Knowledge of Energy Resources and Next Generation Energy Choice Behaviour: A Case Study of Kazakhstan

Abstract

Renewable energy (RE) is key to averting the climate crisis, and public support is central to its successful implementation. In this study, we examined the impact that knowledge of energy policy and energy issues has on public support for different energy types. This was achieved through the use of an online survey of residents of Kazakhstan. As a resource-rich developing country with a history of nuclear testing, Kazakhstan provides an interesting case study. In this paper, we statistically analyze the effect of individual knowledge of energy policy on the choice behavior for next-generation energy, including RE, in Kazakhstan. First, those who support fossil fuel power generation knew that Kazakhstan has abundant natural gas reserves, and those who support wind and solar power generation were aware that Kazakhstan has huge potential for such forms of power generation. It was clear that knowledge of the positive and negative aspects of fossil fuel, nuclear, wind and solar power generation had an impact on the preferences of the respondents. Second, the reasons given for supporting RE included: unlike fossil fuels, there was no danger of them being exhausted; the rise in adoption of RE technology globally; reduction in carbon dioxide emissions; and, addressing climate change. Third, although many women were aware of the advantages and disadvantages of RE, there were high levels of both support and opposition. Respondents with lower income and lower educational attainment tended to oppose RE. People living in East Kazakhstan, the site of nuclear tests, tended to support RE, but due to the high rates of subsidy to the price of electricity, many people preferred the status quo. Finally, preference for next-generation energy in Kazakhstan was associated with higher awareness of the need to protect the global environment.

1. Introduction

Kazakhstan is rich in natural fossil fuel and mineral resources, with nine of its deposits placing in the top ten most plentiful in the world. These resources and their domestically processed products account for 70% of Kazakhstan’s exports. However, desertification, air pollution, soil pollution and water pollution represent serious problems, with almost 4% already suffering desertification and a further 76% categorized as susceptible to desertification [1], and in part a direct consequence of mining and related activities. There is little pollution regulation for these heavy industries, and little enforcement. In addition, many studies have looked at the health and social impact of the Soviet era Semipalatinsk nuclear test site on the water supply around the Aral Sea. Uranium extraction has also led to serious pollution and environmental consequences for Kazakhstan. On the other hand, Kazakhstan has huge renewable energy (RE) potential [2,3,4,5].

The term renewable energy (RE) is used here to include energy sources that are constantly replenished by natural forces such as solar, hydro, and wind power, while fossil fuels such as oil, coal, and natural gas are exhaustible (i.e., non-renewable) resources. Despite its huge potential, only 1 to 2% of Kazakhstan’s energy is generated by RE. As of 2020, there were 130 mostly wind and solar facilities contributing 2 gigawatts of electricity, and large hydropower generators contributing 1.6 gigawatts [6]. Kazakhstan uses non-renewable resources such as oil, coal, and natural gas for thermal power generation to provide for the overwhelming majority of its electricity needs. Nuclear power generation using nuclear fuel is also an exhaustible resource [7]. There are different types of coolants used in reactors including: light water reactors (LWRs) that use ordinary water as a coolant (and account for approximately 80% of the 436 reactors around the world [8,9]), heavy water reactors (HWRs) that use heavy water as a coolant, gas-cooled reactors (GCRs) that use various gases as coolants, and liquid metal cooled reactors (LMRs). The Soviet-era fast-breeder reactor BN-350 that operated from 1973 to 1999, was an LMR, using liquid sodium as the coolant. A total of four research reactors continue to operate in the Semipalatinsk test site and another in the Institute of Nuclear Physics near Almaty. These are used to research reactor safety, materials science, and future reactor technology, often in partnership with other national agencies [6]. There have been several attempts to build commercial nuclear reactors with Russian, American, French, Chinese, and Japanese partners, but have thus far not come to fruition [6].

A study by Dahl and Kuralbayeva [10] argues that technical and legal improvements would be necessary to overcome the environmental and ecological damage caused by the extraction of its mineral wealth. Fyodorov and Kayukov [11] uranium mining had resulted in the accumulation of 240 million tons of radioactive waste. Tomita et al. [12] analysed more than 100 samples of river, ground, spring, and lake water in northeastern Kazakhstan. was lower than that of the Irtysh River, but that groundwater and lake water had uranium levels several times higher than seawater, and the ${}_{234}$U and ${}_{238}$U activity ratio was also relatively high (1.8), indicating ongoing contamination. Kassenova [13] points out that in addition to the health and environmental implications of being a major producer of uranium, there are significant security issues involved, such as nuclear terrorism and instability in the event of anti-nuclear protests or other political instability. RE accounted for only 1.4% of domestic energy consumption in 2018, with about 50% coming from coal and the remainder from oil and natural gas [14]. Domestic household electricity prices are very low, averaging USD 0.041 per kWh in 2021, 21st lowest of 147 countries [15]—although not as low as two of its immediate neighbours (Kyrgyzstan and Uzbekistan). These low prices, along with its abundant fossil fuel reserves are regularly cited as major factors in the low adoption of RE. Kazakhstan’s RE potential is significant, with a conservative 1996 estimate projecting a technically possible 1 trillion kWh of solar energy and 1.8 trillion of wind energy [2].

It is worth noting that although Kazakhstan is a relatively equal society, inequality has been growing and there is a marked east–west, north–south divide within the country, where western and northern regions tend to be more prosperous [16]. East Kazakhstan, home to Semipalatinsk, has an above average standard of living but has been ranked least technologically ready for the information and communications technology sector.

