Using Tai Po as a case study, is clustering move evident in an older of newer shopping area?

Tai Po has two main shopping districts, the old and new town center which are approximately 0.5 km apart and is divided by the Lam Tsuen River (figure 2). The older district south of the river is known as Tai Po Hui (Market) consists of a series of small alleys with low rise buildings consisting of shops on ground levels and apartments above. The rectangular area (highlighted in blue in Figure 2) will be the area used for data collection in the old district.

The newer district north of the river consists of a series of shopping complexes joined together via overhead bridges known as Tai Po Mega Mall. The "mega-mall" can be divided into different sectors A-E, of which only Sectors A, B and C will be used for the purpose of this investigation as the amount of data from these 3 malls should be sufficient.

Hypothesis

I predict that a newer shopping district (Tai Po Center) will have a higher degree of clustering compared to an older district (Tai Po Market).

Shopping centers tend have a high degree of clustering and also a large range of retail types. This is because an indoor shopping mall such as Tai Po Mega Mall will likely be zoned according to retail types for the benefit of customers and also for management. For example, a certain floor in the center may be designated for a certain retail type (such as food). This leads to more competition as shops of similar retail types group can together for maximum profit. Shopping centers also tend to have a wider range of retail types as this attracts the large trade area needed for specialized retail types.

In comparison, older districts have less clustering as shops tend to sell low order goods. Since this has a smaller profit turnover, shops in such areas tend to face less competition compared to newer shopping districts. Since shops in older districts tend to be well established with a loyal customer base, there is little room for similar retail types in close proximity, leading to less clustering. Since the rent is comparatively lower, retail outlets tend to be bigger and cover a wider range of products (such as a general household store); it will create less need for other services and reduce competition.

Methodology

For data collection, an accurate outline map of the areas to be surveyed will be used. A key will be created to distinguish each of the 7 different retail types below:

> Hardware / Household (computer, household, kitchen)

> Services (Agencies, money exchange, insurance, banks etc)

> Recreation (toy shops, hobbies, sports, cosmetics etc)

> Convenience (corner shops, 7-11)

> Restaurants

> Food / Greengrocers (including supermarkets)

> Fashion and Accessories

As this data will be collected for other projects by other students, any other retail types will be ignored as I believe the 7 types above give a sufficient scope of data for analysis.

For the old shopping district, the inner car park areas will be ignored as the area are mainly automobile maintenance stations and not part of the shopping area, therefore it will not be included.

Similarly for the new area, only the area within the two floors of the shopping center will be surveyed. Shops outside by the street are ignored as they are not part of the mall.

Once the data is collected, the nearest neighbor of each shop in each category will be measured on the map. This is done by taking a shop entrance and measuring the closet walking distance to the nearest shop entrance of the same retail type. If the nearest neighbor happens to be on a different level (such as in the mall), the walking distance to the nearest escalator will be taken plus an additional 10 meters for going up.

Once the raw data is collected, the mean distance for each retail type and the area of both regions shall be calculated and substituted into the NNI (Nearest Neighbor Index) for analysis.

The NNI measures spatial dispersions of points (such as settlements) in a given area. These points are said to be either clustered or randomly / regularly dispersed.

Clustered Random Dispersion Regular Dispersion

The NNI Index is calculated as follows:

2D

?

N

A

Key:

N = Number of points

A = Sampling area

D = Mean between each point and its nearest neighbor

This is calculated by finding (?d)/N where d = each individual distance.

NNI values can be compared with the computer generated values to categorize the spatial dispersion:

If the NNI is:

~0 = Clustered

~1 = Random

>2 = Regular