Exploring willingness-to-pay for the restoration and maintenance of reserved forests in the Greater Accra Region of Ghana

Anthony Amoah, Adusei Jumah, Kofi Korle

a School of Sustainable Development, University of Environment and Sustainable Development, Somanya, Ghana & Africa Centre for Economics and Policy Research(ACAEP), Ghana

b Department of Economics, Central University, Accra, Ghana

Keywords:Reserved forest Restoration Maintenance Tobit model Willingness-to-pay CVM

ABSTRACT In recent years, debates on the alarming rate of forest depletion emanating from growth in urban settlement and changed urban land-use patterns have gained prominence across the globe. The present study adopts a demandside management approach to investigate household's willingness-to-pay for the restoration and maintenance of protected forest reserves in three municipalities in Ghana.Using survey data of 733 households from the Greater Accra Region of Ghana, we model the demand for forest restoration and maintenance, respectively, by means of the contingent valuation technique.As part of the findings, the study shows evidence that a household is willing to pay Gh￠50.99 ($8.67) and Gh￠31.12 ($5.29) per annum to restore and maintain the protected forests,respectively. These amounts constitute less than one percent of the average household income per month.Consequently, critical validity tests are conducted to validate the robustness of the results. This study provides willingness-to-pay estimates for forest and forest resources,and its associated determinants.These estimates seek to bridge the information gap and inform policy decisions toward the overarching aim of ensuring sustainable forest management in Ghana.

1. Introduction

Africa's standing as the region with the world's fastest population and urban growth is having a toll on its forests (Clarke, 1993). The driving forces behind Africa's current rapid urbanization are a combination of rural-urban migration and natural increase within towns and cities themselves(Songsore, 2009). As observed by Potts (1997),this is worsened in some regions by forced migrations precipitated by various stresses including ethnic conflicts,wars, droughts and famine.

In 2010,the Greater Accra Region(GAR)-which houses the capital of Ghana, Accra - had an urban population of 90.5 percent (GSS – Ghana Statistical Service, 2012). The centre of this urbanization process in the region is the Greater Accra Metropolitan Area(GAMA),which houses the three urban reserved or protected forests presently understudy. The population of GAMA is projected to more than double to 10.5 million people by 2040 (GSS – Ghana Statistical Service, 2012). Similarly, the National Land Use Planning Committee estimated in 1990 that a unit increase in the urban population requires an additional land area of 33.3 ha for the provision of additional housing,infrastructure,and other social services. Usually, substantial parts of the forest reserves are lost to infrastructural development, especially during road construction and extension of electricity grids—among others. Therefore, as the urban population increases, more forestlands would be converted into other uses, thus limiting the forest resource base(Agyarko,2001).

Cobbinah et al. (2017) have argued that urbanisation and climate change underscore the rapid depletion of environmental resources.Similarly, recent empirical evidences by Amoah and Korle (2020) have provided evidence of urban reserved forest depletion in Ghana and called for efforts towards its sustainable management. Further, they identified human behaviour, institutional failure and climate change as the main drivers of forest depletion. Although the call is justified, the “public”nature of environmental goods like the reserved forests presents a free-riding behavioural problem making it difficult for individuals to pay for direct and indirect forest use or protection of the forest and forest resources.

At stake is a deeper understanding of the factors associated with individual decision-makers’ valuations of non-market benefits obtained from forest conservation. Individuals can make economically optimal decisions regarding investments in forests for such benefits only if they understand the benefits of improved forests on their well-being and possibly their labour market outcomes (Glewwe and Miguel, 2008;Dupas,2011). A critical question facing economists that relates to overcoming the challenge of valuing non-market benefits from improved forests is the extent to which households want to invest in forest conservation. Usually, households have imperfect knowledge about the potential benefits of forest conservation since the economic returns to improved forests are generally unknown.Thus,if decision-makers within households systematically undervalue forest inputs relative to their true long-term benefits,they will rationally underinvest in forest conservation and vice versa(Cummins et al.,2018).Such an understanding underpins consumers’willingness to pay(WTP)for improvement in environmental quality—a concept that is essential for predicting the success of various forest conservation measures in a deregulated market and for helping regulators maintain an appropriate mix of forest conservation support policies (Roe et al., 2001). It is envisaged that a poorly considered valuation of non-market benefits from improved forests can jeopardize investments in their development and threaten innovation failures(Ingenbleek et al.,2013).

The current paper seeks to determine the annual households’WTP for the restoration and maintenance of depleted protected forests in GAMA.The overall goal is to estimate the annual average values that households assign to the protection and restoration of degraded forests.Specifically,we estimate the WTP per annum, based on a hypothetical market and several variables that account for the variation in observed respondent characteristics (socio-economic characteristics) such as income, education,gender,age and married status(married).Specifically,we examine the Achimota Forest,Shai Hills Forest and Gua Kuo Forest(Sacred Grove)in the GAR—the most urbanised region in Ghana.These three forests are endangered protected forests threatened by encroachment and other illegal commercial activities because of the ever-increasing pressures emanating from the intensifying urbanisation of GAMA (Appiah, 2009;Antwi-Baffour,2016).

