, “Do you think the extent to which the disaster caused by the flood can be controlled” was changed to “Do you think the damage caused by the flood can be controlled”. This variable was prone to ambiguity, that is, whether the loss was controllable or the disaster was controllable, and how to control the disaster after it happened.

The purpose and the confidentiality commitment of this survey were firstly introduced to respondents in the questionnaire. The main contents of the questionnaire were divided into three parts after the revision, including 18 questions (for details see Appendix A). The first part was used to measure public flood risk perception, including 4 questions, i.e., likelihood, fear, controllability, and impact (Table 1); the second part was the demographic variables and residential characteristics of the respondents, including 10 questions, which consisted of gender, age, education level, occupation, residence, etc., (Table 2); the third part was the factors that may influence public flood risk perception, including 4 questions, such as residents’ flood experience and the trust in local flood control projects (Table 2).

2.5. Methods

2.5.1. Risk Perception Index

There are several methods that can be used to assess public flood risk perception, and the index method is a widely used method [46]. Here, we constructed a risk perception index (RPI) to assess public flood risk perception, which has been previously adopted by many scholars, such as Ullah et al. [27] and Qasim et al. [47]. RPI for each respondent was computed based on Equation (2).

$$RPI={\displaystyle \sum _{i=1}^n{S}_i}$$

(2)

where S_i = score of assessment indicators, n = the number of indicators. Here, each indicator was considered to have the same contribution to public flood risk perception [48]. The RPI was a relative estimate of each respondent and the lower the value of RPI, the lower the level of flood risk perception.

Table 1. Indicators of flood risk perception.

Indicators	Questions	Alternatives	Sources
Likelihood	How likely do you find a flood in your area within the next 10 years?	Not likely at all → Very likely 1-------------------5	[23,34,49,50,51,52]
Fear	How afraid are you of a flood?	Not afraid at all → Very afraid 1-------------------5	[27,32,34,53]
Controllability	Do you think the damage caused by the flood can be controlled?	Can totally control l → Cannot control at all 1-------------------5	[32,51,53,54]
Impact	To what extent does the flood threaten your life?	Not serious at all → Very serious 1-------------------5	[34,48,51,53]

Table 1. Indicators of flood risk perception.

Indicators	Questions	Alternatives	Sources
Likelihood	How likely do you find a flood in your area within the next 10 years?	Not likely at all → Very likely 1-------------------5	[23,34,49,50,51,52]
Fear	How afraid are you of a flood?	Not afraid at all → Very afraid 1-------------------5	[27,32,34,53]
Controllability	Do you think the damage caused by the flood can be controlled?	Can totally control l → Cannot control at all 1-------------------5	[32,51,53,54]
Impact	To what extent does the flood threaten your life?	Not serious at all → Very serious 1-------------------5	[34,48,51,53]

2.5.2. Statistical Analysis

After collating the questionnaire data, descriptive statistics were firstly performed to depict the characteristics of samples and influencing factors. Secondly, an independent sample t-test and a one-way ANOVA test were carried out to test the relationships between public flood risk perception and its influencing factors. Finally, post hoc tests were used to explore group differences in flood risk perception on each observed dependent variable separately. The entire dataset was analyzed using SPSS.

Table 2. The socio-demographic characteristics and influencing factors distribution of the respondents.

