Saturday, 25 May 2013

PRU13 数据分析 第一章 - 是华人海啸,还是城市海啸?

第13届马来西亚大选闭幕以来,很多人都在争论,造成国阵蒙受其有始以来最差的选举成绩,到底是华人海啸还是城市的海啸?

辩论越来越不明,一些国阵的支持者开始发表针对华社的种族言论。 民联支持者声称,实际上是多元种族的城市选民投了反对票给BN。 在他们看来,这是一个城市海啸。 我希望,通过严格的选举数据研究,我可以让这个争论有一个结论。 

选举成绩数据以及人口统计数据都是从SPR和RTM网站获得的: 
其他的数据源,在文章都有写。 

那只鸡指责华人海啸造成BN有史以来最糟糕的大选结果。 他说得对吗? 恐怕是 


在上图中,每个数据点代表一个国会选区。  
y轴是国阵票的百分比= BN票×100 /总票数。 
(在BN没有竞选的选区,我以亲BN的独立人士作为BN候选人。) 
x轴是%该选区华人选民的百分比。 

 我们可以清楚地看到,BN在华人多数区的得票,巨大的下滑。 这种下滑趋势和上届大选成绩比较已经明显增加。 

另一方面,如果我们看一下马来选民 


我们看到马来人的BN选票比上届来说没有减少。 如上图中所示,非马来人占多数的地区%BN票的下降可能是由于华人票的损失, 所以,总体来说马来人的支持率一直没有实际变化。虽然民联能够靠原有马来选民的继续支持,拿下或持有他们的选区,但是如果没有更多的马来选票,想要入住布城,成立联邦政府是不可能的。

让我们来看看印裔选民。 


我们看到一些迹象表明,印度人选民投了反对票给BN。 由于印度人选民在数量上是少数的,我们在这里看到的趋势,可能是由于马来人和华人投票趋势的影响。 但是,我们可以肯定,印度人的选票没有回流BN。 

民联领导人以及一些独立分析员一​​直说有“城市海啸”,正确吗? 



上图显示同一个“都会指数”的所有选区的平均%BN票。 都会指数被定义为以选区有关的城镇的总人口的对数。 
(马来西亚城镇的人口数据是从 http://www.world-gazetteer.com/wg.php?x=1171193898&men=gcis&lng=en&des=gamelan&dat=32&geo=-152&srt=pnan&col=ohdq&pt=c&va=x 得来) 
都会指数大约把选区区分为: 

2,3:农村 
4:小城市 
5,6:城市 

新山和吉隆坡都会指数6。 在砂拉越诗巫帝这种偏远地方,都会指数2。 

在条形图中,我们可以看到两个大选都有城乡鸿沟 2,3,4类的选民投票给BN,5,6选民投票给反对党。 
然而,在各都会指数都可以看到BN支持率普遍下滑。相对于上届大选来说,并非只是城市选票大幅度下滑,所以不能说是城市海啸。

在下图中,我用不同的指标来衡量选区是城市还是乡村。 


上面的图是%BN票 vs 3G 渗透率。 一个选区的3G渗透率被定义为Celcom移动宽带网络数据库中可以找到在该选区內的所有记录的总数。 这个数据是从celcom网站所提供的。 
http://coverage.celcom.com.my/CelcomCoverageIIPub/faces/CelcomCoverage.jsp

3G渗透率较高的选区为城市选区,有更好的网络通用性。

同样,数据显示在两届大选中都有很明显的城乡鸿沟。 城市的选民和互联网有更多接触,大多数都把票投给反对党。BN支持率,不管是在3G差或者3G优良的选区,都普遍下滑,所以我们不能称之为城市海啸。 

此时,问一个很好的问题:城市马来人支持反对党是否超过农村马来人? 



在此图中, 绿点的选区3G渗透率<= 18,紫色的选区3G渗透率> = 18。 

事实上,在非马来人占多数的地区,尤其是在城市地区,华人选民以压倒性多数票投给反对党。 这个图其实并没有显示出有任何明显的城乡鸿沟。 在马来人占多数的地区,城市和农村实际上%BN票都很相似。 因此,国阵并没有受到来自农村的马来选民支持。 

结论:民联领导人必须接受这个事实,他们赢得了更多的席位是由于绝大多数华人选票的支持, 他们必须接受,无论是在城市或农村地区,在拉拢比PRU12更多的马来选票这一方面上来说,他们的选举策略完全失败。反对党不能单靠华人以及只有一半的马来人支持,来成立联邦政府。 他们必须寻找一种可以攻入BN马来票堡垒的策略,打击BN一直使用的贿选战术。 他们需要弄清楚为什么马来人对他们的竞选宣言冷感。 

另一方面,拿只鸡的华人海啸论虽然原则上没错。当时,他的言论,充斥着对华人社会的挑衅,很不利于民族和解,甚至造成有人鼓吹抵制华人企业。 BN应该反思,为什么国阵已失去了华人支持,并在未来有悔改表现。 

Wednesday, 22 May 2013

PRU13 Data Analysis Part 1 - Was there a Chinese tsunami or an urban tsunami?

Since the closing of the 13th Malaysian General Election, there has been an ongoing debate of whether a Chinese tsunami or an urban tsunami has caused the ruling coalition Barisan Nasional to suffer its worst results in history.

