What the previous analysis established—and did not

In the previous Why Is Tax Law So Difficult? Reading Parenthetical Clauses with Color Coding, I counted parentheses in 20 laws and orders. Parentheses per sentence ranged nearly ninefold, from 0.39 in the Administrative Case Litigation Act to 3.48 in the Corporation Tax Act Enforcement Order.

I initially thought administrative law would explain the density, but counting contradicted that. The general administrative statutes—the Administrative Case Litigation Act, Administrative Complaint Review Act, and Administrative Procedure Act—were all similar to or lower than the Civil Code. Only tax law stood out.

Why tax law? The previous article stopped there. This continues the inquiry.

This count found no strong linear correlation between enactment year and parenthesis density in current text. Some instruments showed an association between delegation wording and parenthesis density, but this does not establish historical effects or causation.Two laws enacted in 2020 differ fourfold in parenthesis density.

The first suspect was the era

One plausible explanation is thatMeiji-era laws were written simply, while modern precision increased parentheses.

The Civil Code dates to 1896, Commercial Code to 1899, and Penal Code to 1907. But enactment year differs from when current, repeatedly amended wording was written. The Companies Act dates to 2005, Administrative Complaint Review Act to 2014, and Worker Cooperatives Act to 2020; enactment dates alone cannot measure historical drafting techniques. First, we examine their simple association with current-text density.

I checked.

Arrange 23 laws and orders by enactment year

I added three Reiwa-era laws to the previous 20, deriving enactment year from legal instrument numbers and sorting oldest first. Parentheses were counted using the previous method.

The figures do not increase from top to bottom. Values of 0.51 and 0.60 near the top give way to 3.48 in the middle and return to 0.50 and 0.68 below. The three latest laws vary at 0.64, 2.52, and 0.83.

A correlation coefficient measures linear association: near 1 indicates strong same-direction movement, near −1 strong opposite-direction movement, and near 0 weak linear association. Nonlinear relationships may exist even at zero.The correlation between enactment year and parentheses per sentence was approximately 0.13. These 23 instruments show no strong linear correlation.(Calculated across 23 instruments)

Two laws a century apart were almost identical

One pairing surprised me.

The Civil Code and Code of Civil Procedure are 100 years apart, at 0.51 and 0.50. The Penal Code and Administrative Complaint Review Act are 107 years apart, at 0.71 and 0.68.However, this compares current texts of laws enacted far apart, not a direct measure of a century's drafting changes.

Only the Commercial Code–Companies Act pair increases by 1.2×. Since the Companies Act separated the Commercial Code's company provisions, it can also be read as only a 1.2× difference within one lineage.

Two laws enacted in 2020 differ fourfold

Consider differences within the same enactment year.Here are two laws enacted in 2020.

They were enacted by the same Diet in the same year. Yet0.64 and 2.52 differ by almost fourfold.The Worker Cooperatives Act is close to the 1899 Commercial Code's 0.60, while the COVID-19 tax act is near the 1965 Corporation Tax Act's 2.68.

The COVID-19 tax act's full title is the Act on Temporary Special Provisions of National Tax Laws to Address the Effects of COVID-19 and Related Matters.A Reiwa-era tax law resembles a Showa-era tax law more than a Reiwa-era cooperative law.In this comparison, subject matter appears more prominent than shared enactment year. But two examples do not establish the cause.

I also measured the 2021 Act on Vesting Ownership of Land Acquired by Inheritance or Similar Means in the National Treasury: 0.83. It concerns inheritance but does not calculate tax.

Check within the same law too

Comparisons across laws mix other factors. Therefore,I compared articles with and without suffix numbers within individual laws.

Articles may be numbered 66-4, for example, to insert new provisions without renumbering existing ones. However, unsuffixed articles may also have been entirely rewritten later.Suffix numbering alone does not identify when current wording was written.

The following is a descriptive numbering comparison, not a comparison of old and new writing.

Ratios range from 0.66 to 1.64; suffixed articles have greater density in some instruments and less in others. This table cannot determine the effect of drafting date.

These comparisons do not separate subject-area and historical effects. Next, examine delegation wording.

Investigating the association with delegation expressions

I counted the remaining candidate:occurrences of “as prescribed by Cabinet Order” and “as prescribed by Ministerial Order.”These indicate matters left to subordinate legislation rather than fully specified in the act.

Recalculating from the rounded figures for the 20 instruments in this table,the correlation between delegations and parentheses per sentence is approximately 0.80.However, the enactment-year comparison used 23 instruments. The coefficients cannot directly identify which factor causes the difference. The Penal Code and Administrative Case Litigation Act have no Cabinet Order delegation expressions, while the Special Tax Measures Act has 3,284.

This table excludes the Labor Standards Act, National Public Service Act, and Road Traffic Act from the original 23. Excluding observations inconsistent with a hypothesis can strengthen apparent correlation. This limited supplementary count does not establish the same association across all 23. Further analysis should compare before and after exclusion consistently and account for subject matter and length.