Despite the limitations of the WTP approach (see e.g., Diamond and Hausman,1994),many researchers(e.g.,Petrolia et al.,2014;Dickinson and Whitehead, 2015; Kotchen et al., 2017; Amoah et al., 2019) have argued that it is one of the few reliable methods in respect of resource valuation that expresses an individual's appraisal of a given resource.Along this proposition, the results from our analysis could provide reliable estimates of the benefits accruing to society of sustainable forest management in Ghana.It is important to acknowledge that all estimates are presented in Ghana cedis with their dollar equivalent for both domestic and international audience to appreciate the context.In addition,June 2019 is used as the base year, where Gh￠1 is exchanged for USD$0.17.

The organization of the paper is as follows: Section 2 reviews the theoretical and empirical literature on forest management and the WTP concept; Section 3 expounds the survey design and model employed in the study; Section 4 interprets the empirical results and Section 5 concludes.

2. Literature

2.1. Theory of planned behaviour, forest resource valuation and proenvironmental behaviour

Although forests are an important source of income and livelihood for several households, there exists a gap in information regarding people's preferences among countries in Africa(Fisher,2004;Mamo et al.,2007;Heubach et al., 2011). Considering preferences of individuals, Owubah et al. (2001) argue that an individual's willingness to sustain a forest is largely influenced by the value attached to it. Given that values have a strong framing effect on attitudes, which in effect influences WTP behaviour,it goes without saying that this argument can be explained by the theory of planned behaviour.

According to the theory of planned behaviour,intention is the direct precursor of behaviour and is dependent on attitude toward the behaviour(Ajzen,1991).That is,attitude affects behaviour through intention.A strong correlation between attitude and behaviour is observed when attitude is measured towards a specific corresponding behaviour. For example,as a consumer's income increases,other things equal,there is an observed increase in the level of consumption. The increase in income brings about a higher motivation(intention)through an improvement in purchasing power, which leads to an increase in expenditure and consumption(behaviour).

Among models of environmental acts,Kollmuss and Agyeman(2002)suggest a model of pro-environmental behaviour where environmental knowledge, values, attitudes and emotional involvement are factors influencing pro-environmental consciousness. Here, the notion of pro-environmental consciousness constitutes a broader set of personal values and traits, and other internal and external factors. Arguably,synergy among internal and external factors are expected to be the largest influence on pro-environmental behaviour. Possible barriers to pro-environmental actions identified include monetary constraints,habit formation,inadequate incentives,poor environmental consciousness and illiteracy.In addition,Kollmuss and Agyeman(2002)underscore the fact that direct experience of extinction will lead to a stronger influence on people's behaviour on the environment than through indirect knowledge obtained by formal education.

Similarly,factors influencing forests and the environment in general are determined by demographic, economic, psychological and institutional factors(e.g.,Kollmuss and Agyeman,2002;Clark et al.,2003;Onel and Mukherjee, 2016; Blankenberg and Alhusen, 2018). Each of these factors influences behaviour and attitude towards the environment.Thus,any behaviour as influenced by these factors towards reducing the harmful effects of human and non-human activities on the reserved forest and forest resources depletion is deemed to be pro-environmental.

2.2. Deforestation in Ghana: an overview

According to FAO (2010a) deforestation is the transformation from forested land to non-forested land during a certain time. Again, FAO(2010b)defines forest reserve as“forest land within reserves and under protection” (p. 5). The underlying causes of deforestation of forest reserves in Ghana include population growth, urbanization, climate change,inadequate stakeholder participation in forest management,low forest taxes & fees regime, weak institutional structures, inadequate investments in the forestry sector, among others (FAO, 2010b; Quacou,2016). Through policy interventions such as European Union's Forest Law Enforcement, Governance and Trade Programme and the initiative to reduce carbon emissions from deforestation and forest degradation in developing countries, some of the aforementioned causes have been reduced over the years (Hajjar, 2015). However, the problem of deforestation especially in the case of reserved forest and forest resources are yet to be effectively addressed.

Similarly, Oduro et al. (2015) have observed that the current widening gap between the demand and supply of wood, continuous degradation of closed natural forests and the low rate of forest plantation establishment makes a forest transition in Ghana difficult to achieve.However, the analysis of the forest resources development trends demonstrates the potential of rehabilitation and reforestation activities,even under different contexts,for a forest transition in Ghana(Meyfroidt and Lambin, 2011). Implementation of sustainable forest management practices, intensification of forest plantation establishment, supporting and encouraging conservation and integration of trees into farming systems,will accelerate a forest transition in Ghana.

