Skip to content

← Accounting

Books on a Computer

Chapter 34 · Accounts Today

In this chapter 5 sections
  1. What the machine writes
  2. The spreadsheet
  3. What a package adds
  4. Moving the books, and keeping them whole
  5. The government’s own systems

At ten past ten on the third of साउन, a family buys a carton of white paint. Yasmin’s helper, logged in under her own name, picks the paint from a list on the laptop, the price fills itself in, the VAT is added, and a bill comes out of the little printer by the counter. The family pays Rs. 9,605 and goes. Nobody opens the journal. Nobody posts anything to a ledger. Nobody writes a word in a cash book.

And yet, when Yasmin looks that evening, the sale is in her journal, in three ledger accounts, in the day’s cash and in the VAT she will owe at the end of the month. Chapter after chapter has been about writing things down twice, by hand, and checking that the two sides agree. Her shop now does it with one tap. So what was all that for?

The short answer is that the computer has changed who does the writing, and almost nothing about what is written. It has also brought new ways for a set of books to go wrong.

What the machine writes

The books Yasmin keeps on her laptop are kept by an accounting package, a program made for nothing else. When the shop moved to it, at the start of a year, the first thing it asked for was a list of accounts: cash, the bank, Sujan the wholesaler, Pasang and the other customers who buy on credit, sales, purchases, rent, VAT, capital, the bank loan. This list, the chart of accounts, is the ledger’s table of contents (Chapter 8 built it page by page). Every account gets a name and a group, so the package knows which ones belong in the balance sheet and which in the profit and loss account.

Then it asked for the balances. On the day the shop changed over, the last balances of the paper ledger became the first entries of the new one: so much in cash, so much owed to Sujan, so much capital. The package added them up and refused to go on until the debits and the credits agreed. They did, because the paper books had been closed with a trial balance, and that was the first test of the move.

From then on, every sale is a voucher, a form on the screen. The person filling it chooses a customer, the goods and the quantity; the package knows the prices, the VAT rate and which accounts a sale touches. When the form is saved, the entry is made. Here is what it writes.

Try it

The Books: one bill, typed once

Each tap is the next thing done at the counter on साउन 3. A made-up shop and figures.

Ledger balancesDebitCredit
Cash9,605
Pasang0
Sales8,500
Sales returns0
VAT1,105
Total9,6059,605
Log, साउन 3
  1. 10:12 Bill 1 savedthe helper

Step 1 of 4. One bill typed, one debit and two credits written.

The bill for the paint says one thing. The package turns it into three ledger lines, and they are exactly the lines a bookkeeper would write by hand. Cash comes in, so cash is debited with the whole Rs. 9,605. Sales go up by the price before tax, Rs. 8,500, a credit. And the Rs. 1,105 of VAT isn’t Yasmin’s at all: it is money she holds for the government, so the VAT account is credited with it, as Chapter 30 showed when she registered. One debit, two credits, and the two sides add up to the same number. The double entry is all still there. It has only moved inside the machine.

The second bill is a sale on credit to Pasang, a contractor, so it is his account that is debited, not cash. And it is wrong. A box of brushes costs Rs. 1,200; the helper typed 12,000. The package doesn’t notice, and couldn’t. As Chapter 11 found, a program will refuse an entry whose two sides differ, and it posts both halves itself, so the slips of copying and adding that a trial balance hunts for almost disappear. But a wrong number entered on both sides is still a perfectly balanced wrong number.

What happens next is the part of a good package that matters most. In a paper book, a careful bookkeeper never rubs out a mistake: she writes a correcting entry, so that anyone can see both the mistake and the fix. A package can be built to enforce that. Yasmin’s won’t let the helper delete a saved bill at all. Only Yasmin can cancel one, and when she does, the package doesn’t erase it. It writes a credit note, a new document with a number of its own, and the entry goes through sales returns, as Pasang’s three returned tins did in Chapter 7: sales returns and the VAT debited, Pasang credited. Bill 2 is still in the books, and so is its undoing. Then the right bill is made, bill 3, for Rs. 1,356.

At the end of the four steps the ledger’s two sides agree, at Rs. 22,961; sales less sales returns come to Rs. 9,700, what was really sold, and Pasang owes exactly what he should. But the log under the ledger tells a longer story than the balances: at 11:40 the helper saved bill 2; at 16:05 Yasmin canceled it with credit note 1. A list like that, kept by the program and not editable by the people it records, is an audit trail. It is the computer’s version of the bookkeeper’s rule against erasers, and what an auditor like Anisha in Chapter 27 reads beside the vouchers: not only what the books say, but how they came to say it.

So the arithmetic has gone to the machine. The judgment hasn’t. Someone still decides that the brushes were sold on credit and not for cash, that a new scooter is a fixed asset and not an expense, that a payment is rent and not a loan. The package puts each voucher where it is told. If the chart of accounts is muddled, or a helper picks the wrong account from a list, the reports at the end will be beautifully printed and wrong. That is why every chapter before this one still matters: a person who doesn’t know what the entries mean can’t tell when the machine has been told something false.

