The Oil-Fired Grid: What Bangladesh Burns When the Gas Runs Short

Executive Summary. In 2023 Bangladesh generated 21.69% of its electricity from oil, the highest share of the six countries compared here and 13.1 times the peer median of 1.65%. India ran on 0.21%, Vietnam on 1.29%, Thailand on 1.65%, Indonesia on 1.98%. Only Pakistan came anywhere close, at 11.39%. Oil is what a power system burns when it has run out of cheaper options, and Bangladesh burns it at a scale no comparator here matches. It sits on top of a gas-first grid: 65.82% of generation from natural gas, also the highest of the six against a peer median of 12.91%. Gas and oil together supply 87.51% of output, and fossil fuels 98.1%. Renewable output was 1.51% in 2021, the lowest of the six and below Bangladesh's own 8.81% in 2000. For all of that fuel, the system delivers 602.9 kWh per person, fifth of the six and 42% of the peer median, while 44.31% of energy use is imported, second-highest of the group.

A fuel the comparators barely use

The number that defines the Bangladeshi grid is not the one usually quoted.

In 2023, 21.69% of electricity output came from oil. Set that against the comparators: India 0.21%, Vietnam 1.29%, Thailand 1.65%, Indonesia 1.98%, Pakistan 11.39%. The peer median is 1.65%, which makes the Bangladeshi share 13.1 times the median and 20.03 percentage points above it. Bangladesh ranks first of six here, and first is not a place to want on this indicator.

Source: World Bank World Development Indicators, electricity production from oil sources (EG.ELC.PETR.ZS), % of total output, 2023.

Liquid fuel has a specific role in a power system. It starts fast, it can be trucked or barged to wherever the shortage is, and it is priced off a world market rather than a domestic supply contract, which is why it usually sits last in the dispatch order. Every system keeps some. Four of the five comparators keep it under 2% of output. When a country's oil share sits at the top of a peer group by a factor of thirteen, the plants burning it are not covering the evening peak for a few hours. They are covering structural shortfall.

Oil is the symptom, gas is the system

The oil share only makes sense next to the fuel it is standing in for.

Bangladesh generated 65.82% of its electricity from natural gas in 2023, the highest share of the six. Thailand is close behind at 64.16%, then Pakistan 27.72%, Indonesia 12.91%, Vietnam 9.62%, India 2.96%. The peer median is 12.91%, so the Bangladeshi gas share runs about 5.1 times the median.

Coal runs the other way. Bangladesh took 10.59% of its electricity from coal, the lowest of the six, against India 74.43%, Indonesia 69.12%, Vietnam 44.76%, Thailand 14.62% and Pakistan 12.78%, a peer median of 44.76%. Gas and oil together are 87.51% of output. All fossil sources together are 98.1%.

That is a grid with one primary fuel and one substitute, and the substitute is the one priced on a world market. A system built around gas turbines can swing to liquid fuel without building new plant, which is convenient in a shortage and punishing as a standing arrangement, because the swing is settled in dollars at whatever the market asks that quarter.

Two confounders belong here before anyone reads policy failure into the mix. First, fuel shares track geology as much as intent. India's coal share reflects domestic coal; Indonesia's does too, and Indonesia is a net energy exporter, with net energy imports of -82.92% of energy use in 2022. A country sitting on its own gas will build gas plants, and a country whose gas fields disappoint later inherits the plants regardless. Second, six countries is a very small panel. These are cross-country associations at a single point in time, and they identify a pattern, not a cause.

The renewables line points the wrong way

If the oil share were a transitional problem, the renewable share would be climbing to replace it. It is not.

Renewable electricity output was 1.51% of total output in 2021, the lowest of the six. Vietnam reached 41.17%, Pakistan 28.28%, Thailand 19.93%, India 19.13%, Indonesia 15.53%. The peer median is 19.93%, which puts Bangladesh 18.42 percentage points below it.

The trend is the harder fact. Bangladesh's renewable share was 8.81% in 2000 and 1.51% in 2021, a fall of 7.3 percentage points, or 82.9% of the starting share.