Kazakhstan occupies a key position in the “One Belt and Road Initiative” and the “Silk Road Economic Belt” (One Belt) proposed by General Secretary Xi Jinping. President Nazarbayev also recognizes the “Nurly Zhol” initiative as part of the One Belt, linking it to The Belt and Road Initiative and strengthening ties with China [17]. Affordable and Clean Energy is the seventh of the SDGs adopted by the UN General Assembly [18]. The goal is to ensure that countries around the world have access to sustainable energy sources. Kazakhstan uses solar panels exported from China. In addition, one of the largest wind power plants in Central Asia, the Janatas Wind Power Plant, is equipped with Chinese-made generating units [19]. A high percentage of RE facilities in Kazakhstan are made in China. Many countries are encouraging domestic production of RE equipment for reasons such as national security. Under these circumstances, the question facing policy makers is whether Kazakhstan should continue partnering with China, or whether it should seek a more independent route, possibly by decarbonizing by using its own natural gas and uranium.

Aims of the Study

Most national energy policies implicitly or explicitly follow the same approach as the Japanese government of “S + 3E” [20,21]. This is “Energy Security”, “Economic Efficiency”, and “Environment”, with the underlying premise of “Safety”. Nuclear power generation has a safety problem with regard to the disposal of nuclear fuel and a disadvantage with regard to the environment due to radioactive materials and contaminated water. On the other hand, nuclear power generation has the advantage of economic efficiency in that it emits less CO${}_2$ and uses a small amount of fuel to produce a large amount of energy. In recent years, power sources are sometimes evaluated by adding “Macro impact” to S + 3E, i.e., S + 3E + M. Since no power source can meet all of the S + 3E criteria, nuclear power has the advantage of being a base power source that can supply a stable and constant amount of power throughout the year.

In response to the environmental damage caused by the mining and use of its mineral resources, Kazakhstan decided to embrace renewable energy. In 2012, President Nazarbayev included a target of at least 50% renewable energy in the Kazakhstan 2050 Strategy of national renewal [22]. Karatayev and Clark [23] suggest that, if properly implemented, switching from fossil to renewable energy will be beneficial to the environment, economy and society in general. However, Karatayev et al. [3] also point out the considerable regulatory and structural barriers to successfully introducing renewable energy, and the negative consequences this can have. They recommend a switch to solar and wind energy to reduce greenhouse gas emissions. However, little to no research has been done on the preferences of the Kazakh people, or their understanding of the alternatives and issues involved. For a policy to be effective, it is essential that the viewpoint of the public - the main end users, the beneficiaries and potential victims - be fully understood and recognised and that there be a degree of public acceptance [24]. To date, all the previous studies have not addressed this issue from the public’s viewpoint, the level of support amongst the public RE has, nor even the level of public awareness and knowledge of RE, or energy policy in general.

In this paper, we will statistically analyze and consider the role the level of knowledge and understanding has on the energy preferences held by the Kazakh people. We will also examine whether the next-generation energy choices will lead to support for anti-globalization or regional integration. Specifically, the following three hypotheses are posited:

Hypothesis 1.

As has already been described, Kazakhstan is rich in mineral and energy resources. Taking nuclear power as an example, given awareness of this abundance, would a person support use of nuclear power as an energy source? The first hypothesis is Awareness of the reserves informs the preference to support or oppose nuclear power.

Hypothesis 2.

Kazakhstan’s RE potential is very promising. Taking solar power as an example, would awareness that Kazakhstan’s long sunny days are potentially ideal for solar power generation, lead to greater support for the expansion of solar power? The second hypothesis is Awareness of the potential to generate renewable energy informs support for RE.

Hypothesis 3.

We examine the specific case of the Semipalatinsk nuclear test site in East Kazakhstan, which continues to be affected by radioactive fallout [25]. Arguably, knowledge of this would lead to support for renewable energy as a safer and cleaner alternative. The third hypothesis we look at is Knowledge of nuclear contamination has a direct influence on the degree of support for renewable energy amongst East Kazakhstan residents.

2. Materials and Methods

2.1. Data Collection

The survey was conducted online from the 20th to 21st September (Friday–Saturday) 2019, Japan time, through the online survey company SurveyMonkey, distributed by their local affiliate. The survey language was in Russian, with participants from all over the country. A total of 304 participants responded fully to the survey (86.6% completion rate).

To determine a representative sample size the following formula was used:

$$n=\frac{z^2·\widehat{p}(1-\widehat{p})}{{\epsilon }^2}$$

where n is the sample size; z is calculated from the confidence interval (1.96 for an interval of 0.95, 2.05 for an interval of 0.99; this research has opted for a confidence interval of 0.95); $\widehat{p}$ is the population proportion estimated to present the target response, set to 0.5 when unknown; $\epsilon$ is the margin of error, set to 0.05 to account for the largest possible error and produce the largest value of n. This or a similar formula producing very similar results is used across the literature and in online survey size calculators (for example www.calculator.net/sample-size-calculator.html (accessed on 19 June 2023)). Given these parameters, a minimal sample size of 385 is required. In addition, Long [26] states that in multiple regression analysis, the model is stable even with a relatively small number of samples, but in logistic regression analysis, at least 200 samples are necessary because the maximum likelihood method is used, and an additional 20 samples should be added for each explanatory variable. As this research uses the backward stepwise selection method with the maximum number of explanatory variables set at five, yielding a minimum sample size of 300 (200 + 5 × 20), the survey sample size satisfies this minimum requirement. It is also important to discuss whether the power of the ordinal logit model can be measured.