The spreadsheet

The same laptop writes the shop’s letters in a word processor and the slides for the bank in a presentation program, but before a business buys an accounting package, its numbers usually live in a spreadsheet, and many keep both. A spreadsheet is a grid of cells, each holding a number, a word or a formula that works something out from other cells. Change a cell and every formula that depends on it changes too, which makes it so useful for planning: Yasmin and Gaurav built their festival budget in one (Chapter 24), and could try a slow असोज or a late-paying builder in seconds. A depreciation schedule, a loan’s repayments, a month’s wages: anything that is the same sum done over and over, row after row, sits naturally in a grid. How to build one, with formulas copied down and addresses fixed with a dollar sign, is taught in Grade 7 Computing, Unit 5.

What a spreadsheet doesn’t have is the rules. Nothing in an empty grid knows that a debit needs a credit, that a saved bill shouldn’t be changed, or which cells are sums and which are typed in. A formula is just text in a cell, and it can be overwritten with a number by a tired hand, after which it looks exactly the same and never changes again. A formula copied down a column can be copied one row too few. Nobody sees it, because a spreadsheet shows the answers, not the sums.

The most famous example comes not from a shop but from economics. In 2010 two economists, Carmen Reinhart and Kenneth Rogoff, published two versions of a study of government debt and growth. Among its findings was a striking one: in the years since the Second World War, countries whose government debt was more than 90% of everything they produced a year had grown, on average, by −0.1% a year. In the groups below it, growth was around 3 to 4%. The number traveled. After the financial crisis of 2007–09, by the account of the economists who later checked it, it was quoted widely in support of governments cutting their spending.

In 2013 a graduate student at the University of Massachusetts Amherst, Thomas Herndon, set out to reproduce the figure with two economists there, Michael Ash and Robert Pollin. They asked Reinhart and Rogoff for their working spreadsheet, and the two sent it. With it they could follow the published numbers, and found a formula whose range of rows stopped five rows short: the averages for each group of countries were taken over rows 30 to 44 instead of 30 to 49. The five countries left out happened to be the first five in alphabetical order, Australia, Austria, Belgium, Canada and Denmark. Reinhart and Rogoff acknowledged the mistake.

But the story that went round the world, that a spreadsheet slip had changed the course of economic policy, was neater than the truth. The slip was one of three problems in their paper. The other two were choices: the first postwar years of Australia, Canada and New Zealand had been left out, and each country’s average had been given the same weight, whether it spent one year above 90% or nineteen. New Zealand’s single year in the group, 1951, a year of −7.6% growth (copied into the summary as −7.9%), counted for as much as nineteen years of Britain. With every year in and each year counted once, the figure for the most indebted countries became +2.2%, not −0.1%, and most of that gap came from the two choices, not from the five missing rows.

Reinhart and Rogoff accepted the coding error and disputed the rest. The years left out, they said, hadn’t been in their data when they wrote, and giving each country one voice was a method, not a mistake. Even the corrected figures, they pointed out, showed slower growth where debt was high: about 2% against 3 to 4%. How much high debt slows a country down, and which causes which, economists still argue.

That is still a lesson about spreadsheets, and a sharper one. The formula error and the choices were found only because someone outside could open the file and follow every cell back to its source. A set of numbers that can’t be traced back is a set of numbers that can’t be checked, however neat it looks. An accounting package keeps its trail by design. A spreadsheet keeps one only if its maker builds it in: the data in one place, the sums in another, each formula checkable, nothing typed over.

What a package adds

Seen from the counter, an accounting package is a set of forms. Inside, it follows the path of a set of paper books. Papers come in: bills, receipts, purchase orders, a barcode read at the counter. Each is checked and entered as a voucher, numbered in order, so that a missing number is as plain as a torn-out page. The vouchers are posted to the ledger’s accounts, each with a code and a group in the chart of accounts. Everything is stored in a database, a store of records that several people can search and add to at once, which is how the helper can bill at the counter while Yasmin reads the ledger, with copies kept elsewhere. The reports are read off the stored books: the trial balance is every account, the profit and loss account the income and expense groups, the balance sheet the rest. That is why they come out the moment the vouchers are in: the package reads pages it already has.

Two things are added on top. One is a list of who may do what. Yasmin’s helper has a login of her own, and it only makes bills; she can’t cancel one, change a price or see the bank account, and the shop’s books stay behind Yasmin’s password, as Chapter 26 had them. The other is the log. Both are old rules, two keys and no eraser, that now hold without anyone remembering them and leave a mark when they are broken.

A business can buy such a package, made for many businesses and set up for its own, or have one built when nothing on sale fits. What it gives back is speed, fewer slips, reports at any moment and any record found in seconds. It costs money: the computers, the program, power and internet, and replacing them as they wear out or the technology moves on. It needs trained people, because someone who doesn’t understand the entries will type wrong ones into the right boxes. It makes the business depend on the machine: when the power or the laptop fails, the books stop. And it brings a danger paper never had, records that can be lost, stolen or changed all at once.

