2.3.1. Descriptive data

Table 1 provides descriptive information about the household foresters in Yen Bai Province, Vietnam (Nghiem, Reference Nghiem2013). The average age of respondents was 46 years, average level of education was 7.6 on a scale up to 12 (with 12 being final year of high school), and almost all respondents (85 per cent) were male. Nearly half (47 per cent) of these foresters belonged to minority groups, namely Tay, Cao Lan, Dao and Nung, with the rest being from the Kinh group. The average household size was 4.84 and over half of the adult family members were labourers. Eighty per cent of the respondents were classed as ‘middle-class income’ (including all sources of income) according to government standards. That is, the income level was higher than 2.4 million VND or US$141.18/person/year on average. About 6 per cent were classified as poor and the rest as relatively ‘well-off’ (Government of Vietnam, 2005).

Table 1. Descriptive statistics about the sample of household foresters in Yen Bai Province, Vietnam (n = 271)

With regard to production information, table 2 shows that the actual harvesting age of household-planted forests was 5.18 years. The histogram results (figure 1) suggest a similar distribution for harvest age for the two species (at 5.21, S.D. = 1.27 and at 5.17, S.D. = 1.52 years for E. urophylla and A. mangium, respectively). Interestingly, revenue per ha from agricultural activities was 14 times higher than for forestry. These findings were consistent with the foresters' desire, noted during household interviews, to shift more of their land to non-forestry agricultural use. The data also suggested that almost all households (99 per cent; 267/271) effectively ‘own’ their forest land, which means they have a government lease. Ninety-one per cent of the owners applied a clear-cut practice to harvest their forests. Among the 271 respondents, 32 (12 per cent) grew E. urophylla only, 78 (29 per cent) grew A. mangium only, and 161 (59 per cent) grew both species.

Figure 1. Distribution of harvest age (years) for E. urophylla grown in household foresters' planted forests in Yen Bai Province, Vietnam

Table 2. Production information for the household foresters in Yen Bai Province, Vietnam

The results of the Wilcoxon–Mann–Whitney U-tests suggested that there was a statistically significant difference in: the number of forest stands with regard to ethnicity; the number of forest stands with regard to financial status; planted forest revenue with regard to financial status; and planted forest area with regard to financial status. These results were as expected. For example, families belonging to the Kinh group normally migrate from other regions and tend to participate in activities outside the forestry sector, so it was not unexpected that they had fewer forest stands compared to those of people in other ethnic groups. It is also not surprising that wealthier families had more planted forest area and earn more compared to poorer families. The other forest-related variables were not significantly statistically different with regard to household size or other respondent characteristics.

Further analysis of household characteristics with regard to forest management using a district fixed effects regression suggested that financial status, sex, ethnicity and age were statistically correlated with forest management variables (table 3). Moreover, financial status was negatively correlated with forest management variables at 1 per cent significance. That means that the more well-off families had larger planted forest area, more forest stands and higher income from forests. However, household size and education consistently showed no relationship with forest management dependent variables. Furthermore, the signs of coefficients and their magnitudes were reasonable. The standard error values suggested that the coefficients were quite stable. Results from the simple linear regression were quite similar in terms of coefficient signs and magnitudes compared to those from the district fixed effects analysis.

Table 3. Relationships between forest management practices and household characteristics in Yen Bai Province, Vietnam, using a district fixed effects model

Notes: *Significant at 10% level; **significant at 5% level; ***significant at 1% level.

^a Harvesting age; ^b no. forest stands; ^c revenue from planted forests; ^d planted forest area.

2.3.2. Planting costs and timber prices

To estimate the cost functions for the two species, planting costs obtained from the survey were adjusted by inflation rates (World Bank, 2008) to 2007 values. The average planting cost of E. urophylla per ha was 6.85 million Vietnamese dong (VND) (US$402.94) ha⁻¹, and 0.07 million VND (US$4.12) higher than that of A. mangium. The majority of the foresters sold stumpage trees at an average stumpage timber price of 0.37 (21.76) and 0.33 (19.41) million VND (US$) cubic meter⁻¹ for E. urophylla and A. mangium.