Through the use of the Backward Stepwise Selection method, we have tried to estimate with a small number of explanatory variables and maintain the principle of ‘Parsimony’. Usually in regression analysis, the higher the number of explanatory variables, the higher the contribution rate. However, the maximum likelihood method minimizes the number of explanatory variables and introduces ‘AIC minimization’, which strikes a balance between the complexity of the model and the goodness of fit with the data, but this can result in a small pseudo-$R^2$ (McFadden’s pseudo-$R^2$ in this paper). In addition, since the number of samples is close to the minimum, the possibility of in-homogeneous variance cannot be ruled out. In general, when estimating ordinal logit models, we test for heterogeneity of residuals and measure robust standard errors if they are estimated to be heterogeneous.

When the survey company distributes the Web-based survey, it is presented as a “survey on next-generation energy in Kazakhstan”, which means that residents who are interested in next-generation energy such as RE are more likely to respond to the survey. Therefore, there is a possibility of self-selection bias caused by the inclusion of the intention to support next-generation energies such as REs. This self-selection bias may lead to a higher probability of choosing RE as the next generation energy source than the average citizen. Self-selection bias is a particular problem for web-based surveys and is a limitation of such surveys. In an attempt to overcome this limitation, the survey items were designed to allow respondents to select the use of non-renewable fuels such as natural gas and uranium as the next generation energy sources in Kazakhstan.

In this study, the analysis is conducted mainly using Stata, but when Backward Stepwise Selection is selected using this platform, robust standard errors cannot be calculated. The model in this paper is constrained in two respects: (1) it assumes a significance level of 10% and thus has a “significant trend”, and (2) it does not use robust standard errors to correct for heteroskedasticity [27]. Power represents the ability of a hypothesis test, and once the target power is set, it is possible to calculate a priori how large the sample size should be [27]. In recent years, power has been emphasized as an indicator of the performance of a test. G*power [28] is a world-renowned software capable of calculating power. However, G*power cannot be used to estimate the power of ordinal logit models. Therefore, in this paper, we followed the theory of Long [26] to negotiate the sample size of the ordinal logit model.

The survey was composed of 27 items, mostly with Likert-type five-scale options to choose from. Support for a statement was presented as ‘strongly oppose’ = 1, ‘somewhat oppose’ = 2, ‘neither support or oppose’ = 3, ‘somewhat support’ = 4, and ‘strongly support’ = 5. Other questions allowed the participants to choose multiple responses from a list of options, such as choosing reasons for supporting RE.

It should be noted that although every effort was made to obtain a representative sample, SurveyMonkey are unable to accurately target the survey by region, and are limited by the reach of their local affiliate. There are therefore a disproportionate number of respondents from Almaty, which has a larger population, and from people in their 20s to 40s. There are less middle-aged and older people, so a degree of regional and age bias can be assumed. This paper intended to address the question of whether residents of East Kazakhstan differ in their support for RE when compared to other respondents, given their experience or knowledge of nuclear tests, but given that most responses are from the more populous provinces, it proved difficult to satisfactorily answer this.

2.2. Analysis Methodology and Variables

The dependent variables modelled in this study are support for different types of energy:

• Overall support for expansion of RE
• Support for expansion of solar energy
• Support for expansion of wind energy
• Support for expansion of hydroelectric energy
• Support for expansion of fossil-fuel-based energy
• Support for expansion of nuclear energy

According to EIA (US Energy Information Administration) statistics, Kazakhstan’s power generation ratio (2021) is 87.92% for fossil fuels, 9.19% for hydropower, 2.90% for renewable energy, and 0.0% for nuclear power. Of renewable energy, solar power accounts for 1.74%, wind power 1.12%, and biomass/waste for 0.04%. According to the latest EI (Energy Institute) statistics (2022), fossil fuels account for 88.21% of Kazakhstan’s energy supply, of which 67.41% is coal-fired and 20.75% is natural gas-fired, and 0.05% from petroleum.

We used sequential logit regression to model support for RE and the other dependent variables. The analysis was conducted using the STATA (version 17) statistical analysis software. Dependent variable response categories were combined when the differences between values were not statistically significant or when the number of respondents was too small. In these cases, the calculation gives only the optimal estimated result, considering the values of AIC and likelihood ratio. For each model, the backward selection method was used to eliminate variables with a significance above 20%, leaving only the variables that were significant at the 1–10% level. In the following tables, cut indicates a threshold variable that corresponds to:

$$\begin{array}{cc}\hfill Pr(y=1)& =Pr(\beta x<cut1)\hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill Pr(y=2)& =Pr(cut1<\beta x<cut2)\hfill \end{array}$$

(2)

where y is the category of the dependent variable, x is the explanatory variable, and $\beta$ is the parameter.

Independent variables include “knowledge of energy sources” and “knowledge of energy development issues”. A total of seven demographic variables were also used in the analysis (

Table 1. Demographic Same below. characteristics of survey participants (n = 304).

