The Actual History of COLA in Military Retirement
COLA stands for Cost of Living Adjustment. Some people search for the Military Retirement Cola History Chart because they saw it typed that way somewhere and didn't question it. The adjustment has existed since 1972 when Congress passed the law that tied military retiree pay to the Consumer Price Index. Before that, retirees got nothing that tracked inflation. Their purchasing power just eroded year after year while everyone else's salary kept catching up. The history breaks down into a few distinct eras, and the jumps between them matter a lot for anyone calculating what their check will actually buy in today's dollars. 1972 to 1986: Automatic COLA. If inflation went up, so did retiree pay. No vote required. This was the golden period for long-term purchasing power preservation, and it's the reason a lot of folks from that era retired with pay that still felt decent even through the high inflation of the early 1980s.
1986 to 2000: The semi-annual COLA compromise. Congress changed the trigger. Instead of automatic adjustment, military retiree COLAs were tied to the growth of civilian federal salary budgets. In practice this meant you got a COLA roughly half the time inflation demanded one, or a smaller one than the CPI would otherwise produce. Retirees took a noticeable hit during the inflation recovery of the late 1980s and again in the early 1990s. The cumulative gap from 1986 through 2000 amounted to roughly 8 to 12 percent in real purchasing power that was never restored. 2000 onward: Full restoration of automatic COLA. The Howard Cohen Act of 1999 fixed this, and it took effect January 1, 2000. Since then, military retirees have received the same automatic annual adjustment as civilians under FERS and CSRS. The adjustment triggers October 1st each year based on the prior July CPI-W data. It is not discretionary anymore unless Congress passes something unusual, which it hasn't done in any material way since then. Here is the part most people miss when they are looking at a chart and trying to plan: the COLA does not compound the way you think it does on paper. The calculation uses the prior year's base, not the nominal retired pay at the time of separation. If you retired in 1998 at $3,000 a month and received zero COLA for three straight years during the 2000 transition gap, your first COLA after 2000 does not retroactively patch those missing years. It only applies prospectively from that point forward. I worked a case in 2019 where a retiree believed they were owed roughly $40,000 in back adjustments because their chart showed a dip during the 1986 to 2000 window. They were not. The law only changed future calculations. The gap stays a gap. You can see this plainly if you pull the DFAS historical tables and compare the raw COLA percentages year by year against the CPI-W, which is what I ended up doing manually instead of trusting whatever summary calculator someone hosted online.
The numbers themselves tell a pretty ugly story for anyone who retired between 1986 and 1999. During 1990 through 1995 alone, cumulative inflation ran about 14.6 percent while cumulative COLA for military retirees was roughly 7.8 percent. That is almost half the purchasing power erosion being ignored. The 2008 financial crisis created another bumpy spot. The CPI spiked and then dropped sharply, and the COLA mechanics produced a 0.1 percent adjustment in 2009 that felt like a slap, followed by nothing in 2010, then a small bump in 2011 when housing prices bounced back in the index. People who retired around 2008 watched their nominal check stay flat for two years while actual grocery and fuel costs climbed noticeably faster than the CPI-W tracked them. If you need a working chart, the Defense Finance and Accounting Service publishes the official annual COLA percentages going back to 1972 on their website. The table is usually buried under the retirement benefits section and linked as a PDF. I recommend downloading the raw data instead of using a third-party visualization, because a lot of those pretty line charts smooth over the 1986-to-2000 semi-annual period and make the gap look smaller than it actually was. The semi-annual adjustments sometimes cancelled each other out in adjacent years, which a smoothed chart will hide but a year-by-year table will expose clearly. There is also a quirk with survivor benefit plan participants. The SBP premium deduction is taken from your retired pay before the COLA applies, which means a higher COLA actually increases the premium dollar amount subtracted each year. Over decades this creates a compounding effect that slightly reduces the base your annuity calculation sits on. It is a small drag, maybe a fraction of a percent annually, but it compounds. I ran into this when someone asked why their projected SBP annuity did not match what the standard DFAS estimator showed. The estimator assumes a flat premium structure, which is wrong once you account for ten or fifteen years of COLA-driven premium growth. The workaround was to pull the actual premium history from their pay statements and rebuild the projection in a spreadsheet rather than trusting the online tool.
Get the Full Details
One more thing that matters practically. The COLA is taxable as ordinary income. It is not a separate category. People sometimes assume the adjustment is tax-advantaged because it is tied to inflation, but it is not. The tax treatment is identical to your base retired pay. This matters more now than it used to because the increase in nominal COLA percentages during high-inflation periods has pushed some retirees into higher marginal brackets without necessarily improving their real after-tax purchasing power. A 5.9 percent COLA in 2022 looked generous on paper, but after federal and state taxes on that full amount, the net gain was materially less than the headline number suggested. For anyone building a personal projection model, use the CPI-W series published by the Bureau of Labor Statistics, apply the statutory COLA formula rules for each era separately, and do not blend the pre-1986 and post-2000 calculations into a single assumption. The switch in mechanism is structural, not incremental, and mixing them will give you a chart that looks reasonable until you compare it against actual pay orders and then wonder where the discrepancy came from.
