Taxation changes behaviour. Taxes on goods and services makes them more expensive, and in most cases people buy fewer of them as a result. Taxation on incomes makes work less attractive, motivating some people to do less of it. Taxes on business usually fall on those who buy the products of those businesses, often reducing demand to below the level it would reach without taxation. Taxes that people regard as unfair or oppressive will often lead people to take steps to avoid their burden. Taxes on inheritance lead people to dissipate their wealth early, or they break up the capital pools that enable heirs to invest. Dynamic models of the economy try to take account of this. Doubling the tax on tobacco products, for example, does not yield twice the revenue. It might lead to twice the smuggling, though. Vast increases in the duty on alcohol does not raise revenue in proportion, thought it does lead some drinkers to move down to cheaper booze. Some tax increases actually yield less revenue because of the behavioural responses they trigger.
Arthur Laffer famously noted that a 0% tax on incomes yields no revenue, as does a 100% tax. The graph line that links those two zeros is the Laffer Curve, and it has a peak somewhere whose rate yields the most revenue. Even this moves as people adjust to the new status quo. If a 50% rate of income tax is lowered to 40%, it is highly likely that it will soon bring in more revenue. The rate reduction makes work more attractive, so people do more of it. It also makes complex and expensive avoidance schemes less necessary, as people opt just to pay the lower rate tax instead.
When the economy in question had adjusted to the 40% top rate, however, more revenue might then be raised by lowering it to 35%. The optimum revenue-raising rate depends on the status quo to begin with. In many cases it might be appropriate to lower taxes to raise more revenue, rather than to hike them. It is never likely that higher rates will yield more revenue in proportion to the increase.