NO.	Variables	Frequency	Percentage	NO.	Variables	Frequency	Percentage	NO.	Variables	Frequency	Percentage
1	Gender			6	Occupation			11	Number of floods experienced
	Male	98	41.9		Work in government	25	10.7		0	126	53.8
2	Female	136	58.1		Student	88	37.6		1	79	33.8
	Age (years)				Farmer	30	12.8		2	19	8.1
	<18	61	26.1		Worker	16	6.8		≥3	10	3.8
	18–44	135	57.7		Work in company	59	25.2	12	Affected by the flooding
	45–60	36	15.3		Self-employed	7	3		Not at all	5	4.7
	>60	2	0.9		Unemployed	9	3.8		Low	8	7.5
3	Marital status			7	The location of residence				General	31	28.9
	Married	100	42.7		Rural	109	46.6		High	44	41.1
	Unmarried	132	56.4		Town	55	23.5		Very high	19	17.8
	Divorced	2	0.9		City	70	29.9	13	Ability to mitigate floods
4	Education level			8	Distance from rivers (meters)				Very bad	29	12.4
	Primary school and below	15	6.4		<500	50	21.4		Bad	65	27.8
	Middle school	39	16.7		500~1000	52	22.2		General	119	50.9
	High school	88	37.6		>1000	132	56.4		Good	16	6.8
	College and above	92	39.3	9	Dike protection				Very good	5	2.1
5	Income per month (dollars)				Yes	92	39.3	14	Trust in flood-control projects
	USD 0–316	109	46.6		No	63	26.9		Very distrustful	5	2.1
	USD 317–791	73	31.2		Unclear	79	33.8		Distrust	29	12.4
	USD 792–1266	39	16.7	10	Located in flood storage areas				General	104	44.5
	>USD 1266	13	5.6		Yes	36	15.4		Trust	78	33.3
					No	98	41.9		Very trustful	18	7.7
					Unclear	100	42.7

Note: NO. represents the number of the question in the questionnaire.

Table 2. The socio-demographic characteristics and influencing factors distribution of the respondents.

NO.	Variables	Frequency	Percentage	NO.	Variables	Frequency	Percentage	NO.	Variables	Frequency	Percentage
1	Gender			6	Occupation			11	Number of floods experienced
	Male	98	41.9		Work in government	25	10.7		0	126	53.8
2	Female	136	58.1		Student	88	37.6		1	79	33.8
	Age (years)				Farmer	30	12.8		2	19	8.1
	<18	61	26.1		Worker	16	6.8		≥3	10	3.8
	18–44	135	57.7		Work in company	59	25.2	12	Affected by the flooding
	45–60	36	15.3		Self-employed	7	3		Not at all	5	4.7
	>60	2	0.9		Unemployed	9	3.8		Low	8	7.5
3	Marital status			7	The location of residence				General	31	28.9
	Married	100	42.7		Rural	109	46.6		High	44	41.1
	Unmarried	132	56.4		Town	55	23.5		Very high	19	17.8
	Divorced	2	0.9		City	70	29.9	13	Ability to mitigate floods
4	Education level			8	Distance from rivers (meters)				Very bad	29	12.4
	Primary school and below	15	6.4		<500	50	21.4		Bad	65	27.8
	Middle school	39	16.7		500~1000	52	22.2		General	119	50.9
	High school	88	37.6		>1000	132	56.4		Good	16	6.8
	College and above	92	39.3	9	Dike protection				Very good	5	2.1
5	Income per month (dollars)				Yes	92	39.3	14	Trust in flood-control projects
	USD 0–316	109	46.6		No	63	26.9		Very distrustful	5	2.1
	USD 317–791	73	31.2		Unclear	79	33.8		Distrust	29	12.4
	USD 792–1266	39	16.7	10	Located in flood storage areas				General	104	44.5
	>USD 1266	13	5.6		Yes	36	15.4		Trust	78	33.3
					No	98	41.9		Very trustful	18	7.7
					Unclear	100	42.7

Note: NO. represents the number of the question in the questionnaire.

3. Results

3.1. Descriptive Statistics for Samples

3.1.1. Socio-Demographic Characteristics of Respondents

The personal characteristics of 234 respondents are presented in Table 2. Out of the 234 respondents, 41.9% were male and 58.1% were female, 42.7% were married and 57.3% were unmarried. The number of respondents aged from 18 to 44 and from 45 to 60 accounted for 57.7% and 15.3%, respectively. About the education, most respondents claimed to have a college and above education (39.3%). With regard to income, 46.6% earned less than USD 791 per month, while only 5.6% earned more than USD 1266. Seven occupation types were classified and most respondents were students (37.6%) or worked in a company (25.2%), respectively.