Variable		n	%
Gender	Male	113	37.2
	Female	191	62.8
Age	under 29	78	25.7
	30–39	115	37.8
	40–49	82	20.0
	over 50	29	9.6
	Average/SD	37.4	10.6
Children under 12	Present	191	63.8
	Not present	131	43.1
	Average/SD	3.7	1.3
Occupation	Office worker	101	33.2
	Engineer/specialist	49	16.1
	Self-employed	40	13.2
	Public employee	27	8.9
	Homemaker	24	7.9
	Other	63	20.7
Education	∼Senior high	25	8.3
	College/trade school	51	16.9
	University	182	60.5
	Graduate	46	15.3
Region	Almaty	110	36.2
	Akmola	41	13.5
	East Kazakhstan	24	7.9
	Karaganda	24	7.9
	Other	105	34.5
Income (KZT) ${}^1$	under 70,000	62	20.4
	70,000–90,000	40	13.1
	90,000–120,000	57	18.8
	120,001–160,000	49	16.1
	over 1,600,000	96	31.6
	Average/SD	133,926	80,960

¹ Income brackets have been combined for clarity.

). Three of these are presented as dichotomous:

• Gender (male = 1, female = 0)
• Region (East Kazakhstan = 1, non-East Kazakhstan state = 0)
• Presence of children under 12 years old in the household (yes = 1, no = 0)

The remaining four demographic independent variables are continuous:

• Age
• Number of household members
• Education
• Income (net)

For age and income, respondents were given brackets to choose from. Education was scored from 1 for high school up to 4 for graduate school.

Ogui et al. [29] conducted an epidemiological survey of residents around the Semipalatinsk nuclear test site, and found that the occurrence of both neoplastic and cardiovascular diseases varied greatly by gender, age and ethnicity (they found a tendency for cardiovascular diseases amongst high-dose men, and cardiovascular disease and ischemic heart disease amongst high-dose women), suggesting these are important factors to consider. As it was expected many participants would be victims of fallout from the Semipalatinsk test site, we included an additional dichotomous variable:

• Exposure to fallout from Semipalatinsk test site (yes = 1, no = 0).

In addition to these independent variables, selected variables were applied to different models. For example, in modeling support for RE, participants were asked to make multiple selections from lists of potential arguments for and against RE. These are treated as binary variables.

Dependent variable categories were combined when the differences between values were not statistically significant or when the number of respondents was too small. In these cases, we made our selection based upon optimal AIC values and likelihood ratios. The backward selection method was used to eliminate variables with a significance level above 20%, leaving only the variables that were significant at the 1–10% level.

3. Results

3.1. Descriptive Results

3.1.1. Demographic Details of Participants

Table 1 summarises the demographic details of the survey participants. The first point to note is the disproportionate number of women participants (62.8%). Almaty, as the most populous region, accounts for 36.2%. Akmola, where the capital is located, accounts for 13.5%. East Kazakhstan, where the nuclear test site is located, accounted for 7.9%. The average age was 37.4, and the largest age group was 30–39 (37.8%). This is somewhat older than the national average of 30. The majority of respondents completed university (59.9%), or college or professional school (16.8%), and a further 15.1% completed graduate school. This is slightly more, but comparable to the national average. The average monthly income of the participants was KZT 133,926, about USD 356.70 (at the time of the study, KZT 100 was approximately equal to USD 0.266). The national average income for September 2019 was USD 487, so our participants had a somewhat lower income than average. According to the United Nations (2009–2017), the Gini coefficient for Kazakhstan was (27.5), the ninth lowest out of the 152 countries with available data, meaning Kazakhstan had one of the lowest levels of income inequality in the world. However, given the incomes reported in this survey, there appears to be a greater spread among our participants than the UN data suggests for the country as a whole.

3.1.2. Knowledge of Energy Reserves and Related Issues

As Figure 1 illustrates, there was a general awareness of Kazakhstan’s fossil fuel reserves and potential capacity for renewable energy, although a large proportion were unaware of the cheap domestic energy prices (44.7%). The most recognised energy source was natural gas, a major part of the energy market, with 84.9% of respondents aware of it. Awareness of uranium reserves was also very high (74.3%), not surprising given the role uranium has played in the country’s economy and history. Wind energy potential had 70.7% awareness, and solar energy potential 60.5%.

Participants showed a high level of awareness of the dangers to the health and environment different types of energy generation can cause (Figure 2). In particular, there was very high awareness of the dangers of nuclear testing (88.2%) and the impact on health and the environment of coal-powered generators (80.3%).

There are 15 hydroelectric power plants producing 50 MW or over dating back to the Soviet era, mostly in the south and east of the country. Hydroelectricity is the main source of RE in Kazakhstan, making up 12.3% of total electricity capacity [30]. Expanding this capacity will have an impact on the surrounding natural environment, including damage from construction, and 56.9% of participants reported awareness of these concerns.

Transmission loss is a serious problem, with an average of 21% loss and up to 50% in rural areas, due to the distance between the electricity generators and consumers, and the age of the transmission lines [31]. Only 41.1% of respondents indicated an awareness of this issue.

As illustrated in Figure 3, there is widespread support for the expansion of solar and wind energy (97.0% and 95.7%), but there is equally widespread support for further expansion of fossil fuels (89.9%). However, there is strong opposition to the expansion of nuclear power (with 30.92% strongly disagreeing with it).

Respondents were asked why they supported the expansion of RE (

Table 2. Reasons for supporting RE expansion (multiple responses possible).

