Adjusted vs Unadjusted Stock Prices: Which Series Answers Your Question?

Before choosing adjusted or unadjusted prices, decide what you are trying to measure. Reconstructing a historical quoted price, comparing price behavior across a split, and studying an investment return are different tasks. A column labeled close cannot answer all three without additional definitions. The most useful first step is to read the provider's explanation of each field and write down the intended use in your own research notes.

Alpha Vantage's documentation distinguishes raw daily fields from adjusted closing values and split and dividend events. That is a specific example of why field-level documentation matters; other datasets may use different conventions. The inspection workflow below is an original research framework, not a validated investment strategy. Its purpose is to help you select coherent inputs and describe the resulting calculation honestly.

Primary-source context: Alpha Vantage: Daily and Daily Adjusted Data Documentation.

Imagine a fictional share trading at 100 currency units immediately before a two-for-one split. Ignoring any separate market movement, the post-split price would be 50 and a holder of one share would have two. The combined value remains 100 in this simplified example. A raw price comparison of 50 with 100 shows a 50% decline, but that calculation ignores the changed number of shares.

A split-adjusted historical series can restate the earlier price on the later share basis. In the example, comparing an adjusted earlier value of 50 with 50 avoids representing the split itself as a loss. This is explanatory arithmetic, not a claim about how every vendor implements adjustments. Record the adjustment convention and verify how related fields are treated before applying formulas across the event.

Now consider an invented one-period example with a starting price of 50, an ending price of 49, and a cash distribution of 1 per share. Ignoring taxes, fees, reinvestment, and all other effects, the end value plus cash equals the starting value. The price-only change is negative 2%, while the simple holding-period return including the distribution is zero. Real analyses need explicit assumptions about dates and cash treatment.

Do not assume that an adjusted closing field alone makes every other field compatible. If you divide a raw daily high by an adjusted close from the same row, a corporate-action adjustment may make the ratio unsuitable for your intended question. Likewise, do not add distributions again to a return series that already incorporates them. State whether the output represents price change, a distribution-inclusive calculation, or another provider-defined measure.

Suppose a spreadsheet compares a fictional stock's raw historical high of 102 before a split with a later adjusted close of 51. The researcher describes the current price as half its previous high. That conclusion mixes share bases. Under the simplified two-for-one example, the comparable earlier high would be 51, so the supposed gap disappears before considering any actual price movement.

The repair begins by identifying the research question: distance from a comparable historical high. The researcher obtains consistently adjusted high and close fields, or performs a documented transformation using verified event information. If only adjusted closes are available, the researcher changes the question to a close-based comparison instead of inventing adjusted highs.

The notebook preserves the original mistake, the chosen convention, and the transformation date. It also states that correcting this error does not validate any strategy based on historical highs. It merely makes the comparison internally coherent under the stated assumptions.

Record the provider, field names, currency, session coverage, timezone, adjustment setting, and download date. Identify how splits and distributions are represented. Check whether open, high, low, close, and volume use compatible conventions for the proposed formula. Keep the event records needed to explain conspicuous discontinuities rather than assuming every unusual move is a market event.

Choose one corporate-action example and reconcile it manually. Specify whether your calculation includes cash distributions and whether reinvestment is assumed. Avoid adding an event twice through both an adjusted series and a separate cash-flow calculation. Preserve raw inputs when permitted so transformations remain inspectable.

If a historical download changes, compare its adjustment metadata before concluding that the vendor made an error. Revised event information or a different requested convention may explain the difference. Document the unresolved case instead of quietly substituting another series.

Consider a hypothetical investor holding ten shares priced at 80 each. Initial wealth is 800 currency units. A two-for-one split changes the holding to twenty shares. Assume the next observed raw price is 41 and a subsequent cash distribution is 0.50 per post-split share. Finally, the raw price is 40.50. With no transactions, reinvestment, taxes, or fees, the investor finishes with shares worth 810 and cash of 10. Ending wealth is 820, so the holding-period gain is 20 divided by 800, or 2.5%.

A chart calculation needs to describe the same economic question before it can be compared with this ledger. On a split-only basis, the initial price becomes 40. Comparing 40.50 with 40 gives a price gain of 1.25%. Including the 0.50 distribution on that same share basis produces 2.5%. The apparent disagreement between those percentages is explained by cash treatment. It does not require either calculation to be numerically wrong. Notice also that the intermediate price of 41 does not determine the final holding-period return.

Use this reconciliation as a worksheet with separate columns for date, raw price, share count, cash received, cash balance, and total wealth. Add a second block for the transformed series, its adjustment factor, and the return definition. Compare ending wealth only after matching assumptions. If an adjusted series assumes reinvestment, reproduce that reinvestment convention in the ledger before expecting agreement. Do not force the figures to match by inserting an unexplained balancing entry. A residual is a question about conventions, timing, or input quality that deserves an explicit answer before the comparison is used elsewhere.