2.3.3. Forest management for timber production

The survey results suggested that about 73.4 per cent of households used their own capital to invest in their forests (see table A1 in the online appendix, available at http://journals.cambridge.org/EDE), so lower interest rates from bank loans, a policy currently used by the Government of Vietnam to encourage investment in planted forests, may not be relevant to the majority of household foresters (Government of Vietnam, 2009). The most likely reason is that risk aversion makes foresters avoid borrowing money from banks. Family financial status also played an important role in the decisions of household foresters regarding tree cutting, tree species and compensation levels for delaying harvest. A total of 93.4 per cent of foresters stated that they would delay their harvest if they were financially supported by the government with an average amount of 5.5 million VND (US$323.53) ha⁻¹ year⁻¹.

2.3.4. Participating in a carbon pooling project

In the survey, foresters were also asked about their willingness to participate in a carbon pooling arrangement. Under such an arrangement, the land of several households is bundled to form a larger project size. The households must act together regarding types of dominant tree species, harvest decisions and benefit sharing mechanisms. Besides the income from selling the timber, the participants would be paid for carbon sequestration. A pooling arrangement would help to reduce transactions costs and could take advantage of economies of scale (Grieg-Gran et al., Reference Grieg-Gran, Porras and Wunder2005; Gong et al., Reference Gong, Bull and Baylis2010).

The results suggested that the majority of households (89 per cent) would agree to participate in such a carbon pooling arrangement (see online appendix, table A2). Regarding the potential contractual arrangements involved, nearly half of the respondents (47.3 per cent) said that a contract between households and the investor would be necessary. When asked if they could foresee any obstacles to a carbon pooling agreement, nearly half (42.4 per cent) had no view about whether or not obstacles would occur. This response may have been because carbon sequestration and the CDM are relatively new concepts to household foresters in Vietnam.

The results of the Wilcoxon–Mann–Whitney U-tests suggested that there was a statistically significant difference in the mean of carbon pooling agreement with regard to age of the respondents, and in respect of the reasons for being involved in a carbon pooling project with regard to householder ethnicity, education and financial status. Other household characteristics showed no significant difference in the mean of these variables.

Further analysis of household characteristics with regard to carbon sequestration management using district fixed effects regression suggested that age, ethnicity and household size also significantly correlated with carbon payment method (table 4). Ethnicity was also correlated with ‘reason for being involved in a carbon pooling project’, but not financial status and education as suggested by the mean test as above. This suggests that these two variables had no impact on carbon sequestration management at the margin. In addition, education was correlated to the carbon payment method.

Table 4. Relationships between carbon sequestration management and household characteristics in Yen Bai Province, Vietnam, using a district fixed effects model

Notes: *Significant at 10% level; **significant at 5% level; ***significant at 1% level.

^a Carbon payment method; ^b reason for involving in a carbon pooling project; ^c carbon pooling project agreement.

2.3.5. Payments for carbon sequestration

For the annual payment scheme, it was proposed that foresters would be given annual carbon service payments from forest canopy closure or from year four onwards until forest harvesting. For the upfront payment scheme, foresters would be paid for carbon services in full at forest establishment and they would have to sign a pledge to keep their forests until they reached a certain age. The two proposed carbon payment schemes, annual vs. upfront payments, were favoured by the respondents, at 51.7 and 48.3 per cent respectively. The survey results indicated that foresters would want an average amount of 21.4 million VND (US$1,258.8) ha⁻¹ rotation⁻¹ from carbon sequestration benefits for them to keep their trees until age 11 years on average.

An analysis of the characteristics of household foresters in the annual and upfront payment groups suggested that households who chose the annual payment tended to belong to a low-income group, were more dependent on income from planted forests, were more concerned about money for their living and were less interested in a carbon pooling project. With regard to income levels, the data indicated that households favouring the annual payment were characterized by earning less income in total compared to those of the upfront payment group (by 3.7 million VND (US$246.7) per year). More than half of the respondents (56.1 per cent) who belonged to minority communities, who were recognized as poorer in comparison to the Kinh group, also favoured annual payments.

