Merger Savings Forecasts vs Reality: A Paired Comparison

Every merger announcement comes with a promise: the combined company will save money. Costs will fall, overlaps will vanish, and "synergies" will deliver hundreds of millions. Years later, few people check. Statistics gives us a simple way to ask whether such forecasts are, on average, honest.

Why the comparison must be paired

Imagine ten completed deals where the companies forecast annual savings of 120, 80, 200, 150, 60, 90, 300, 45, 110 and 170 million dollars, and later reported actual savings of 95, 85, 140, 100, 40, 70, 190, 50, 60 and 120. Deals differ enormously in size, so comparing the two lists as unrelated groups would be noisy. The informative number is each deal's own gap between forecast and result.

Those gaps are −25, +5, −60, −50, −20, −20, −110, +5, −50 and −50. The paired samples t-test calculator gives a mean shortfall of 37.5 million, with p ≈ 0.007. The confidence interval calculator for unknown variance, applied to the gaps, runs from about −62 to −13 million. Overall, realized savings were about 72% of the forecast.

A test that does not assume normality

Ten deals is a small sample, and the gaps are skewed by one large miss. The Wilcoxon signed-rank calculator for paired data uses only the ranks of the gaps and gives p ≈ 0.008, the same message. Eight of the ten deals fell short. When two methods agree, confidence in the finding grows.

Why forecasts lean optimistic

Optimism is a documented pattern. Research on corporate acquisitions found that overconfident CEOs made acquisitions at higher rates, with the odds of a deal 65% higher, and that markets reacted much more negatively to their announcements, about −0.9% against −0.12% for other CEOs (Malmendier and Tate, 2008).

Incentives play a part too. Forecasts are made to win approval from boards, regulators and shareholders, while the people checking results years later face different incentives. Evidence-based reviews of merger research, such as the one from the US Federal Trade Commission, treat stated expectations with caution.

Cautions about our own example

Our ten deals are hypothetical, and real data would raise further questions. Were the deals chosen because they were completed, ignoring failed ones? Did companies define "savings" the same way before and after? Are realized savings measured fairly when the business also changed for unrelated reasons?

Statistics cannot answer those questions, but it tells you how large a gap is and how likely it would be by chance. Here, a systematic shortfall is far more likely than coincidence.

Reading the ratio

A ratio is easier to remember than a difference. In total, realized savings were about 72% of the forecast, though the average of the ten individual ratios is about 77%, because the large deals missed by more. Which figure you quote matters. The total-weighted ratio describes the money, while the average ratio treats every deal equally. State which one you used.

Also look at the direction of the surprises. In our data only two deals beat their forecast, and both by just 5 million, while the misses ran as large as 110 million.

Errors that lean one way point to bias, not bad luck.

Next time a deal promises big savings, look for the follow-up. A forecast is only a claim until someone compares it with the outcome.