The decision to protect forest resources depends on the structure of tenure arrangement and whether the benefit stream reaped by the consumer outweighs the benefit stream from alternative uses(Owubah et al.,2001).It follows that an individual's willingness to engage in sustainable forestry practices would be determined by forest tenure practices, individual preferences and characteristics. The new national reforestation plan following the Modified Taungya System (MTS), provides stakeholders with greater economic benefits and livelihood sources(Acheampong et al., 2016). The MTS incorporates benefit sharing arrangements that recognize farmers' tenure rights, unlike the former Taungya system where benefits accrued to the Forestry Commission and landowners only.This new system of forestry management has led to the successful reclamation of some degraded forest cover,sustainable future requirements of wood industries,reduced land scarcity,reduced poverty– as farmers are paid for services such as planting and tending, and generation of revenue for the country and other stakeholders (FAO,2008;Acheampong et al.,2016).

FAO(2020)attests to the fact that variations in forest area over time indicate changes in demand for land for other uses. Growing urban settlement accompanied with changing land use pattern have increased demand for land for constructional purposes (Agyarko, 2001). The tremendous rise in the proportion of population moving into urban areas in the study area has implications for the supply of urban land for housing and for the provisions of infrastructure and other social services.

2.3. Empirical approaches to forest resource valuation

There exists a plethora of procedures for assigning prices to nonmarket goods and services within the environment (e.g., Jorgensen and Syme, 2000; Johnston, 2006; Atkinson et al., 2012; Baker and Ruting,2014).These approaches are broadly categorized as revealed and stated preference methods.Revealed preference techniques use actual individual behaviour in a market environment while the stated preference techniques use expressed or expected behaviour under hypothetical market situations. Among the class of stated preference techniques appropriate for non-market hypothetical analysis, the contingent valuation(CV)method is the most widely used(Iranah et al.,2018).

The CV method quantifies individual WTP or willingness to accept a change in the quality or quantity of an environmental good (Ciriacy-Wantrup,1968;Hanemann,1994).It is a commonly used method in quantifying ecosystem services, recreation and other externalities accruing to consumers. As a shadow price technique, it involves asking potential beneficiaries to state their WTP for a good or service for which market prices are either non-existent or are a poor reflection of their true worth(Baker and Ruting,2014).

In line with obtaining accurate estimates,Getzner et al.(2018)posits that the validity and reliability of stated preferences and measures obtained through CV methods or choice experiments are conditional on descriptions of the scenarios to be valued. Johnston et al. (2017) also point out that stated preference questionnaires must clearly outline actual conditions, expected changes in quality and quantity, ensuring that these attributes are well understood by respondents.In this regard,objectivity in measurements and perceptions of attributes measured must be considered,as employed in this study,to quantify the stated amounts about forest reserves accurately.

Somefew related studies have been conductedin Ghana's forestry sector making use of CV method.Navrud and Vondolia(2005)find that increasing entrance fees for visitations to Kakum National Park(located in the central region)from$3 to$9 for domestic tourists and from$10 to$37 for foreign tourists will maximize revenues for managing the Park.In a related study,Ansong and R?skaft(2014)examine the WTP of 300 respondents from 10 communities around Subri Forest Reserve in the Western Region of Ghana.The authors observe a mean monthly WTP of about$2.3 towards the provision of sustainable forest services.Older and higher income earning respondents are willing to pay higher than younger and lower income earning categories.The findings suggest that residents close to forest reserves are likely to support forest management initiatives that ensure sustainable usage.These studies,however,do not have instances of zero valued WTP which would require a more suitable technique.

Incidentally, socio-economic or demographic factors are important variables associated with forest conservation in Africa. These factors include age, gender, education, income,perception of the environment,dependency ratio and system of land tenure among others(Endalew and Wondimagegnhu, 2019). With regard to income, Vincent et al. (2014)establish that rising household income is associated with increased WTP for forest resources.Higher incomes earners are more likely to spend on nature-based tourism and conservation. However, Iranah et al. (2018)and Kamri(2013)suggest that some high-income earners may exhibit a lesser WTP because they expect their tax obligations to suffice in contributions towards social welfare.

3. Materials and methods

3.1. Contingent valuation survey

This study relies on the CV method of eliciting non-market information through a hypothetical market to estimate the value a person places on two goods. The CV method relies on an accurate description of the good in question. Using a hypothetical market scenario to describe the good in question, the following restoration and maintenance questions were presented to respondents:

I would want to find out from you, if you value forest and forest resources in Ghana,particularly in the Greater Accra Region.By forest and forest resources we mean the forest is not depleted, people are not allowed to indiscriminately cut down trees in the forest, and all forest habitats have been restored and well maintained. The air in the community has greatly improved, tourism and its associated benefits have increased in your community etc. Generally, we know that every good thing comes at a cost.You may be required to pay an amount that will be factored into your property rate provided by the District Assembly.