Item	n	%
RE contributes to reduced climate change	152	50.0
Prices will fall as capacity increases	123	40.5
Unlike fossil fuels, energy source is infinite	117	38.5
An appealing application of science and technology	101	33.2
Ease of installation on homes, factories, etc.	98	32.2
It contributes to local economies	86	28.3
Zero or low pollution	81	26.6
It is durable	70	23.0
It is becoming the international norm	58	19.1
It is efficient at a small scale	55	18.1
The feed-in tariff system is beneficial	34	11.2

). The RE contribution to reducing climate change was given by 50% of respondents, followed by the belief that prices will fall as capacity increases (40.5%). The benefits of the feed-in tariff were only cited by 11.2% of participants.

Respondents also gave their preferred reasons for opposing RE expansion (

Table 3. Reasons for opposing RE expansion (multiple responses possible).

Item	n	Percentage
Too dependent on the weather, seasons, and time of day	89	29.3
Construction of large numbers of low output units will be expensive	77	25.3
Unstable supply	75	24.7
It does not fit in with the surrounding environment	72	23.7
It can be an eyesore	68	22.4
Battery technology is not yet sufficiently advanced	60	19.7
It is impractical	54	17.8
It can endanger the local ecosystem	52	17.1
The power transmission network is insufficient	44	14.5
The output is too low	37	12.2
Electricity bills will increase	36	11.8
Satisfied with current bill	31	10.2

). The main reasons given were the dependence on weather and seasonal conditions (29.3%), the cost of producing a large number of low-output units (25.3%), and related to the first reason given—an unstable supply of electricity (24.7%).

Support for RE falls into three broad categories: reducing climate change, positive economic impact, and a move to a more localised and clean energy source. Opposition falls into four broad categories: unreliability, unreliable technology, negative environmental impact, and negative economic impact.

3.2. Analysis

In this section, we discuss the results of our efforts to model the dependent variables.

Table 4. Logit regression analysis for energy preferences (n = 304).

Item	Fossil Fuels Coefficient (SE)	Nuclear Coefficient (SE)	Hydroelectric Coefficient (SE)	Wind Coefficient (SE)	Solar Coefficient (SE)
cheap electricity			0.128 (0.087)
knowledge of natural gas reserves	0.210 * (0.127)
wind power potential				0.510 *** (0.136)
solar power potential					0.706 *** (0.147)
dangers of coal	0.168 (0.119)
health & nuclear testing		−0.166 (0.105)
transmission loss		0.312 *** (0.090)			−0.154 (0.146)
male		0.824 *** (0.235)	0.317 (0.238)
age		0.030 *** (0.090)	0.018 * 0.011
East Kazakhstan			0.666 (0.427)
Household size				−0.240 * (0.142)
children under 12				0.961 ** (0.397)
education					0.368 * (0.207)
cut1		−1.332 *** (0.455)
cut2	0.759 (0.522)	−1.964 *** (0.433)	0.203 (−1.162)	1.911 *** (0.614)	0.368 (0.941)
cut3	−0.765 (0.516)	−3.138 *** (0.462)	−1.162 * (0.487)	0.614 (0.614)	−1.326 (0.924)
Likelihood ratio	500.5 *	740.0 ***	611.2 *	266.0 ***	248 ***
AIC	508.5	754.0	623.2	276.0	258.0
${\chi }^2$	8.4	47.5	10.6	21.2	17.1
McFadden $R^2$	0.017	0.060	0.017	0.074	0.099

***, **, * indicate significance at 1%, 5% and 10% respectively. cut represents responses: for fossil fuels, hydroelectric, wind and solar cut1 represents ‘somewhat aware’ and cut2 represents ‘aware’. For nuclear power, cut1 represents ‘somewhat oppose’ and ‘oppose’; cut2 represents ‘neither’ and cut3 represents ‘somewhat support’ and ‘support’. Models include knowledge of energy reserves and potential (Figure 1), awareness of energy issues (Figure 2), and demographic variables (Table 1). The backward stepwise selection method was applied to eliminate variables with a significance above 20%, and only those with significance of 1 to 10% are given.

gives the results of the sequential logit modelling for each of the dependent variables. Some of the responses were combined in order to produce statistically significant results. Thus, for example, the responses ‘oppose’ and ‘somewhat oppose’ for nuclear power were combined and are represented by cut1.

Knowledge of natural gas reserves had a positive impact on support for fossil fuels (0.210). For nuclear power, positive variables included male (0.824) awareness of transmission loss (0.312), and age (0.030). Age also had a positive relationship with support for hydroelectric power (0.018). Larger household size had a negative relationship with support for wind power (−0.240), whereas those with young children tended to support it (0.961), as did awareness of its potential (0.510). Solar power had positive associations with education (0.368) and awareness of its potential (0.706).

The initial interpretation of these models suggest that the null of the first two hypotheses (awareness of energy potential informs support for that energy source) can be rejected.