One drafting pattern connecting delegation and parentheses

The reasoning is as follows.When delegating to a Cabinet Order, legislation must specify the scope of delegation in the provision.The principle of taxation by law prevents a blank delegation. Parentheses help delimit that scope.

Consider Article 25(3) of the Corporation Tax Act, one sentence of 196 Japanese characters. “As prescribed by Cabinet Order” appears4 times. Pink marks those four occurrences; yellow marks parentheses.

Where a domestic corporation receives approval of a rehabilitation plan or another equivalentprescribed by Cabinet Orderevent occurs, and that corporation has revalued its assets in accordance withprescribed by Cabinet Orderan appraisal, the assets(excluding those unsuitable for recognition of revaluation gains asprescribed by Cabinet Orderspecified)shall have revaluation gains in theprescribed by Cabinet Orderspecified amount included in gross income for the fiscal year containing the date of those events, notwithstanding paragraph (1).

Three of the four occurrences are outside parentheses, butthe third is inside, creating an exclusion.After referring to “those assets,” the provision delegates which assets are excluded to a Cabinet Order. Each use of that pattern adds a parenthesis.

In constructions placing exclusions in parentheses, delegation wording may increase the distance between connected sentence elements. This is one syntactic example, not proof that delegation causes overall parenthesis density or reading difficulty.

Some provisions have unmatched parentheses

During counting, I encountered something unexpected:provisions with unequal opening and closing parentheses.

Of 496 articles in the Special Tax Measures Act,60 had unmatched counts.Some had eight extra closing parentheses. These were not typos; they came from substitution provisions.

Substitution provisions quote part of another provision and replace it. An excerpt from Article 68-4 of the Special Tax Measures Act follows.

In Article 75-4(1) of the Corporation Tax Act, “including.)” is replaced by “including.), the provisions of Chapter III of the Act on Special Measures Concerning Taxation (Special Provisions of the Corporation Tax Act) (including orders based thereon; the same applies in paragraph (3)), and provisions prescribed by Cabinet Order under Article 68-4 of that Act (Special Provisions for Electronic Filing)”

The four Japanese characters corresponding to “including.)” are themselves the substitution target. Quoting a fragment of another provision meansonly a closing parenthesis appears in the text.Accumulated fragments make parentheses unmatched across the article.

There is another practical consequence:the full stop inside “including.)”.I counted sentences using full stops outside parentheses. That rule also split at the quoted full stop in “including.” Article 68-4, actually one sentence, was counted as nine.

This article's figures were initially corrupted by the same error.After changing to split and count each article separately before summing, the Special Tax Measures Act's sentence count changed from 251 to 5,212—over twentyfold. One change in counting rules can do that.

How to use this when reading

The measurements suggest three things to do before reading a provision.

• Mark “as prescribed by Cabinet Order” first.Each occurrence points to material outside the provision. The Corporation Tax Act averages 0.71 per sentence. It is safer not to assume the article alone resolves the question
• Distinguish delegation expressions inside parentheses from those outside.Those inside often define exclusions or limitations. Skipping them can misstate scope
• In substitution provisions, do not treat quoted full stops as sentence boundaries.“Including.)” is part of the quoted wording, not the end of the sentence

Enactment year alone cannot determine current-text readability.In this table, the Civil Code and Code of Civil Procedure have almost identical parenthesis density.Delegation expressions help identify subordinate legislation to read, but their count cannot measure all difficulty.

Sources and method

All text came from legislative data published through the Digital Agency's e-Gov Laws and Regulations Search. Enactment years came from instrument numbers. Counts cover main provisions, excluding supplementary provisions and schedules. The previous 20 instruments retain the earlier parenthesis counts; the three Reiwa laws were newly counted by the same method.

• e-Gov Laws and Regulations Search Digital Agency
• Corporation Tax Act e-Gov Laws and Regulations Search
• Act on Special Measures Concerning Taxation e-Gov Laws and Regulations Search
• Civil Code e-Gov Laws and Regulations Search
• Penal Code e-Gov Laws and Regulations Search
• Worker Cooperatives Act e-Gov Laws and Regulations Search
• Act on Temporary Special Provisions of National Tax Laws to Address the Effects of COVID-19 and Related Matters e-Gov Laws and Regulations Search
• Act on Vesting Ownership of Land Acquired by Inheritance or Similar Means in the National Treasury e-Gov Laws and Regulations Search
• Previous article Why Is Tax Law So Difficult? Reading Parenthetical Clauses with Color Coding

Sentences were extracted article by article and split at full stops outside parentheses. Opening parentheses, both full-width and half-width, were counted. Delegations count occurrences of wording prescribing matters by Cabinet Order, Ministerial Order, or regulations. Correlation uses Pearson's product-moment coefficient.

Different counting methods produce different numbers, as the twentyfold sentence-count difference illustrates. This article uses counts calculated separately by article and then summed.