Restoration or Maintenance Question:“Are you willing to pay X-amount1Note: X-amount is a number ranging from Gh￠10, Gh￠20, Gh￠30, …, Gh￠200 per annum which were obtained during the pilot survey and randomly assigned to respondents.per annum(Refer to Fig.1)for the……..(restoration or maintenance)of the forest and forest resources in your district?”

From Fig. 1, whether good 1 or good 2, the respondent is presented with a randomized starting amount which was determined from the pilot survey.This amount ranges from Gh￠10($1.70)to Gh￠200($34).If the respondent agrees to the starting bid(Yes answer),a higher(second)bid is presented for the respondent to make a choice decision.However,if the starting bid is rejected(No answer),a lower(second)bid is also presented for the respondent to make a choice decision.The same process is applied for the third bid.In all cases,based on the third(higher or lower)bid and now resorting to an open-ended format,the respondent is given the opportunity to finally state his/her maximum WTP for the good in question.This strategy gave all respondents a triple-bound open-ended bidding game format in determining their maximum WTP for either good 1 or good 2.

Fig. 1. Bidding game for restoration (Good 1) and maintenance (Good 2).

3.2. Survey design

To apply the CV method,the bidding game format was included in a structured questionnaire which was administered to households in three major municipalities of the GAR. These municipalities are homes to the three respective forests under investigation.The forests are the Gua Kuo Sacred Forest, which is located in the Ga North Municipality, the Achimota Protected Forest in the Okaikwei North Municipality and the Shai-Hills Protected Forest which is located in the Shai Osudoku Municipality.According to the last(2010)population census of the nation,the Region is made up of a population of 4,010,054 inhabitants with 1,036,426 number of households (GSS, 2012).

A standard well-structured questionnaire,which follows the format of the Ghana Statistical Service for National Surveys, was developed and used as the main research instrument.The questionnaire was structured into sections which include demographic questions, forest and environment related questions, and measures of forest worth to respondents using contingent valuation methods.

Fieldwork for the study began with a pilot survey, which was completed during the early of March 2019 in different communities which were not included in the final sample used.Queries,comments and concerns were raised for discussion,review and eventual revision of the questionnaire. The final questionnaire was further reviewed and approved by experienced survey experts. The survey was conducted inperson. Actual data collection based on a stratified sampling technique was applied within the three municipalities during the period April–June 2019.

Following Amoah et al. (2019), each municipality was sub-divided into a number of communities to constitute unique strata. Housing units with household heads were randomly selected in each stratum after it has been well listed.Altogether,a total number of 733 household heads constituted the number of observations used for the study.This comprises 248(33.83%)from Okaikwei North Municipal Assembly,243(33.15%)from Ga West Municipal Assembly and 242(33.02%)from Shai Osudoku Municipal Assembly. It is important that the calculated sample size yielded approximately 400 observations.Yet,we oversampled to obtain a relatively larger sample for a better representation of the population.Respondents were given the opportunity to participate in a draw of Gh￠5(USD$0.85) worth of mobile phone recharge as a way of incentivizing them.

3.3. Econometric specification

3.3.1. Modelling

The current study seeks to estimate the WTP for forest restoration and maintenance. It relies on the maximum amount respondents are willing to pay for the restoration,and maintenance as the dependent variables in our estimated models.The relationships can be generally specified in an econometric model as:

where WTPirepresents the WTP for respondent i, Xiis a vector of the independent variables which are mostly household level characteristics.The phi(φ)and beta(β)are the unknown parameters to be estimated.uiis a stochastic term that follows a standard normal distribution.

In WTP studies a non-response or expression of unwillingness to pay by respondents often leads to zero values in the series for the dependent variable. In such probable instances, as in this study, the bidding game with open ended responses may be prone to some biases.Stewart(2013)argues that the Tobit approach is more appropriate in such cases where the dependent variable have zeros. However, the censored Tobit model appears to be a more powerful tool compared to the Logit and Probit models commonly used in the literature. This is because the censored Tobit model has inherent treatment of responses by truncating all outliers, whether too low or high in the data generating process,as well as dealing with zero entries. Stated differently, the Tobit model has traditionally been used to account for censoring in a data.It is generally based on the assumption that“the same stochastic process determines both the value of continuous observations on the dependent variable and the discrete switch at zero”(Blundell and Meghir,1987,p.2).In spite of the appropriateness of the Tobit model for the present study, Amore and Murtinu(2021)argue that Tobit models may lead to bias estimates when a) the nature of the dependent variable is wrongly determined, b) the distribution of the residuals is skewed, and c) “the difference between selection concerns and censored data” exists (p. 339). For limited dependent variable or dichotomous or binary dependent variables, the Logit/Probit models have been used(Wiersema and Bowen,2009).