McFadden Pseudo-$R^2$ is a measure of the goodness of fit of a model in logistic regression analysis and is interpreted as equivalent to the coefficient of determination ($R^2$) in linear regression analysis (OLS). However, unlike the $R^2$ of OLS, the pseudo-$R^2$ based on log-likelihood is not the ratio of explained variance to total variance, but rather represents the degree of improvement in the likelihood of the estimated model relative to the model including only the intercept. Thus, in general, a pseudo-$R^2$ closer to 1 is also a statistically better model, but it cannot be evaluated using the same criteria as the OLS $R^2$. McFadden pseudo-$R^2$ in Table 4 is particularly low (0.017 to 0.099). On the other hand, the results of the likelihood ratio test in these models reject the null hypothesis that the regression coefficients of all explanatory variables are zero. The statistical significance of the individual regression coefficients, which is necessary to understand the structure, is also confirmed for several variables. Based on the above, we will proceed with the interpretation of the results, assuming that the validity of the estimation model is assured. The null hypothesis tested by the likelihood ratio test is rejected in each of the models in the tables. The dependent variables were ordered into five response categories, but because the differences between categories were not statistically significant, the categories were combined to improve the responsiveness of the models. The response categories for the dependent variables were reduced to cut1 (‘disagree’ and ‘somewhat disagree’), cut2 (‘neither’), and cut3 (‘somewhat agree’ and ‘agree’). The backward stepwise selection method was used to produce the model with the best fit using the AIC (Akaike Information Criterion), reducing the number of variables while retaining as much explanatory information as possible. The marginal effects have been excluded due to space limitations.

The initial interpretation of these models suggests that the null of the first two hypotheses (awareness of energy potential informs support for that energy source) can be rejected.

Marginal Effects of RE Model

Table 5. Marginal effects of RE model.

Item	Support for RE Coefficient (SE)	Oppose $\mathit{dy}/\mathit{dx}$ (SE)	Somewhat Oppose $\mathit{dy}/\mathit{dx}$ (SE)	Neither $\mathit{dy}/\mathit{dx}$ (SE)	Somewhat Support $\mathit{dy}/\mathit{dx}$ (SE)	Support $\mathit{dy}/\mathit{dx}$ (SE)
Reduces climate change	0.306 * (0.160)	−0.013 (0.08)	−0.009 * (0.005)	−0.031 * (0.017)	−0.037 (0.20)	0.089 * (0.046)
No need for fuel	0.615 *** (0.180)	−0.023 ** (0.009)	−0.016 ** (0.007)	−0.058 *** (0.018)	−0.072 *** (0.023)	0.169 *** (0.045)
International norm	0.593 ** (0.255)	−0.018 ** (0.008)	−0.013 ** (0.006)	−0.050 *** (0.019)	−0.068 ** (0.028)	0.148 *** (0.052)
Too dependent on the weather	0.499 ** (0.195)	−0.017 ** (0.008)	−0.012 ** (0.006)	−0.046 *** (0.018)	−0.059 ** (0.023)	0.134 *** (0.047)
Bills will rise	−0.526 ** (0.231)	0.034 (0.023)	0.019 (0.013)	0.062 ** (0.031)	0.060 ** (0.026)	−0.175 ** (0.084)
cut1	−1.539 (0.174)
cut2	−1.282 (0.153)
cut3	−0.713 (0.130)
cut4	−0.199 (0.124)
Likelihood ratio	−240.3 ***
AIC	498.6
$X^2$	48.76
Pseudo $R^2$	0.075

***, **, * indicate significance at 1%, 5% and 10% respectively. The model combines reasons for supporting RE (Table 2) and opposing RE (Table 3). The backward stepwise selection method was applied to eliminate variables with a significance above 20%, and only those with significance of 1 to 10% are given. dy/dx indicates marginal effect. The p-values for each marginal effect are omitted due to space limitations. cut1 represents the threshold for ‘oppose’ and ‘somewhat oppose’; cut2 represents threshold for ‘somewhat oppose’ and ‘neither’; cut3 represents the threshold for ‘neither’ and ‘somewhat support’; and cut4 represents the threshold for ‘somewhat support’ and ‘support’.

gives the marginal effects of reasons for supporting or opposing renewable energy on overall support for RE. As with Table 3, Table 5 used the backward stepwise selection method, resulting in a more parsimonious model with a lower AIC. Firstly, arguments in favour of RE have a positive impact on support for RE: ‘reduces climate change’ (0.306), ‘no need for fuel’ (0.615) and ‘international norm’ (0.593). On the other hand, the argument against RE ‘too dependent on the weather’, also had a positive impact on support for RE. Concern that ‘bills will rise’ had a negative impact on support for RE, as expected. The marginal effect of ‘reduce climate change’ was negative for the first four categories (‘oppose’ to ‘somewhat support’), but positive for ‘support’ (0.089). The same pattern is true for the other three variables contributing to support for RE, with ‘no need for fuel’ being particularly strong, with those selecting this variable having a 16.9% increased probability of supporting RE. The pattern is reversed for those who selected ‘bills will rise’, suggesting those who selected this were 17.5% more likely to not show support for RE, all other variables remaining the same.

3.3. Demographic Variables and Reasons for Supporting or Opposing RE

Table 6. Demographic variables and reasons for supporting or opposing RE.

