# The year-on-year trend also points into the expected direction for all of the

metrics, and is in fact very significant for many of them. The interaction variable

# ‘Post*Year’ in seven out of ten cases has a different sign from the trend variable,

indicating a softening (or reversal) of the performance trend after privatisation. As

expected the model finds oil prices to play an important role in explaining e.g. higher

profitability and capital expenditure, but the net effect of privatisation on firm

performance remains substantial, even when controlling for oil prices.

# Over the seven-year period around privatisation, the ‘typical’ NOC – assuming the

average of the fixed unit effects α_{i}, and a constant oil price of US$50/barrel –

improves its return on sales by 3.6 percentage points (0.6 p.a.), increases output per

employee by 30% (CAGR 4.4%), and capex by 47% (CAGR 6.6%); total output is up

by 40% (CAGR 5.8%), and the ratio of employees over assets falls by 35% (CAGR

6.9%); the leverage ratio of debt over equity drops by 16 percentage points (2.7 p.a.),

and the dividend payout increases by 16 percentage points (2.6 p.a.). The timing of

performance improvements varies between metrics, but overall the importance of the

pre-privatisation period is striking. In what might be called the ‘anticipation effect’ of

privatisation, a significant part of the total benefits are realised in the run-up to the

handover of property rights. Whilst capital expenditure is ramped up substantially

following privatisation, physical output and employment intensity improve steadily

throughout the seven-year period, and in terms of profitability all of the

improvements are already realised in anticipation of the deal. As to the size and nature

of operating cost reduction: in the three years leading up to privatisation the to-be-

privatised firms manage to reduce their unit production costs by 7% and cut their

employment levels by 5%, but based on the point estimates neither of these are

sustained beyond this date, and neither of them is of statistical significance.

18