The zero WTP amounts for restoration and maintenance of the forests constitutes 16.10% and 17.19%, respectively. Using ordinary least squares (OLS) or Logit and Probit models suggests that quite a large number of observations will be excluded from our dataset. Against this background,we present a Tobit model in the form:

where “Bid” is a bid parameter captured in the model as a control measure for stating point bias.This is the initial amount that all respondents are mandated to answer first before proceeding with other bids. A statistically insignificant bid parameter is an indication of no stating point bias in the model.“l(fā)nincome”is the natural logarithm(ln)of household take-home income.This measure is considered as an essential variable in modelling demand. It is expected that higher levels of income will positively drive WTP for forest restoration and maintenance.“Age”of the respondent is measured in years.This is used as a measure of experience over time. It is expected that older people have much more positive experience with the protected forests and will demonstrate higher WTP.“Gender” differences matter in behavioural choices. The direction of gender effect for the protected forests is ambiguous. However, on average, women are more inclined to be concerned about the environment than men.Gender is included as a dummy where male is denoted as 1 and female is denoted as 0.“Edu”(Education)is included as a dummy.Educated respondents are denoted as 1, while those without education are denoted as 0. The educated here are respondents who have been to school whiles uneducated are respondents who have not been to school.A priori,education is expected to vary positively with sustainable forest management. “MS” (Married) is also included as a dummy. Those who are married are represented as 1,while the unmarried are represented as 0. Corporate effect is expected to influence WTP decisions. By implication,those who are married should be expected to pay more as compared to the unmarried.Objective environmental knowledge is captured in two ways,local objective environmental knowledge(LOK)and international objective environmental knowledge (IOK). In the case of the former,respondents indicated their knowledge of any local environmental issue or law, whiles in the case of the latter, respondents indicated their knowledge of an international environmental issue or law.In both cases,those who indicated yes,were again asked to validate their position with an example or illustration. Those who indicated correctly were considered as having objective knowledge.We expect objective environmental knowledge (local or international) to have a positive association with WTP. The error term (u) is assumed to be normally distributed with a constant standard deviation and zero mean.

3.3.2. Model diagnostic tests

The commonly identified econometric problems that easily impact on estimates of cross-sectional data include multicollinearity, heteroscedasticity and endogeneity. To avoid multicollinearity among the explanatory variables,we used the pairwise correlation test to investigate the correlation properties of the variables. Given that the highest coefficient of 0.496 was obtained for the relation between WTP and Bid,we conclude that there is no severe multicollinearity in the model (Supplementary Material Table S1).

Heteroscedasticity was accounted for by estimating the model with robust standard errors.Therefore,it is expected that the negative effect of heteroscedasticity will not severely impact our results. Generally, given that our independent variables are represented by the socioeconomic characteristics of the respondents, and that these vary given the unique characteristics of each respondent,WTP is not expected to influence any independent variable. In effect, we do not expect reverse causality,misspecification and omitted variable bias to have a severe impact on our estimated coefficients.Results of the F-test shown in Tables 4 and 5 reject the null hypothesis that the overall models are not statistically significant. Thus,all the estimated models are statistically fit.

3.4. CVM validity diagnosis

In line with the CVM criticism by Hausman (2012), we test for hypothetical bias and income effect. With respect to hypothetical bias, we followed the ex-post and ex-ante treatment approaches of Loomis(2011,2014)and Amoah et al.(2019).As earlier mentioned,respondents were made aware that their support for the restoration and/or maintenance is by paying an annual amount through property rates to the Municipality.This approach is augmented with an ex-post evaluation of the data.This is achieved by deriving the rational income ratio(RIR)which represents the share of mean income that is allocated to restoration or maintenance by the respondent. This is expressed mathematically as:

The rule of thumb is,there is hypothetical bias if the RIR falls below a threshold of 20 percent.This threshold implies that it will be illogical for a respondent to allocate 20% or more of his/her income to forest restoration or maintenance. The test results show that for restoration and maintenance - given the 95% confidence interval - the expected mean WTP values for the full sample are similar to the sub-samples except when the subsample is 20%.As a result,there is overwhelming evidence that the hypothetical market design is not hypothetically biased. Additionally, the non-independence in the WTP values of the sub-samples were controlled for using the bootstrap confidence intervals (See Bateman et al.,2002;Amoah et al., 2019).

To realize the second validity test which is a standard theoretical validity test, we employ the statistical one-sample t-test which assumes that the income parameter equals zero. If the test fails to reject the null hypothesis, it means there is no income effect, and that, its statistical equivalence equals zero.The results as shown in Tables 1 and 2 provide overwhelming evidence that the income effect is non-zero,and that,WTP for both restoration and maintenance vary with respect to income.

Traditionally, the scope sensitivity test (scope effect) has been investigated in WTP studies. However, recent evidences by Heberlein et al.(2005)and Desvousges et al.(2012)have argued that this test is less informative or better still uninformative in providing validity evidence to support WTP estimates. Consequently, the scope-effect test is excluded from the current study. The argument at this juncture is that the hypothetical validity and theoretical validity tests provide enough basis to validate our estimates.