Variable	Reasons for Supporting RE
	Price (SE)	Fuel (SE)	Install (SE)	Economy (SE)	Durable (SE)
male	−0.345 ** (0.157)			−0.353 ** (0.168)	−0.296 * (0.170)
children			−0.168 *** (0.006)		−0.008 (0.006)
age	0.011 (0.007)			0.018 ** (0.008)
education		0.126 (0.091)	0.191 *** (0.096)
salary		0.054 ** (0.270)
constant	−0.542 ** (0.270)	−0.978 *** (0.358)	−0.959 ** (0.368)	−0.368 ** (0.287)	−1.142 *** (0.134)
Likelihood ratio	−202.1 **	−198.9 **	−185.3 ***	−178.8 **	−161.6 **
${\chi }^2$	6.05	7.30	11.53	8.50	4.84
Pseudo $R^2$	0.015	0.018	0.030	0.024	0.015
Variable	Reasons for opposing RE
	Weather (SE)	Eyesore (SE)	Storage (SE)	Impractical (SE)	Wildlife (SE)	Bills	Satisfied
male	−0.611 *** (0.168)	−0.505 *** (0.179)	−0.304 * (0.178)		−0.439 ** (0.188)
children	−0.008 (0.005)	0.008 (0.006)				0.009 (0.007)	0.011 (0.007)
age							−0.032 ** (0.012)
education	0.163 (0.101)					−0.249 * (0.110)	−0.346 ** (0.118)
salary				−0.068 * (0.037)			0.069 ** (0.035)
E. Kazakhstan		0.497 * (0.279)				0.596 ** (0.300)
constant	−0.838 ** (0.411)	−0.778 *** (0.146)	0.523 *** (0.136)	−0.688 *** (0.149)	−0.808 *** (0.102)	−0.486 (0.428)	0.696 (0.611)
Likelihood ratio	−174.3 ***	−154.7 ***	−146.6 **	−140.2 **	−136.3 **	−105.4 **	−87.0 ***
${\chi }^2$	19.00	13.69	8.75	3.93	5.65	10.40	20.84
Pseudo $R^2$	0.052	0.042	0.029	0.014	0.020	0.047	0.107

***, **, * indicate significance at 1%, 5% and 10% respectively. The backward selection method was applied to eliminate variables with a significance above 20% and only those with a significance between 1 and 10% have been included.

models the reasons for supporting and opposing RE with demographic variables. As with the previous regression models, the backward stepwise selection method is applied to produce a model with the best fit using the AIC. The likelihood ratio test rejects the null hypothesis that all the coefficients are zero. So, despite the low pseudo R2, we will discuss the model under the assumption that the model is valid.

Firstly, looking at gender, women were more likely to cite prices falling due to RE (‘prices’), RE’s positive impact on the economy (‘economy’) and the durability of RE (‘durable’) as reasons for supporting RE (−0.345, −0.353 and −0.296, respectively). On the other hand, women were more likely to oppose RE when citing dependence on the weather (−0.611) and its visual incompatibility with the surrounding environment (‘eyesore’, −0.505). Respondents with higher levels of income were more likely to consider that RE did not need a fuel source (‘fuel’) as a positive argument for RE; however, the same group were also more likely to be satisfied with their current electricity rates (‘satisfied’, 0.069). Those with lower levels of income were more likely to see RE as impractical (‘impractical’, −0.068). Respondents with higher levels of education tended to see ease of installation (‘install’) as a positive argument (0.191). Those with lower levels of education were more likely to see the possibility of bills increasing due to RE (‘bills’, −0.249) and were more likely to be satisfied with current rates (‘satisfied’, −0.346). Those without children saw ease of installation as a positive argument (−0.015) but were also concerned about the inadequacy of storage technology (‘storage’, −0.015). Residents of East Kazakhstan were more likely to see the RE facilities as an eyesore (0.497) and be concerned with the possibility of their bills increasing (0.596). The dangers of nuclear testing did not feature as a significant variable, suggesting the null of Hypothesis 3 (“knowledge of nuclear contamination has a direct influence on the degree of support for RE amongst East Kazakhstan residents”) cannot be rejected.

4. Discussion

Kazakhstan is rich in fossil and mineral resources, and as a result, fossil fuels have been the country’s main source of electricity. As environmental problems become more serious, Kazakhstan is shifting from fossil fuels to renewable energy sources. As of 2022, while countries around the world are committing to reduced coal consumption, natural gas is seen as a realistic and immediate replacement energy source. In their “One Belt, One Road” initiative, President Xi Jinping has made the reduction of coal consumption a pillar of their energy policy and has recommended natural gas and renewable energy sources such as wind and solar power as alternatives. In implementing its low-carbon energy strategy, Kazakhstan is using solar panels imported from China and Japan. In 2021, one of the largest wind farms in Central Asia was built in southern Kazakhstan with the help of Chinese financing. Kazakhstan is seen as having a key role in China’s “One Belt, One Road” initiative. However, as mentioned above, most of the existing case studies of Kazakhstan have focused on social and medical research relating to the Semipalatinsk Nuclear Test Site and on the consequences of the Aral Sea disaster. This paper takes Kazakhstan as a case study to examine the role knowledge plays in influencing the public’s choice behavior for next-generation energy sources, including renewable energy. Several points are worthy of highlighting. The first hypothesis was in relation to support for thermal power generation. Given that those with knowledge of Kazakhstan’s natural gas reserves supported thermal power generation, the null hypothesis “Knowledge of Kazakhstan’s large reserves of uranium does not influence support for nuclear power generation” can be rejected. Karatayev and Clarke [23] state that Kazakhstan relies on fossil fuels for energy and that GHGs have a negative impact on the environment and people’s health. This suggests that many participants support the switch to natural gas, which emits less GHGs than coal and oil. On the other hand, even though the participants knew that Kazakhstan has large reserves of uranium, they did not support nuclear power generation. Therefore, the second null hypothesis “Knowledge of Kazakhstan’s large reserves of uranium does not influence support for nuclear power generation” was not rejected. Fyodorov and Kayukov [11] state that reducing the environmental impact caused by uranium mining is an important national policy in Kazakhstan. The analysis in this paper reveals that few of the participants desire expansion of nuclear power generation just because of the available large reserves of uranium. The next set of hypotheses looked at renewable energy. The null hypotheses (3) “Knowledge of Kazakhstan’s wind power potential does not influence support for the use of wind power generation” and (4) “Knowledge of Kazakhstan’s solar power potential does not influence support for the use of solar power generation” are both rejected by the analyses. In other words, knowledge of RE potential was a determining factor in support for the development of those energy sources, regardless of awareness of their strengths and weaknesses.