Furthermore, the annual payment group obtained more of their income from productive forests and less income from other sources, even though they provided more outside labour, such as work for other individuals or organizations. This is because the outside work performed by the annual payment group was low skilled, and thus they were paid less compared to the upfront group. This result followed from the fact that the level of education in the annual payment group, on average, was lower than that in the upfront payment group (7.34 and 7.95 over the scale of 12, respectively). The data implied that the annual payment households were more dependent on planted forests, and that this was probably the major reason why they favoured annual payments (i.e., to secure a constant source of income). The reasons given for harvesting decisions made by households also clearly support this hypothesis. Around one-third (31 per cent) of households who preferred payment annually stated that they cut their trees because of financial reasons, i.e., lack of money for their everyday basic needs, compared to 24 per cent in the other group.

The next characteristic of household foresters who favoured annual payments was that they seemed to be more worried about money for living than the other group. In particular, the majority of the respondents would agree to delay harvest if they were subsidized by the government, especially those in the annual payment group. When asked why, 47 per cent of the annual group related their answers to financial ability, living costs and debts, compared to 42 per cent of the upfront payment group. Again, in the expression of household foresters' opinions about the carbon pooling project, the respondents who preferred the annual payment were more concerned about capital and money (10 per cent) than those who chose the upfront payment (4.6 per cent).

Household foresters in the upfront payment group seemed to be more optimistic than those in the other group. This was also suggested by this group being more willing to engage in the carbon pooling project compared to the other group. Furthermore, when asked if they anticipated any difficulties in the implementation of the carbon pooling project, 30 per cent of the respondents in the upfront payment group believed there would be no obstacles while only 18 per cent of those in the other group had that belief.

In conclusion, household foresters who preferred the annual payment seemed to be poorer, more financially vulnerable (that is, more dependent on planted forest income) and less sure about the success of the carbon pooling project. It was likely that this group did not choose the upfront payment with a commitment because they were concerned that the commitment might bring inflexibility in times when they needed to sell their trees for everyday basic needs or living costs. However, these results may be influenced by bias, given that most respondents were male (85 per cent), reflecting the more dominant role played by men in the financial decisions of a family in this country. Even though the survey requested that the respondents answer on behalf of their families, it was not possible to ascertain to what extent their answers represented their personal view only (see Nghiem, Reference Nghiem2013 for more discussion about survey bias).

3.1.1. The single stand forest model

The objective of the single stand forest model was to maximize the net present value (NPV) from harvesting timber and sequestering carbon. At the end of the planning horizon, foresters were supposed to harvest the whole forests. In this study, the NPV did not include land rents since once the land is used in forestry, it cannot be rented. So renting is another investment option instead of growing trees for the land. Instead of being calculated over the infinite chain of cycles of planting on the given forest stand as in the Faustmann formulae, the NPV was calculated over 50 years with a one-year time step as in the mathematical model below.

Let v(.) be the discounted sum of the timber value (V _t ) and the carbon sequestration value (A _t ).

$$\eqalign{& \quad x_{t}\colon \hbox{the age of the stand at period}\, \, t. \cr &t\colon \hbox{time period with a step of one year}.}$$

The model objective was to maximize the discounted revenue from timber and carbon sequestration (Nghiem, Reference Nghiem2011, Reference Nghiem2013, Reference Nghiem2014):

(1) $$v\lpar x_t\rpar = max\sum_{t=1}^{50} \lpar 1+r\rpar ^{-t} \lpar V_t + A_t\rpar$$

subject to:

(2) $$x_{t+1} = \lpar x_{t} + 1\rpar \comma \; \hbox{ age of the stand at period }t + 1$$

(3) $$x_{t} \ge 0\comma \; \hbox{non-negative age constraint}$$

where:

(4) $$r \quad \quad \quad \hbox{discount rate}$$

(5) $$V_t = q\lpar x_t\rpar . P-G \quad \hbox{timber value at the period}\, \, t $$

(6) $$q\lpar x_{t}\rpar \quad \quad \quad \hbox{timber volume ha}^{-1} \hbox{of stand in period}\, \, t $$

(7) $$P \quad \quad \quad \hbox{price of timber cubic meter}^{-1} $$

(8) $$G \quad \quad \quad \hbox{initial planting cost of timber} $$

(9) $$A_t = Q_c \lpar x_t\rpar . P_c \quad \quad \hbox{value of carbon sequestered at period}\, \, t $$

(10) $$Q_{{\rm c}} \quad \quad \quad \hbox{amount of carbon sequestered tonne ha}^{-1} $$

(11) $$P_{{\rm c}} \quad \quad \quad \hbox{carbon price tonne}^{-1} $$

Functions and parameters used in the model are presented in the online appendix, table A3. The model was coded in GAMS 2010.