Source: World Bank WDI, renewable electricity output (EG.ELC.RNEW.ZS), % of total output, 2021, the latest year with comparable coverage across the six.

Three qualifications, because this figure is easy to overread.

It is a share, not a quantity. Total generation grew over the same window (electricity use per person rose 6.07 times between 2000 and 2023), so a renewable source of roughly fixed size would see its share collapse without a single turbine being switched off. Part of the decline from 8.81% to 1.51% is arithmetic in the denominator, not retreat in the numerator. This series carries shares, not installed capacity, so it cannot separate the two.

It is a single aggregate. The series does not break renewable output down by technology, and the peer comparison is not like for like: Vietnam's 41.17% and Pakistan's 28.28% rest partly on mountain river systems that a delta does not have. The fair comparison for Bangladesh is solar, and this data does not isolate it.

It is a different vintage. Renewable output here is 2021; the fuel shares above are 2023. The two years should not be spliced into a single accounting of one grid.

What survives all three qualifications is the direction. Whatever the denominator did, the country that ended the period with the group's heaviest reliance on oil also ended it with the group's thinnest renewable share, and a smaller share than it carried a generation earlier.

What the grid actually delivers

An expensive fuel mix could still be defensible if it bought abundance. It has not.

Electric power consumption was 602.9 kWh per person in 2023, fifth of the six, ahead only of Pakistan at 518.3. Thailand consumed 2,965 kWh per person, Vietnam 2,585, Indonesia 1,445, India 1,182. The peer median is 1,445 kWh, so Bangladesh sits 842.3 kWh below it, at 42% of the median.

The growth record is genuine and should not be buried under the ranking. Consumption per person was 99.25 kWh in 2000 and 602.9 kWh in 2023, a rise of 503.6 kWh, or 507.4%, and the population it serves reached 173.6 million by 2024. The economy grew 4.2% in 2024 on a GDP of US$450.1 billion, and it grew on this grid.

Source: World Bank WDI, electric power consumption per capita (EG.USE.ELEC.KH.PC), kWh, Bangladesh against comparators, 1990-2023.

The point is the combination, not either half. Bangladesh pays for the most oil-dependent generation mix in the group and receives less electricity per person than every comparator except Pakistan. A country using 21.69% oil to reach 2,965 kWh per head would be making an expensive but comprehensible trade. Using it to reach 602.9 means the liquid fuel is not funding growth in supply. It is funding the maintenance of a supply that is already thin.

The wires are not the alibi

The standard explanation for a stressed South Asian grid is the distribution network: power generated, then lost or stolen before it is billed. That explanation does not carry here.

Transmission and distribution losses were 7.9% of output in 2023, third of the six. Pakistan lost 15.35% and India 14.16%. Thailand lost 7.16%, Vietnam 6.58%, Indonesia 6.45%. The peer median is 7.16%, so Bangladesh runs about a tenth above the median.

Third of six is not a clean bill of health, and the WDI loss series mixes technical losses with unbilled consumption, so it is a blunt instrument. But it is nowhere near the outlier the oil share is. On losses Bangladesh is ordinary; on oil it is thirteen times the median. The binding constraint is upstream of the wires, in what the plants are burning.

Where the exposure sits

The last figure sets the risk. Net energy imports were 44.31% of energy use in 2022, second-highest of the six behind Thailand at 57.12%, against Pakistan 39.79%, India 36.53%, Vietnam 34% and Indonesia at -82.92% as a net exporter. The peer median is 36.53%, so Bangladesh imports about 1.21 times the median share.

Put the two facts together. Nearly half of energy use is bought abroad, and 87.51% of electricity comes from the two fuels most exposed to world prices. That is a generation system whose operating cost is set outside the country and settled in foreign currency, on a grid whose customers are largely domestic and largely price-sensitive.

What this data cannot do is price it. There are no fuel costs, no subsidy figures, no capacity payments and no plant-level dispatch records in this data. The claim here is about exposure and fuel shares, not about taka per kilowatt-hour, and anyone who wants the second number will have to go to the accounts, not to WDI.