The three most favored reasons for supporting RE were that it is not reliant on a finite fuel source, the widespread international adoption of RE, and the contribution it makes to reducing global climate change. Therefore, the fifth null hypothesis “Knowledge of Kazakhstan’s potential for renewable energy does not influence support for the use of renewable energy” can be rejected. Demographic factors played some part in determining support. Women were aware of both the advantages and disadvantages of renewable energy but were equally likely to oppose it as to support it. Those with lower income and lower education were less likely to be supportive. In East Kazakhstan, the site of historical nuclear testing, it was assumed by the researchers that knowledge and experience of nuclear testing would lead to support for safer and cleaner renewable energy over the alternatives. However, the results do not support this assumption due in part to current low energy prices and the fear that adoption of RE would increase prices. From this, we find that the sixth null hypothesis “Knowledge of Kazakhstan’s potential for renewable energy does not influence support for the use of renewable energy” cannot be rejected.

As with many other countries, Kazakhstan is committed to a low-carbon energy strategy that takes advantage of its abundant energy resources. In addition, it is required to reduce coal consumption and promote alternative energy sources such as renewable energy. To summarize our research, it was found that support for next-generation energy was strongest amongst those who were more conscious of the need to protect the global environment and were aware of Kazakhstan’s huge potential for renewable energy. As there are no other studies of Kazakhstan addressing these issues, we are unable to compare the results with previous research. This study is very much a tentative first step. However, the results of this study do suggest that there is widespread support for renewable energy, despite the awareness of Kazakhstan’s huge reserves of fossil fuels and uranium. Although Kazakhstan will probably be a huge supplier of fossil and nuclear fuel under the One Belt, One Road initiative, Kazakhstan is willing and able to develop a de-carbonised economy and enact policies that reduce GHG emissions and promote renewable energy, providing a model for the Eurasian region.

5. Conclusions

This paper statistically analyzed the impact knowledge had on support for different energy sources. The results revealed the following points. We found that knowledge of a particular energy source generally led to support for that energy type. The greater the awareness of Kazakhstan’s fossil fuel resources, the greater the support for non-renewable power generation; the greater the awareness of Kazakhstan’s potential for renewable energy, the greater support for RE power generation. The reasons given for supporting RE included the fact that it is not a finite resource, the fact that RE is becoming increasingly common around the world, and the fact that it contributes to tackling GHG emissions. Women tended to be more knowledgeable about RE and were aware of both its advantages and disadvantages. Additionally, many of them were both proponents and non-proponents of RE. However, those with lower income and education levels were less likely to support RE. Those surveyed in East Kazakhstan were divided on their support for RE, divided in part by awareness of the dangers of nuclear testing, and concern over increased energy prices. Those who supported next-generation energy in Kazakhstan were highly conscious of environmental issues and climate change. However, residents of East Kazakhstan with low average incomes were less likely to accept RE, even if they had been affected by nuclear tests. They may be concerned that the widespread use of RE will lead to higher electricity costs, partly because the electricity tariff in East Kazakhstan Oblast (8 KZT/kwh) is 37.5% lower than that in Nursultan (11 KZT/kwh). Kazakhstan’s Gini coefficient is 27.5% (2020, UN statistics), the same rate as Norway (USD 81,550, 2018, IMF statistics), which has the fourth highest GDP per capita globally. Although Kazakhstan has low-income inequality by global standards, regional income inequality does exist. GDP per capita in the East Kazakhstan Oblast is lower than the national average. Even East Kazakhstan residents that have experienced nuclear testing may not accept renewable energy. This may result in higher electricity prices, because their incomes are lower than the national average. Given the rise in global instability due to conflict, climate change, and the rise of populism, it is not surprising that many people are uncertain of the future and skeptical of new technologies that require new lifestyle patterns that can be disruptive and may result in failure. Globalization and integration into the global economy brings the promise of prosperity and other social benefits, but can also bring inequality and disruption.

In future research, we will return to examine how attitudes towards next-generation power in Kazakhstan have evolved as it becomes either more isolationist or more integrated into the regional and global economy. The results of the survey in this paper do not satisfactorily reveal whether there were any differences in the responses of those who knew about the effects of the nuclear tests and those who did not. A web-based survey has inherent limitations, and even more so when conducted in countries and regions with less than ideal cellphone and internet coverage. As discussed above, budget constraints meant the survey size was kept to a minimum while remaining statistically useful. We were unable to accurately target East Kazakhstan through the use of quotas. This would have produced more accurate and useful data that would possibly have allowed the null hypothesis addressing knowledge of nuclear testing on support for RE to be rejected. To address this, future research will need to target East Kazakhstan residents, either through the use of an in-person survey instrument, a paper-based survey, or the establishment of a survey platform that offers better presence in the region.