The debate is getting ugly with some BN supporters hurling all sorts of racist remarks towards Chinese community. PR supporters defended Chinese community by claiming that a multi-racial bulk of urban voters have actually voted against BN. It was a urban tsunami in their opinion. I hope that by presenting a rigorous study of the election data, I can settle the debate and put the matter to rest.

The data as well as the demographics statistics used in this study is obtained from SPR and RTM websites:
Other supporting data sources will also be specified in the course of the analysis.

Our Prime Minister, Mr. Najib blamed on a Chinese tsunami for the loss of support for the government in its worst ever general election results. Is he correct? I am afraid he is.


In the figure above, each data point represents a parliamentary constituency. The y-axis is the percentage of BN votes = BN votes x 100 / total votes.
(In constituencies where BN did not contest, I have taken the BN-friendly independent candidate as the BN candidate.)
The x-axis is % Chinese registered voters in that constituency. We can see clearly a vast drop of support for BN in Chinese-majority areas.
This drop of support has increased significantly from PRU12.

On the other hand, if we look at the same type of figure plotted for the Malay voters:


We see no decrease of Malay votes for BN from PRU12. The drop of % BN votes in the non-Malay-majority areas is likely due to the loss of Chinese votes as shown in the previous figure. So overall there has been no actual change in Malay support for PR. Although PR were able to win or hold on to their seats because of the continued support of Malay voters, without any significant increased Malay support, it is impossible for them to capture Putrajaya and form the federal government.

Let us look at the same type of figure again for Indian voters.


We do see some indications that Indian voters have voted against BN. Because Indian voters are minority in numbers, the trend we see here may have been due to influences by Malay and Chinese voting trends. However, we can be sure that Indian votes have not returned to BN.

Pakatan leaders as well as independent analysis groups have claimed there has been an "urban tsunami", are they correct?



The figure above shows the average % BN votes in all constituencies characterized by a same "metropolitan index".
Metropolitan index of a constituency is defined as the logarithm of the total population of the city or town associated with the constituency.
(The population data of Malaysian cities and towns is obtained from World Gazetteer http://www.world-gazetteer.com/wg.php?x=1171193898&men=gcis&lng=en&des=gamelan&dat=32&geo=-152&srt=pnan&col=ohdq&pt=c&va=x)
Metropolitan index in general measures how urban that constituency is. Roughly speaking:

2, 3: rural
4: sub-urban
5, 6: urban

Areas like Johor Bahru and Kuala Lumpur have metropolitan index 6. Area like Sibuti in Sarawak receives metropolitan index 2.

From the bar chart, we can see that there is an urban-rural divide in both general elections. With voters in categories 2,3,4 voted more for BN and voters in 5,6 voted more for the opposition.
However the drop of support for BN from PRU12 is universal among all metropolitan indices. Because an urban tsunami would require a significant increased support of urban voters for the opposition relative to the last general election, therefore it can be said that an urban tsunami has not happened in this general election.

In the next figure, I use a different metric to measure how urban the constituency is.


The figure above is a plot of % BN votes against 3G peneratration. 3G penetration of a constituency is defined as the number of records in Celcom mobile broadband coverage database that can be found to have an address that matches the constituency. This data is obtained from the broadband coverage lookup tool provided by Celcom website.
http://coverage.celcom.com.my/CelcomCoverageIIPub/faces/CelcomCoverage.jsp

The higher the 3G penetration number, the more urban the constituency is, the better the internet availability is in that constituency.

Again, the data shows a clear urban-rural divide in both general elections. The urban folks who have better access to internet have voted for the opposition. However the drop of support for BN from PRU12 to PRU13 is universal in poor 3G areas and good 3G areas. Since urban-rural divide is not new because it was present in the last general election, we cannot call this general election a result of an urban tsunami.

At this point, a question can be asked: do the urban Malays support the opposition more than rural Malays?



In this figure, I plotted % BN votes against % Malay registered voters. The green points are constituencies with 3G penetration <= 18, the purple points are constituencies with 3G penetration >= 18.

Factor in the fact that Chinese voters in the non-Malay-majority areas have voted overwhelmingly for oppositions, especially in urban areas. This chart has not shown any clear urban-rural divide. In Malay-majority areas, urban and rural seats actually show same % of BN votes. Therefore BN did not receive its support from rural Malays alone.

In conclusion. PR leaders have to accept the fact that they have won more numbers of seats compared to PRU12 due to overwhelming Chinese support. They must accept that their general election strategy has failed completely in wooing more Malay votes than PRU12, whether in urban or rural areas. The opposition cannot form a federal government with high Chinese support but with the same amount of Malay support. They must find a way to break into BN Malay vote base, beat the dirty but effective vote-buying tactics that BN has been using. They need to figure out why PR manifesto has not appealed to the Malays more than cash handouts.

On the other hand, PM Najib has gotten it right about losing Chinese support. However his remarks of "Chinese tsunami" was made in a setting that seems provocative and insensitive to the Chinese community. This does not help in national reconciliation and has since invited retaliative actions against the Chinese community such as the call for boycotting Chinese businesses. BN should reflect on why BN has lost Chinese support and perform better in future to gain it back.

EDIT:
A good analyst has pointed out that there were some outlier data points (total 9) in PRU12 data which had 100% BN votes. These were the seats where BN had had a walkover win. These seats where PR had given up would have likely yielded high BN votes if PR had contested. Taking them away will result in imbalance between PRU12 and PRU13 data. 100% is of course still too high. So I have adjusted 100% to 90% to make it more realistic. This adjustment had not affected any of the analysis and conclusions.