What would change this conclusion

A checkable test. If the oil share of electricity output falls below 10% and the renewable share rises above 5% in the WDI releases covering 2030, the argument that the grid is structurally locked into liquid fuel is wrong, and the oil block was a bridge that got dismantled. If the oil share is still near 21.69% in 2030, the bridge was the destination.

Three moves, with owners and signals.

1. Publish monthly generation by fuel, and the cost of each block. The reason this essay stops at exposure rather than cost is that the public record stops there. An annual WDI share cannot show whether the oil plants ran on the evening peak or on baseload. Owner: Power Division, with BPDB. Success signal: a published monthly series of generation and fuel cost by plant, in the public domain, within one budget cycle.
2. Buy the substitute, not more of the swing. Every additional liquid-fuel peaking contract deepens the same dependence the last one created. Contracting firm solar with storage attacks the exact block that oil now covers. Owner: SREDA, with BERC on tariff design. Success signal: renewable output above 5% of total output in the WDI release covering 2030, which would be the first move back toward the 8.81% share of 2000.
3. Report unserved energy as a standing national statistic. The oil block is buying reliability, and no one can weigh its cost against its benefit while the benefit is unmeasured. Owner: BPDB, published alongside BBS releases. Success signal: a regular published series of unserved energy and outage duration, which would let the next version of this comparison test the trade instead of describing it.

The counterargument

The strongest objection is not that the numbers are wrong. It is that they are being read backwards.

Bangladesh has the lowest coal share of the six, at 10.59% against a peer median of 44.76%, while India runs at 74.43% and Indonesia at 69.12%. On that view the oil is the price of refusing coal, and this essay is criticising a country for the cleaner path. The objection continues: gas and oil turbines can be sited near load and built quickly, without the coal supply chain and the local air pollution that come with the alternative.

Two answers. First, this data cannot settle the emissions comparison: there is no carbon intensity, no particulate measure and no health cost anywhere in the series behind this essay, so anyone claiming Bangladesh's mix is cleaner than India's is making a claim beyond what is shown here. Second, and more directly, the avoided coal did not become renewable generation. It became imported liquid fuel, on a grid that reached 1.51% renewable output in 2021, the lowest of the six. Refusing coal is a defensible choice. Refusing coal and arriving at 21.69% oil, 98.1% fossil generation and 602.9 kWh per person is not the same choice, and should not borrow the credit for it.

---

Data sources: World Bank World Development Indicators, retrieved from the BDPolicyLab data lake, 2026-08-10. Comparators are India, Pakistan, Indonesia, Thailand and Vietnam; "peer median" is the median of the five comparators reporting in the stated year, excluding Bangladesh. Generation fuel shares and transmission losses are 2023, renewable output 2021, energy imports 2022, population and GDP 2024, these being the latest years with comparable coverage across the group. All comparisons are cross-country associations in a panel of six and do not establish causation.

Sources

1. World Bank WDI, Electricity production from oil sources (% of total), EG.ELC.PETR.ZS: https://data.worldbank.org/indicator/EG.ELC.PETR.ZS
2. World Bank WDI, Electricity production from natural gas sources (% of total), EG.ELC.NGAS.ZS: https://data.worldbank.org/indicator/EG.ELC.NGAS.ZS
3. World Bank WDI, Electricity production from coal sources (% of total), EG.ELC.COAL.ZS: https://data.worldbank.org/indicator/EG.ELC.COAL.ZS
4. World Bank WDI, Renewable electricity output (% of total electricity output), EG.ELC.RNEW.ZS: https://data.worldbank.org/indicator/EG.ELC.RNEW.ZS
5. World Bank WDI, Electric power transmission and distribution losses (% of output), EG.ELC.LOSS.ZS: https://data.worldbank.org/indicator/EG.ELC.LOSS.ZS
6. World Bank WDI, Electric power consumption (kWh per capita), EG.USE.ELEC.KH.PC: https://data.worldbank.org/indicator/EG.USE.ELEC.KH.PC
7. World Bank WDI, Energy imports, net (% of energy use), EG.IMP.CONS.ZS: https://data.worldbank.org/indicator/EG.IMP.CONS.ZS