3.1.2. Timber growth function

Timber growth functions of E. urophylla and A. mangium were estimated from reliable and experimental data obtained from a government source (Vietnam Ministry of Agriculture and Rural Development, 2003). A quadratic function provided the best fit functional form (see online appendix, table A3). For short rotation ages, a power timber growth function was used to predict timber volume in the literature (Ministry of Forestry, 1996). Furthermore, as shown in figure 1, rotation ages for household-planted forests in reality could be as short as two years; thus it is appropriate to use this functional form.

3.1.3. Carbon sequestration function

Carbon sequestration functions, which were estimated based on reliable experimental data for E. urophylla and A. mangium in planted forests by Vo et al. (Reference Vo, Dang, Nguyen, Nguyen and Dang2009), were used in the model. Since Vo et al.'s functions were not suitable for application to planted forests beyond seven years of age, their carbon sequestration functions were adjusted as follows. First, using stand age and tree density data from the Tables of Site Index (Vietnam Ministry of Agriculture and Rural Development, 2003) and tree density at planting from the survey, a relationship was calculated between stand age and tree density. Secondly, the amount of carbon uptake was obtained by running the carbon sequestration functions from Vo et al. with respect to stand age and tree density for E. urophylla and A. mangium. Finally, the amount of carbon uptake and the associated stand age were used to re-estimate the carbon sequestration functions for use in the optimization models.

3.2.1. The optimal rotation age

This section reports the optimal rotation ages for timber only (the timber rotation age) and carbon sequestration models (the carbon rotation age) for both E. urophylla and A. mangium. According to the results in table 5, at the chosen discount rate, the optimal timber rotation age for E. urophylla was nine years. Adding carbon benefits had no effect on the optimal rotation age for E. urophylla. This is because both the rates of carbon uptake and of timber growth were faster in early years; however, the latter rate was much faster, making the optimal rotation age remain the same (i.e., the carbon sequestration rate is low so that it does not change the optimal rotation age).

The optimal timber rotation age for A. mangium was 11 years at the chosen discount rate. When carbon sequestration had a price of US$3 tonne⁻¹, the optimal carbon rotation age for A. mangium was shorter by two years. This is because of the fact that higher timber productivity outweighed the effect of carbon sequestration on the optimal rotation age for enterprise planted forests. In terms of NPV, the carbon rotations all had higher NPVs as expected, with carbon having a value (table 5).

The finding that the optimal carbon rotation was shorter than the timber-only rotation (more so for A. mangium than for E. urophylla) was in contrast to the results of van Kooten et al. (Reference van Kooten, Binkley and Delcourt1995) and Gutrich and Howarth (Reference Gutrich and Howarth2007). This is probably because of the nature of fast-growing trees in this tropical country study, where carbon uptake and timber growth are greater in the early years compared to the slow-growing tree species in the analyses in the previous two studies. However, it should be noted that the results of van Kooten et al. and this study's results are not easily comparable since forest owners here did not have to pay a tax when carbon was released at the time of harvesting as in the research by van Kooten et al.

3.2.2. Sensitivity analysis to discount rate

The results of a sensitivity analysis of the optimal rotation age to discount rates are shown in Table 5. At positive discount rates, the optimal rotation length for E. urophylla ranged from eight to 10 years. There was hardly any difference between the two except at discount rates 5 per cent and higher, when the optimal rotation age for carbon becomes slightly shorter than the optimal timber rotation age. For A. mangium, the optimal timber rotation age varied from nine to 17 years depending on the discount rates and with the carbon optimal rotation being shorter by several years, especially at lower discount rates (1–5 per cent). Both the timber and carbon optimal rotation ages decreased with an increasing discount rate.

Table 5. The optimal rotation ages for timber only and carbon sequestration values for household-planted forests in Yen Bai Province, Vietnam

Notes: ^a Optimal rotation age; ^b net present value; ^c mean for all discount rates from 1% to 10%, some of which were not reported here.