4. Results

This study aims at examining households WTP for restoration and maintenance of protected forests in the GAR of Ghana. The descriptive statistics of the relevant variables are presented in Table 3.

Out of a total number of 733 respondents,we recorded approximately 17% non-responses hence the full sample used for this study is 627 observations. The unconditional mean WTP for the restoration (WTPr) of the protected forest is Gh￠53.32 ($9.06) per annum while that of maintenance (WTPm) is Gh￠34.82 ($5.92) per annum. Both have some zero observations, hence need to use a Tobit model instead of the usual OLS.

The average take-home income of respondent's household is Gh￠1102.46 ($187.42) per month (i.e., Gh￠13,229.52 ($2249.02) per annum) with a minimum of Gh￠145 ($24.65) and a maximum of Gh￠6200($1054).The average age of the respondent is about 35 years with the minimum and maximum age being 18 and 81 years respectively.The age and income distributions show a good spread of the different categories of people. The sample had about 57% of the respondents being males while 43%were females.This gives a fair description of the male dominated feature of most household heads in Ghana.

Most people were found to be educated. In addition, 67% of the respondents were found to be married or living with a partner as legitimate couple,while 33%were single(or divorced or widowed)as of the time of the interview. The study sought to categorize objective environmental knowledge into domestic or local and international. An average of 31%was found to have international objective knowledge, while 29% was found to have local objective knowledge.This disparity is plausible given that both social and the traditional media dominate with more international environmental issues than country specific environmental issues and laws.

Table 1 Hypothetical bias and income effect tests for restoration of the forest.

Table 2 Hypothetical bias and income effect tests for maintenance of the forest.

Table 3 Descriptive statistics.

The data used for the study was obtained using the bidding game format. Given that this technique is prone to starting point bias, we controlled for this bias by means of a bid parameter. From Table 4, we included the starting point bid as the bid parameter for all respondents.We found evidence of a positive and statistically significant relationship between the bid parameter and the average WTP for the restoration of the protected forests.This implies that the starting point bias could have been more but for its control(Boyle et al.,1997;Amoah et al.,2019).In Table 4,three regression results with three thresholds ranging from Rr ≤100%to Rr ≤20%are presented.We acknowledge that similar result is obtained even after extending the threshold to Rr ≤10%.Although our interpretation of the results is based on the full sample,yet we find robust and consistent significant results across all sub-samples estimated (see Table 4).In line with consumer demand theory,income is considered as an important determinant of consumer or household demand.Consistent with theory, the coefficient of household income is positive and statistically significant at 1% level of significance. This means that if income should increase by 1%, WTP will increase by approximately Gh￠4($0.68)per annum.Thus,increase in income is associated with WTP for the restoration of the protected forests.Czajkowski et al.(2017)reported similar income-effect findings for forest management in Poland.

Age is used in this study as a proxy for experience.A priori,we expect age to have a positive effect on WTP for the restoration of the protected forests because experience with biodiversity or a protected forest with respiratory health and tourism benefits will relatively be associated with higher WTP. We found age to be negative and highly statistically significant across all samples. This implies that a one-year increase in a respondent's age decreases WTP by about Gh￠0.48($0.0816). This suggests that younger people are more willing to pay for forest restoration relative to older respondents.This could be explained by cohort effect of older respondents who may want the younger generation rather to pay for the restoration as they may live to enjoy the benefits.Our evidence is consistent with the findings of Alvarez and Larkin(2010).

The effect of gender disparity regarding sustainability of the protected forest is not assumed to be the same across gender.Indeed,there exist a gender gap in environmental views and opinions. So, we used female as the reference category and found the coefficient to be positive and statistically highly significant. This implies that, being a male relative to female, is associated with a WTP of approximately Gh￠12.88($2.19), i.e., males are inclined to pay more for the restoration of the protected forests as compared to the females.As argued by Longhi(2013)and Zelezny et al. (2000) women are more inclined to exhibit pro-environmental behaviours than men. Similarly, one would expect women to be more concerned about the environment than men,however,our results is counter intuitive yet corroborates with the findings of Kamri (2013), and Amoah and Addoah (2021). This is plausible in patriarchal households where the household's economic burden rests on the male. In such societies, the economic burden for the safety and joint environmental services is reposed in the male.

IOK has a positive and statistically highly significant coefficient,i.e.,respondents with IOK are associated with a WTP of approximately Gh￠9.67($1.64)as compared to those without knowledge.This is plausible given that international environmental issues dominate both the traditional and social media in Ghana. LOK, Edu (education) and MS (married) were found with their expected signs, albeit insignificant. As with the approach used for the restoration of the protected forests in Table 4,we controlled for starting point bias and found a positive and statistically significant relationship between the bid parameter and the average WTP for the maintenance of the protected forests in Table 5.Thus,controlling the bid parameter has lessened the effect of starting point bias on our results.It is important to acknowledge that the bid parameter's influence on the WTP for restoration is higher than that of WTP for maintenance.