3.1.2. Other Important Influencing Factors

The respondents’ characteristics, risk response ability, flood experience, and the trust in flood-control projects were other important influencing factors of public flood risk perception. The percentages of respondents living in rural areas, towns, and cities were 46.6%, 23.5%, and 29.9%, respectively. There are some differences between cities and towns. Firstly, cities are residential areas formed by the agglomeration of non-agricultural industries and a non-agricultural population, while towns are dominated by a non-agricultural population. Secondly, their functions are different. Cities have the function of administrative jurisdiction, while towns have fewer functions and can basically meet the needs of life. Up to 43.6% of respondents’ residence is less than 1000 m away from the river. The percentage of the respondents protected by dikes was 39.3%. Only 15.4% respondents were located in flood storage areas. The percentages of respondents who have and who have not experienced floods accounted for 46.2% and 53.8%, respectively. Most (58.9%) respondents who have experienced flooding believed that the flood had a great impact on them. With regard to the ability to mitigate floods, respondents showed a low level. In particular, the general level accounted for 50.9%, and only 8.9% had a good ability to cope with the flood. When asked about their trust in flood-control projects, respondents with “very distrust” and “distrust” answers accounted for 14.5%, followed by “general” (44.4%), “trust” (33.3%), and “very trust” (7.7%). The sample distribution is basically reasonable, which can better represent the overall situation of residents in Jiaozuo city.

3.2. Assessment of Public Flood Risk Perception

Equation (2) was used to calculate the RPI scores. The grade of each respondent’s risk perception was confirmed based on the mean value (MV) and standard deviation (SD) of the RPI scores. If the RPI score > MV + 1 SD, the respondent was in the high grade, if the RPI score < (MV−1 SD), the respondent was in the low grade, and other RPI scores were in the moderate grade [55]. Here, the MV, SD, Min, and Max value of RPI were 2.79, 13.24, 5, and 19, respectively. Therefore, the extents of low, moderate, and high grades were (5,10.45), (10.45,16.03), and (16.03,19), respectively.

According to the results, 68.4% respondents’ flood risk perception level belonged to a moderate grade. Only 42 and 32 respondents belonged to high and low grades, which accounted for about 17.9% and 13.7%, respectively. Moreover, the mean scores of likelihood, fear, controllability, and impact were 2.89, 3.23, 3.49, and 3.62, respectively. When asked about “How likely do you find a flood in your area within the next 10 years?”, about 37% of respondents’ answers were “not likely at all” and “not likely”; no more than 3% were “very likely”. When asked about the fear of a flood disaster, 12.8% were “very afraid”, 56.7% were “afraid” or “general afraid”, and only 3.8% were “not afraid at all”. However, 58.1% believed that the damage caused by the floods could not be controlled and only two respondents thought it was completely controllable. Over half (60%) of the respondents thought the floods had a large or strong impact on their families and lives.

3.3. Influencing Factors of Public Flood Risk Perception

The ANOVA test was used to test the relationship between public flood risk perception and its influencing factors (

Table 3. ANOVA results between public flood risk perception and its influencing factors.

Variables	p Value	Variables	p Value
Gender	0.146	Distance from rivers	0.772
Age	0.000 *	Dike protection	0.047 *
Marital status	0.000 *	Located in flood storage areas	0.001 *
Education level	0.000 *	Flood experience	0.000 *
Income per month	0.049 *	Affected by the flooding	0.001 *
Occupation	0.000 *	Ability to mitigate floods	0.000 *
The location of residence	0.011 *	Trust in flood-control projects	0.000 *

Note: * denotes statistical difference (p < 0.05).

) in Jiaozuo city. The group differences in flood risk perception on each observed dependent variable were explored using post hoc tests.

3.3.1. Socio-Demographic Characteristics

The ANOVA results (Table 3) showed that there was no significant correlation between flood risk perception and gender, and other influencing factors (age, marital status, education, income per month, occupation) had statistically significant differences in flood risk perception.

