The Collocation of Industries in Agriculture-Food-Tourism in Vietnam

Abstract

Gokan examines the collocation of agriculture-food-tourism industry using the clusters identified in the process of deriving Mori and Smith's index. Collocation is evident around Ha Lon Bay, in the rural cities of northern Vietnam, in some coastal cities in the middle of Vietnam, around rural cities some distance from Ho Chi Minh City, in regions near Ho Chi Minh City and in some regions in the Mekong Delta. Because rural cities already have clusters related to agriculture-food-tourism industry, strengthening these links will help to foster further such clusters. In order to connect industries, the success stories and case studies introduced in this chapter may be helpful.

Appendices

Appendix A: Deriving the Global Extent and Local Density

The process of deriving the GE and LD can be divided into three steps: detecting the clusters, determining the essential containment and then calculating the GE and LD.

The first step is to find a cluster scheme ${\mathbb{C}}^{*}$ with a maximum value for the BIC among the candidate cluster schemes. Cluster scheme $\mathbb{C}$ is simply one or more disjoint clusters, C _j , for j =1 ,  …  , k, and the residual set of all non-cluster regions. The BIC increases with a larger log-likelihood of ${\widehat{P}}_{\mathbb{C}}$, the location probability of cluster scheme $\mathbb{C}$ for each basic region, given an observed location pattern x. However, the BIC decreases with the penalty term composed by ${k}_{\mathbb{C}}$, which expresses the number of clusters in cluster scheme $\mathbb{C}$, and n, which expresses the number of establishments in the total area, such that ${BIC}_{\mathbb{C}}={L}_{\mathbb{C}}\left({\widehat{P}}_{\mathbb{C}}|x\right)-\frac{k_{\mathbb{C}}}{2} \ln n$. The location probability of cluster scheme $\mathbb{C}$for cluster j=1 ,  …  , k, ${p}_{\mathbb{C}}(j)$ can be rewritten as ${\widehat{p}}_{\mathbb{C}}(j)={n}_j(x)/n$, where n _j (x) expresses the sum of the number of establishments in cluster j=1 ,  …  , k and $n\textcolor[rgb]{1,0,1}{}$expresses the total number of establishments in all regions. In the above log-likelihood functions, n _j (x) and n _r are related with a sequence of independent location decisions by individual establishments. Each region is included as part of a cluster C _j , j=1 ,  …  , k, or the residual set of all non-cluster regions. Thus the right hand-side of the log-likelihood function of the location probabilities (i.e. the probability that a randomly sampled establishment is located in a region within a certain cluster) expresses the law of total probability. Here the log-likelihood function can be divided into two parts. The first term gives the location probabilities of n _j establishments being located in a cluster C _j , j=1 ,  …  , k. The second term gives the location probability that n _r establishments are located in region r in cluster C _j , j=1 ,  …  , k, given that individual establishments choose their location randomly within each cluster, such that

$$ {L}_{\mathrm{\mathbb{C}}}\left({\widehat{P}}_{\mathrm{\mathbb{C}}}|x\right)={\displaystyle {\sum}_{j=0}^{k_{\mathrm{\mathbb{C}}}}{n}_j(x) \ln {\widehat{p}}_{\mathrm{\mathbb{C}}}(j)+{\displaystyle {\sum}_{j=0}^{k_{\mathrm{\mathbb{C}}}}{\displaystyle {\sum}_{r\varepsilon {C}_j}{n}_rln\frac{a_r}{a{C}_j}}}}, $$

where a _r expresses the economic area in region r and ${a}_{C_j}$ expresses the economic area in cluster C _j , j=1 ,  …  , k. A new cluster is formed and compared with other possible modifications of the current cluster scheme in order to find the highest value of the BIC. In the process of expanding an existing cluster by adding regions consisting of the d-convex solid of the core part only, the shortest path distance is used. The “core” part is used in the analysis in Sect. 4, whereas the “core” and “member” parts are used in Sect. 3. For more detail on “core” and “member” parts, see Mori (2014).

The second step is divided into two parts. The first determines the essential clusters, which are the most significant in terms of incremental contributions to the BIC given that the sum of the incremental contributions to the BIC exceeds a certain proportion (λ) of ${BIC}_{\mathbb{C}}$. Following the recommendation of Mori and Smith (2014), the value of λ is set to 0.91. The second part of the second step determines the smallest convex-solid set in the total area containing the set of essential clusters. Intuitively, a convex-solid set means that the set is connected and has no dents in its perimeter and no internal cavities. Then the regions in the smallest convex-solid set can be regarded as an essential containment.

In the last step we can obtain the GE for an industry by dividing the total economic area of essential containment for an industry by the total economic area of the whole country. Similarly, we can obtain the LD for the industry by dividing the total economic area of the essential clusters of the industry by the total economic area of the essential containment.

Finally, we test whether the BIC of the best cluster scheme is significantly better than that under a random location pattern generated by a Monte Carlo test. Among all industry studies, the null hypothesis of complete spatial randomness was rejected, with p-values of virtually zero.

Appendix B: Data for Analysis

“Basic regions” are taken to be district-level divisions, which are subsubdivisions of Vietnam. Considering the consistency within the regions in the industrial census data of 2011 and after removing islands, we focus on 649 regions. “Economic regions” are calculated by subtracting the area of forests, lakes, marshes and undeveloped areas from the total area of the region using a satellite image in Global Land Cover 2000. The travel distance between each pair of neighboring regions is calculated with a shape file on road networks. The shortest-path distances are computed as in Mori and Smith (2014). The employment data used in this analysis are based on the VSIC of 2007. The number of employees across 79 industries, which I described earlier, is taken from the 2012 establishment census of Vietnam.