Nonetheless, both bid parameters are not more than Gh￠0.23 ($0.04),hence we conclude almost non-existent starting point bias for both restoration and maintenance.

Table 4 Regression results for restoration.

Table 5 Regression results for maintenance.

Household income obtained the expected positive sign; however, it appears insignificant in the full sample and the sixty-percent sub-sample.In contrast, we found evidence of a positive and statistically significant relationship between income and WTP for the maintenance of the protected forest areas in the other relatively smaller sub-samples (40%and 20%). This implies that a smaller fraction of the respondents considers maintenance as partly their responsibility and not solely to the Municipal Assembly or the government.Stated differently,a greater fraction of the respondents possibly considers maintenance as the responsibility of the government or the Municipal Assembly as it is the responsibility of the government or the Municipal Assembly to maintain public goods. Thus,for most of the respondents, restoration may be seen as their shared responsibility.However,the same cannot be said for maintenance.

In line with the results for restoration, we found age to be negative and statistically significant across all samples. This implies that a oneyear increase in respondent's age decreases WTP for maintenance by Gh￠0.26($0.044).Indeed,older people are less willing to pay for forest maintenance relative to younger respondents. As earlier indicated, this behaviour can be attributed to cohort effect of the older age who may want the younger generation to pay for the maintenance of the protected forest areas. This is considered as a strategy of making the younger generation more responsible for sustainability of the forests.The WTP for restoration as influenced by age is about 83%higher than that of WTP for maintenance.

Similarly,the effect of differences in gender and WTP behaviour can be observed in the maintenance of the protected forest areas. Using female as the reference category,we have evidence of a positive coefficient which is statistically highly significant. This means that being a male relative to female, is associated with a WTP of approximately Gh￠7.47($1.27), i.e., females are relatively less inclined to pay for the maintenance of the protected forests as compared to their male counterparts.The WTP for restoration as influenced by gender is over 72%higher than that of WTP for maintenance.

Unlike our earlier results in Table 4 which found education to have a positive yet insignificant relation with WTP for restoration, the coefficient for education is now found to be positive and statistically significant.This means that education is key to driving behaviour and attitude towards ensuring sustainability of protected forests.This observation is,however, true only for the full sample and not the subsample. The subsample results for restoration converges with the subsamples for maintenance. The WTP value for restoration as influenced by education is about 93%lower than that of maintenance.

Similar to the results for restoration,the coefficient for IOK is found to be positive and statistically highly significant. Thus, respondents with IOK are associated with a WTP of approximately Gh￠5.14 as compared to those without knowledge.By way of comparison,respondents with IOK are WTP about 95% more for restoration as compared to maintenance.The WTP value for restoration as influenced by education is about 88%higher than that of maintenance.

Again,LOK is found to be positive and insignificant in the full sample.LOK not being statistically relevant to WTP is also found in Amoah and Moffatt (2021). However, it is found to be positive and marginally significant in all the sub-samples. The WTP value for restoration as influenced by education is over 700% lower than that of maintenance.Consistent with the results on WTP for restoration,being married has the right sign, albeit insignificant. Thus, being married is not found to be a statistically significant variable in making WTP decisions regarding restoration and maintenance of the protected forests. In line with insignificant socioeconomic factors,Carrasco-Diaz et al.(2019)found similar evidence.

Overall, the current study has shown that not all the drivers of WTP for restoration are drivers of maintenance.This observation is important as it seeks to help policy makers focus on determinants based on specific projects (i.e., be it restoration or maintenance). Lumping them together for policy purposes may not yield the expected results as the drivers for each project are found to be unique.

WTP and percentage share of household income: In line with the theory of consumer demand, the expected mean WTP is the predicted amount based on the WTP determinants.In Table 6,for the full sample,we can argue that respondents in the GAR are willing to pay Gh￠50.99($8.67) per annum for the restoration of the protected forests. The amount should be an addition to the current property rights being paid by households or property owners.This constitutes 0.39%of a household's annual income and can be depicted as affordable. Now, assuming that each household is willing to contribute for the costs of restoration and forest management through property rates, then the total WTP for the entire region will be Gh￠52,847,361.70 ($8,984,051.49) per annum.With reference to specific forest areas, the WTP of respondents who reside near the Achimota forest,Gua Kuo forest and Shai Hills Forest are Gh￠54.74 ($9.31), Gh￠53.07 ($9.02) and Gh￠44.93 ($7.85), respectively.These sums constitute reasonable or affordable(＜0.5%)shares of their respective annual incomes.

Also,respondents are willing to pay Gh￠31.12($5.29)per annum for the maintenance of the protected forests. This constitutes 0.24% of a household's annual income. The entire region's total WTP for maintenance is Gh￠32,253,577.10 ($5,483,108.11) per annum. The WTP for maintenance per forest ranges between Gh￠25($4.25)to Gh￠35($5.95)which also considered as affordable(＜0.3%).