Moving the books, and keeping them whole

Books on paper are hard to lose all at once, and hard to change without a trace. Books on a computer can vanish with one stolen laptop or failed disk, and a record can be changed without a mark. So the move from paper, and every year after it, needs care that paper never asked for.

The move itself tests the books. The closing trial balance of the old books must equal the opening balances of the new, account by account. The list of what each customer owes must add up to the same total in both, and so must what the shop owes each supplier. The stock on the shelf must match. Yasmin did each check before the package took over, and kept the closed paper books. If the new books start wrong, every year after them is wrong too, however neatly the computer adds.

Then there are copies. A backup is a copy of the books kept somewhere else, made often enough that losing the laptop costs a day’s work and not a year’s. How to make backups that survive a fire, a theft or a mistake (three copies, on two kinds of storage, one of them away from the building) is in Grade 8 Computing, Unit 8. For a set of books, one more test matters: a backup is only a backup if it can be opened and read again, so it is worth trying, now and then, to bring one back.

Nepal’s law has a view on all of this. The Electronic Transactions Act, 2063, says that where a law asks for a record to be kept in writing, a record kept electronically counts as long as the Act’s procedures are followed (section 4). Where a law asks for a record to be kept for a certain time, an electronic one counts only if three things hold (section 6): it must stay accessible for later use; it must be kept in a form that can show it again exactly as it was made, sent or received; and it must keep the details of where it came from, where it went, and the date and time. Those three conditions are a description of a good backup and a good audit trail. And the same Act makes it a crime to destroy, delete or alter information in a computer knowingly, meaning to cause an organization wrongful loss, with a fine of up to Rs. 2 lakh, up to three years in prison, or both (section 46).

The tax office has gone further, and recently. Every bill Yasmin issues is, for a VAT-registered shop, a tax invoice, the document her customers use to take off the VAT they paid (Chapter 30). So the Inland Revenue Department, which collects VAT, cares a great deal whether a bill can quietly disappear. Since the Finance Act of 2083, the VAT Act’s section 14क lets the department set the standards for issuing bills electronically; lets it order, by public notice, the taxpayers it names to issue their bills electronically and connect them to its own central billing monitoring system, केन्द्रीय बीजक अनुगमन प्रणाली, called CBMS, or else to use the billing system it provides; and has it set procedures for the security and reliability of the billing software and machines. The Income Tax Act’s section 81(5) now reads the same. And both Acts now set a fee: a taxpayer issuing bills electronically under these sections who uses software that can delete or correct the data pays Rs. 5 lakh, and Rs. 1 lakh for breaking their other rules.

That fee is about exactly what Yasmin’s package did with bill 2. Software that lets a saved bill be deleted lets a sale disappear from the books, and with it the VAT the government was owed. The law’s answer is the bookkeeper’s old rule, written for machines: never erase; cancel, and leave both in sight.

The government’s own systems

The government keeps its books on computers too, most of them in systems run by the Financial Comptroller General Office.

The Treasury Single Account, TSA, is the one account from which the federal government pays, first tried in 2009/10, as Chapter 31 told: offices send payment orders and the money stays in one place until it is paid. Provinces have their own, the PTSA. The Revenue Management Information System, RMIS, is where every rupee of revenue paid at any bank is entered, and a deposit doesn’t count until it is there (Chapter 30); a directive of 2077 brought the companies licensed for electronic payments into it, so that revenue can be paid by card or online. The Computerized Government Accounting System, CGAS, is where federal and provincial offices keep their books, and their spending counts once a payment order goes through it to the TSA. And SuTRA, the subnational treasury regulatory application, keeps the budgets, books and reports of the local levels: by a government decision of 30 April 2019 it runs in all 753 of them.

Around these sit others, among them the budget systems BMIS and LMBIS, the second for the line ministries, and the Financial Management Information System, FMIS, through which the yearly statement is put together. Together they are the government’s version of Yasmin’s package: a chart of accounts (the codes of Chapter 29), vouchers, roles that let one office order a payment and another make it, and reports that come out of the books already kept.

In early 2022 the Office’s bulletin reported CGAS linked to the Auditor General’s own audit system, so that auditors can see spending as it happens, and SuTRA linked to the national payment network, so that local levels can pay electronically. In November 2026 its website led to a new sign-in, IPFMS, integrated public financial management services, which calls itself the Office’s integrated system, and whose training videos carry CGAS’s name beside its own. The names on the screens change; the double entry underneath doesn’t.

A computer has taken away the copying and the adding, and with them most of the slips that trial balances used to catch. It hasn’t taken away a single decision about what happened and where it belongs, and it has added a new danger, books that can be changed or lost without a trace, which the law and every good package now answer with the oldest rule in bookkeeping. Yasmin’s helper still has to know that brushes on credit are Pasang’s debt. Yasmin still has to know that a mistake is canceled, not erased. And anyone who can read a ledger can still follow every number the machine prints back to the bill it came from, which is exactly the point.

Found a mistake? Please tell us.