Values in bold were the baseline scenario. The results were based on the assumptions of a carbon price (via the CDM or equivalent) equal to 0.051 million VND/m³ or US$3/m³, and a timber price of 0.37 million VND (US$21.76)/m³ for E. urophylla and 0.33 million VND (US$19.41)/m³ for A. mangium.

Given that the actual harvesting age for households was five years, the optimal results from a societal or government perspective indicated the value of longer rotations at all discount rates. This result suggests that policy intervention is potentially desirable to bring the actual harvesting age in line with the optimal rotation age.

3.2.3. Sensitivity analysis to carbon price

The carbon price varied from US$3 to US$10 (equivalent to 0.051–0.17 million VND) tonne⁻¹. In the case of E. urophylla (see online appendix, figure A1), the optimal carbon rotation age decreased slightly with an increasing carbon price, from nine to eight years. The NPVs for carbon rotation age increased with an increasing carbon price as expected. A. mangium experienced a similar relationship between the carbon rotation age and the carbon price to that found with E. urophylla. The households' rotation age for A. mangium was very sensitive to carbon price, decreasing from nine to three years with an increasing carbon price, suggesting that carbon sequestration benefits had a significant impact on the optimal rotation age. The rotation age for E. urophylla was not as sensitive to carbon price as that for A. mangium because of the difference in parameters in the timber growth and the carbon sequestration functions for these species. The NPVs for carbon rotation age for both species increased with an increasing carbon price as expected.

The relationship found between rotation length and carbon price was in contrast to the findings of van Kooten et al. (Reference van Kooten, Binkley and Delcourt1995), but was in agreement with the results of Englin and Callaway (Reference Englin and Callaway1993) with amenity value. Diaz-Balteiro and Rodriguez (Reference Diaz-Balteiro and Rodriguez2006) found that the carbon price level has little effect with no clear pattern on the rotation length.

3.2.4. Sensitivity analysis to carbon payment scheme

The impacts of various carbon payment schemes on the optimal management strategy suggested that carbon value, with a carbon price of 0.051 million VND m⁻³or US$3 m⁻³, could be paid to foresters at the beginning of the rotation. To be eligible to receive the carbon payment at the start of the rotation, foresters would have to commit to keeping the trees until the end of the rotation. Alternatively, the carbon payment could be paid every year, or at the end of the rotation. The NPV values in these two payment schemes are comparable since the difference in discounting time periods of these schemes was incorporated into the optimization model.

Sensitivity analyses also suggested that payment schemes had no significant impact on the optimal rotation age. In particular, the optimal rotation age remained the same at nine years in all payment schemes for E. urophylla. The optimal rotation age for A. mangium was nine years in the ending payment scheme, and 10 years for the other two schemes. However, the payment schemes had an impact on the NPV. A lump-sum payment at the beginning of the rotation increased the NPV slightly compared to annual payment and payment at the end of the rotation.

3.2.5. Sensitivity analysis to the structure of timber price

Harvest size of timber was assumed to be positively correlated to timber age based on the preliminary fieldwork of this survey in Yen Bai Province. Let harvest age and the price of timber aged seven be T and P, respectively. Then, P is 0.37 million VND (US$21.72) m⁻³ for E. urophylla and 0.33 million VND (US$19.41) m⁻³ for A. mangium. Let the price of timber, which is harvested at any age T, be P _t as follows:

$$P_t=\left\{{ {\matrix{P-0.15\colon for T \lt 5 \cr P-0.05\lpar 7-T\rpar \colon for 5 \le T \le 6 \cr P\colon for T=7 \cr P+0.1\lpar T-7\rpar \colon for 7 \lt T \le 14 \cr P+0.8\colon for T \gt 14\cr } } } \right.$$

The results suggested that a varying timber price made the optimal rotation ages significantly longer compared to those of a constant price. In particular, the varying timber price (P) lengthened the rotation ages to 14 years (NPV 14.94 million/878.82 VND/US$ ha⁻¹) for E. urophylla and 15 years (NPV 59.24 million/3,484.71 VND/US$ ha⁻¹) for A. mangium, which were significantly longer compared to those of the constant price, 9 and 11 years, respectively. The varying timber price was a more realistic assumption compared to the constant price; however, it was not applied to the whole analysis because of great uncertainty in the timber price functions.