Post hoc tests (

Table 4. Multiple Comparisons of factors influencing public flood risk perception.

Variable	Mean Difference	Std. Error	Sig.	95% Confidence Interval
				Lower Bound	Upper Bound
Age (vs. <18 years)
18–44	−1.940 *	0.401	0.000	−2.73	−1.15
45–60	−3.183 *	0.547	0.000	−4.26	−2.1
>60	−0.877	1.87	0.640	−4.56	2.81
Marital status (vs. unmarried)
married	−1.813 *	0.352	0.000	−2.51	−1.12
divorced	−2.053	1.893	0.279	−5.78	1.68
Education level (vs. high school)
Primary school and below	−2.977 *	0.754	0.000	−4.46	−1.49
Middle school	−1.469 *	0.519	0.005	−2.49	−0.45
College and above	−0.628	0.403	0.120	−1.42	0.17
Income per month (vs. >USD 1266)
<USD 316	−0.564	0.812	0.488	−2.16	1.04
USD 317–791	−1.172	0.833	0.161	−2.81	0.47
USD 792–1266	−1.821 *	0.886	0.041	−3.57	−0.08
Occupation (vs. student)
Work in government	−0.486	0.59	0.411	−1.65	0.68
farmer	−2.653 *	0.551	0.000	−3.74	−1.57
worker	−1.449 *	0.708	0.042	−2.84	−0.05
Work in company	−1.836 *	0.438	0.000	−2.7	−0.97
Self-employed person	−1.029	1.023	0.315	−3.04	0.99
Unemploymed	−3.664 *	0.911	0.000	−5.46	−1.87
(vs. town)
Rural	−1.370 *	0.455	0.003	−2.27	−0.47
City	−0.732	0.496	0.141	−1.71	0.24
Dike protection (vs. no)
Yes	−0.373	0.453	0.412	−1.27	0.52
Unclear	−1.121 *	0.468	0.017	−2.04	−0.2
Located in flood storage areas (vs. no)
Yes	−1.854 *	0.53	0.001	−2.9	−0.81
Unclear	−1.110 *	0.387	0.004	−1.87	−0.35
Flood experience (vs. 0 times)
1 time	−1.858 *	0.379	0.000	−2.61	−1.11
2 times	−2.064 *	0.651	0.002	−3.35	−0.78
≥3 times	−2.043 *	0.868	0.020	−3.75	−0.33
Affected by the flooding (vs. very low)
Low	−3.055 *	1.236	0.014	−5.49	−0.62
General	−1.875	1.546	0.226	−4.92	1.17
High	−3.419 *	1.307	0.009	−5.99	−0.84
Very high	−3.432 *	1.28	0.008	−5.95	−0.91
Ability to mitigate floods (vs. high)
Very low	−3.369 *	0.838	0.000	−5.02	−1.72
Low	−2.130 *	0.751	0.005	−3.61	−0.65
General	−1.269	0.716	0.078	−2.68	0.14
Very high	−1.037	1.378	0.452	−3.75	1.68
Trust in flood-control projects (vs. very trustful)
Very distrustful	−3.856 *	1.133	0.001	−6.09	−1.62
Distrust	−5.814*	0.672	0.000	−7.14	−4.49
General	−2.777 *	0.572	0.000	−3.90	−1.65
Trust	−0.774	0.586	0.188	−1.93	0.38

Note: * with statistical difference (p < 0.05).

) identified that the respondents aged 45–60 had the highest flood risk perception compared with others. The reasons may be that the respondents aged 45–60 had more flood experiences and thought more about the family safety, thus they were significantly higher in the perception of likelihood, impact, and fear than other groups. For marital status, the married people had significantly higher risk perception than unmarried people. The formation of this characteristic, on the one hand, was related to the family responsibilities and pressure of married people. On the other hand, it was determined by age and, in general, a married person was older than an unmarried person. It was also found that the higher the education level, the lower the flood risk perceptions, possibly because people with a high education had more abilities and more confidence to control the damage caused by flood disasters [32].

