The power of incumbency

OK – post election depression done with – here’s some analysis.
Prior to the expenses scandal it was conventional to think that the combination of new technology and the £10k p.a. MP’s communication allowance would increase the power of incumbency. So let’s see if the incumbency effect exists. Here’s the percentage drop in Labour votes for every (pre election) Labour held seat in Wales in 2010 along with incumbency status:
Seat Retireree % Drop
Aberconwy Y -8.5%
Islwyn Y -15.1%
Pontypridd Y -15.4%
VoG Y -7.8%
Clwyd South Y -6.8%
Swansea West Y -7.2%
Llanelli N -4.5%
Carmarthen West N -4.0%
Ynys Mon N -1.3%
Torfaen N -12.1%
Newport East N -8.2%
Newport West N -3.6%
Caerffili N -10.5%
Rhondda N -12.8%
Merthyr N -16.8%
Cynon N -10.5%
Cardiff South & P’arth N -7.7%
Cardiff West N -3.6%
Cardiff North N -1.9%
Swansea East N -5.1%
Ogmore N -7.1%
Bridgend N -6.6%
Gower N -4.0%
Wrecsam N -9.2%
Alyn & Deeside N -9.2%
Aberafan N -8.1%
Neath N -6.3%
Vale of Clwyd N -3.6%
Delyn N -4.9%
Arfon N -3.5%

The simple average drop for seats with a retiring MP was 10.1%. The average drop in the other seats was 6.9%.
Is it that simple? No – some normalisation is possible. The most striking pattern of the data is the big drops in the South Eastern coalfield seats – all drops of over 10% in the sample are in this category (Merthyr’s been different since Keir Hardie’s day btw). 2 of the 6 in the retiree category but only 5 of the other 24. If we normalise the drop percentage for the retiree category the comparison drops become 9.2% cf 6.9%. Significant? I think so but am more than willing to take advice from those who remember statistical confidence arithmetic…