5. Discussion

Based on the results of this study, we show that, on average, households have exhibited a positive behaviour towards the restoration and maintenance of reserved forests and forest resources in the GAMA of Ghana. In the full sample, we show that the expected mean WTP for restoration of reserved forest and forest resources is Gh￠50.99 ($8.67)per annum which constitute 0.39% of household monthly income.Similarly,the expected mean WTP for maintenance of reserved forest and forest resources is Gh￠31.12($5.29)per annum which constitute 0.24%of household monthly income. According to previous WTP studies such as Whittington et al. (1990), McPhail (1993) and Amoah (2017), given that the WTP estimates as a share of household's monthly income is less than 5%, we conclude that the estimates of less than 0.5% for both restoration and maintenance fall within a reasonable range of affordability.Although,the estimates are not the same across forests,we posit that all estimates fall within our definition of affordable.By implication,these estimates do not present an unbearable situation to households in the study area. Rather it presents an opportunity for resources to be harnessed towards the restoration and maintenance of the reserved forests.

In line with economic theory,marginal benefit of forest consumption declines over time as its consumption increases. Generally, restoration precedes maintenance hence its marginal benefit should be expected to be greater than its maintenance. It stands to reason that for restoration,WTP can only be as large as the time the forest and forest resources are restored.However,that cannot be the case for maintenance which is not time or attainment bound. For restoration, we expected a one-off payment which is likely to be larger than the recurring maintenance spending. Impliedly, we expected WTP for restoration to be larger than maintenance which we find. In the case of Loomis et al. (1991), they found that for wetlands and contamination, WTP for improvements/restoration was always higher than the corresponding WTP for maintenance. Further, we show that WTP for restoration is approximately 64% greater than WTP for maintenance. The positive attitude towards restoration and/or maintenance is not new as several other studies (Amoah, 2011; Mueller, 2014) have found corroborating findings.

Overall, the WTP estimates are determined by socioeconomic and behavioural factors of the respondent. For restoration, the statisticallysignificant factors in the full sample include income, age, gender and international objective knowledge. For maintenance, the statistically significant factors in the full sample include age, gender, education, international objective knowledge, and local objective knowledge. The determinants in the sub-samples for both restoration and maintenance are largely consistent. Several empirical studies (example, Czajkowski et al., 2017; Amoah et al., 2019; Endalew and Wondimagegnhu, 2019;Amoah and Moffatt, 2021) have highlighted the relevance of socioeconomic and behavioural factors in determining WTP for environmental goods and services.

Table 6 WTP and share of household income by forest.

6. Conclusions

This study presents arguments from the perspective of the literature to support the claim that reserved forests have experienced depletion over the years at an alarming rate in Ghana. This problem is shown to emanate from institutional failure, climate change, and behavioural factors (Amoah and Korle, 2020). Towards a demand-side management approach,this study relies on a well-validated CV method and finds that a household is willing to pay Gh￠50.99($8.67)and Gh￠31.12($5.29)per annum for the restoration and maintenance of the forests, respectively.These amounts are reckoned to be very reasonable as they fall below 0.5% of a household's annual mean take-home income. The primary payment vehicle for realizing these amounts is through the property rate.The evidence is very informative to the Forestry Commission and the Municipalities in determining the extent to which the property rate can be raised toward protected forest sustainability. Admittedly, this study addresses the demand-side with no attempt at the supply-side due to unavailability of data. Further research may consider providing supply-side evidence to facilitate a cost and benefit analysis. By this,policy makers will be in a better position to determine the economic viability of restoring and maintaining the protected forests in the GAR and other parts of Ghana. With the “affordable” nature of our estimate and the low property rates in Ghana,we are confident that policy makers will embrace the recommendations towards the restoration and maintenance of forest and forest resources in the GAMA.

Funding

This work received financial support from the Global Greengrants Fund, UK/Europe/USA(Grant Number:2018-2472).

Availability of data and materials

All data used are available upon request.

Authors’contributions

Anthony Amoah conceived the idea, participated in questionnaire design, data collection, wrote the methodology, analysed the data, discussed the findings and reviewed the final manuscript. Adusei Jumah wrote the introduction and reviewed the final manuscript. Kofi Korle participated in questionnaire design,data collection,wrote the literature and reviewed the final manuscript.

All authors contributed to the manuscript writing and gave final approval for publication.

Ethics approval and consent to participate

This is a non-experimental study which satisfies all ethical requirements. Again, all the authors have approved the manuscript and agreed with the submission.

Consent for publication

Not applicable.

Declaration of competing interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Acknowledgements

Much appreciation also goes to Global Green Grants for their financial support.The authors also appreciate Henry Akpolu(Dept.of Economics,Central University) for his technical assistance. Lastly, we remain thankful to EFD-Ghana(ENRRI)for their kind assistance.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://do i.org/10.1016/j.fecs.2022.100041.