For respondents’ monthly income, respondents with a higher income had lower flood risk perceptions. The result arises in part for three reasons. Firstly, there may be a positive relationship between income and education. Secondly, high-income respondents thought that they had the ability to deal with the loss caused by flood disasters. Furthermore, there was a classical bias found in seismic risk perception based on the System Justification theory, namely individuals with a high income tended to believe that the world as they perceive it was just, legitimate, and beneficial [56]; thus, they were least likely to see earthquakes as significant risks [57]. Regarding occupation, there were significant differences in flood risk perception in other occupations apart from students, the self-employed, and public officials. The unemployed respondents had the highest perception level.

3.3.2. Residence Characteristics

The ANOVA results (Table 3) showed that there was no significant correlation between flood risk perception and the distance from rivers, but that there were significant differences in three variables: the location of the residence, whether there is a dam or reservoir near the residence, and whether the residence belongs to the flood storage area.

Post hoc tests (Table 4) identified that the rural respondents had the highest risk perception than others; this was because rural residents believed they did not receive much external support and attention, which lead them to feel a high degree of possibility and impact over a hazard.

On the question of whether there is a dam or reservoir near the residence, the respondents who answered “no” had the lowest level of risk perception, which was opposite to the actual flood exposure, indicating that the residents who were not protected by the dam are to a certain extent poorly informed about the risks they faced. On the question of whether the resident lives in a flood storage area, the respondents who answered “yes” had the highest flood risk perception, because they had a higher probability perception of flood hazards than others. Moreover, once it happens, the impact is huge and is almost impossible to control.

3.3.3. Experience of Floods

The ANOVA results (Table 3) showed that respondents with a different number of flood experiences and different flood damage degree had significant differences in flood risk perception. Post hoc tests (Table 4) identified that the respondents who experienced flooding more than three times and were most severely affected by the floods had the highest risk perception level. This may be because the people with more disaster experiences have a clearer and more systematic understanding of the hazards and are also more easily aware of the hazards and the severity of the hazards.

3.3.4. Ability to Mitigate Floods

The ANOVA results (Table 3) showed that there was a significant difference in public flood risk perception in the ability to mitigate floods. Post hoc tests (Table 4) identified that the respondents with a “very low” ability to mitigate floods had a higher risk perception than others. This was mainly because the confidence in their ability to respond to hazards had weakened their fear of flood disasters.

3.3.5. Trust in Flood-Control Projects

The ANOVA results (Table 3) showed that there was a significant difference in public flood risk perception in the trust of public flood protection. Post hoc tests (Table 4) identified that the respondents with “very distrust” in the flood-control projects had a higher risk perception than others. This was mainly because the trust in flood protection projects not only reduced the public’s judgment on the possibility of flooding but enhanced the public’s confidence in hazard response and recovery.

4. Discussion and Conclusions

4.1. Discussion

Our study found that there was no significant correlation between gender and flood risk perception, which was consistent with previous studies [58,59]. Conversely, several authors found that gender had a strong impact on risk attitudes. Generally, the flood risk perception level of males was lower than that of females [8,26,52,60]. The reason for this conclusion is that women have less resources and socioeconomic status and are more vulnerable and more concerned about property damage in the face of floods [42,60,61,62]. Different age groups had different flood risk perception degrees. The result was consistent with Grothmann and other scholars [63]. Furthermore, we also found that respondents with a higher income and education showed a lower flood risk perception. This result was consistent with some previous studies. For example, Kellens et al., reviewed the empirical research on flood risk perception [23,47,64,65] and revealed that income and education had negative impacts on flood risk perceptions [24,32,34]. For occupation, studies showed that occupation partly determined the perception of flood risks. This result was supported by previous studies [66,67,68]. They argued that socio-economic vulnerable groups, that is, those who were objectively more likely to face the adverse consequences of hazards, were more sensitive to risk.

Resident’s residential location, which reflected the physical exposure, was often used to appraise the public risk perception to natural hazards. Some of the literature about risk revealed that residents who lived far away from rivers had a lower flood risk perception [24,27,42]. Ullah et al. [27] found that respondents who lived near rivers suffered a higher physical exposure to floods than others. The results of this study revealed that respondents living in flood storage areas had a lower risk perception compared with the reference group, similar to the study by Botzen et al. [24]. Respondents who lived in cities had a lower risk perception than those who lived in rural areas; the result was consistent with some previous studies [24,32,34]. Many studies revealed the impact of flood experience on risk perception [19,20,69,70,71], consistent with the results obtained in this paper; namely, respondents with direct flood experience had a higher risk perception.

The ability to mitigate floods had a negative impact on risk perception, which was similar to the views of many scholars [19,70]. Consistent with previous studies, the trust in flood-control projects had a negative impact on risk perception [23,24,63,72,73,74].

The limitations of this research were as follows. Firstly, the number of samples was limited to a relatively small area due to the impacts of COVID-19, which may influence the adaptability and accuracy of the results. Secondly, a few respondents were still cautious in answering the questions, which decreased the accuracy of the results. Consequently, they avoided choosing extreme answers to sensitive questions, although they were told the survey was anonymous. Thirdly, more than half of the respondents in our survey did not experience a flood. This means that their answers were not based on their experience. Fourthly, some questions in the questionnaire had certain limitations. For instance, regarding question 11, people had different understandings of floods, whereby some respondents may perceive a small flood as a major event, while others would not. Thus, their answers were somewhat subjective and uncertain. For question 12, it was not clear to which flood event this question refers. In addition, it was not recommended to refer to all these elements in the same questions. For question 13, it was a very tricky and subjective question; we did not know whether the respondents really knew very much about flood response measures.

There are some issues to be further analyzed. Firstly, how perceived risk influences public flood response behavior. Secondly, the systems of trust, such as public trust in government, media, experts, and markets, should be taken into account in future research of risk perception. Moreover, interest-related factors may also affect individual risk perception.

4.2. Conclusions

Based on questionnaire surveys and statistical analysis, the public flood risk perception in Jiaozuo City was analyzed and its affecting factors were selected. There are some interesting findings, as follows: (1) 68.4% respondents’ flood risk perception level belonged to a moderate grade; only 17.9% and 13.7% of respondents belonged to high and low grades. (2) Respondents with less education, lower annual income, more flood experience, aged 45–60, and who lived in rural areas, areas with dike protection, and located in flood storage areas had a higher flood risk perception than others. (3) Marital status, occupational, self-response ability to flooding, and trust with public flood protection also affected residents’ flood risk perception to varying degrees. Hence, risk perception can be considered as a complex process including cognitive and affective features. In view of these results, studies on flood risk perception factors should not only be affected by statistical data but also by people’s social and cultural characteristics. Therefore, the study of risk perception should adopt a more comprehensive method, including background factors such as cultural, religious, historical, and political backgrounds.

The finding of this study is useful for flood management. Firstly, it can help policymakers take reasonable actions to reduce casualties. For example, the results revealed that residents with previous flood experience had a higher flood risk perception than others. Therefore, the communities should cooperate with local governments to regularly carry out evacuation exercises in order to improve residents’ evacuation capabilities during a flood. In addition, according to the survey results, residents with a high education level or a high income generally believed that the flood disaster was highly controllable, yet, in many cases, flood control facilities and government intervention may not work timely or effectively. Hence, it was important to improve their risk awareness through training and education. Secondly, people with a higher risk perception can be identified based on the results, which can help governments take specific measures, such as providing flood insurance to reduce flood losses. Thirdly, flood-control projects had an irreplaceable role, which was the important safeguard to reduce the exposure of the flood-prone areas. From the management perspective, governments can guide the public to protect these projects and recognize their limitations and effectiveness.