An original inspection exercise starts with a hypothetical historical row: raw open 98, high 104, low 96, and close 100. Suppose a documented split-only transformation applies a factor of 0.5 to every price in that row. The comparable transformed values are 49, 52, 48, and 50. The raw intraday range divided by the raw close is 8 divided by 100, or 8%. The consistently transformed calculation is 4 divided by 50, also 8%. This equality is an arithmetic property of applying the same positive factor to numerator and denominator.

Mixing the raw range with the transformed close instead gives 16%. That number may look plausible enough to escape visual inspection, which is why field compatibility matters even when a chart appears smooth. The exercise suggests a useful diagnostic: identify formulas that should remain unchanged under a common scaling of all their price inputs, then check a manually specified event row. Passing this diagnostic demonstrates only the relevant algebra. It cannot certify the provider's event record or an entire historical file.

Volume requires a separate definition. In a deliberately simplified split-only representation, restating ten old shares as twenty new shares while halving their associated price preserves the product of shares and price. A dividend adjustment is not a change in share count, so that split arithmetic does not justify applying an inverse price factor to volume for every kind of adjustment. Record exactly what the volume field measures before combining it with transformed prices. If its convention is unknown, withhold the derived measure instead of inferring compatibility from the fact that the columns arrived in the same download.

A single split example is easy to inspect; a longer horizon needs an event sequence. Imagine a hypothetical share starting at 120, undergoing a three-for-one split, and later a two-for-one split. On the final share basis, one original share corresponds to six shares. A starting price restated solely for those splits is therefore 20. If the final raw price is 22, the price-only gain is 10%. The original share's final holding value is six times 22, or 132, which reconciles to the same percentage before other effects.

Now place a distribution between those splits. Assume each of the three intermediate shares receives 2 in cash. Total cash received by the original holder is 6. On the final six-share basis, that cash corresponds to 1 per final share, not 2. Without reinvestment, ending wealth is 132 plus 6, or 138, and the complete holding-period gain is 15%. This invented sequence illustrates why distribution units must be attached to their event date and share basis. Copying every historical cash amount unchanged into a final-basis calculation can quietly misstate the return.

Build the sequence chronologically before transforming it. For each event, write the share count immediately before and after, the distribution entitlement assumed, and the basis of every quoted amount. Keep unresolved entitlement or effective-date details separate from confirmed arithmetic. The worksheet should also identify whether fractional holdings are allowed in the illustration. A simplified fractional-share assumption can make an educational ledger coherent, but it should not be presented as a reconstruction of an account whose actual cash-in-lieu treatment has not been checked.

Suppose a hypothetical research note asks which observations exceeded a quoted price of 100 at the time. That is a question about historical nominal quotations. Rewriting all prices onto a later share basis changes the question: a pre-split quote of 120 might become 60 after a two-for-one transformation. The transformed value does not mean the original quoted-price condition was unmet. Preserve the raw quote and contemporaneous condition if the objective is to reconstruct what that screen actually displayed.

A different note might ask whether a share exceeded its prior comparable closing high. Here a consistent adjustment basis may be necessary to avoid making the split itself determine the answer. Write these as two separate research questions rather than trying to make one column serve both. The same distinction applies to labels on a chart. A historical adjusted value can support a comparison without being represented as an actual transaction price from that date. The caption should make that meaning accessible to a reader who never sees the data dictionary.

For a reusable threshold worksheet, record the rule in words, the numerical boundary, the units, the historical date, and whether the condition refers to quoted prices or comparable price behavior. Add the input fields and adjustment convention used to evaluate it. Then construct one split boundary case and predict its treatment before computing the result. If the answer changes merely because a later download uses a different adjustment anchor, decide whether that dependence is appropriate for the question. An unexplained change should suspend the interpretation, even when the resulting screen looks more attractive or produces a cleaner historical chart.

Consider two hypothetical downloads of the same date range. In the first, an earlier adjusted close is 50; in the second, it is 49.50. The raw close remains 100. The implied ratios are 0.5 and 0.495. This observation locates the disagreement in the relationship between raw and adjusted values, but it does not establish the reason. A revised event, changed request setting, or changed convention remains a candidate explanation until supported by the relevant records. Do not pick the download that makes the research outcome look strongest.

Preserve a compact comparison record containing request parameters, retrieval times, the two raw values, the two adjusted values, and associated event metadata. Next identify the affected date interval and calculations. If all prices in a particular comparison window changed by the same factor, a ratio within that window may be unchanged. A comparison spanning differently affected dates may move. This distinction helps explain which conclusions require recalculation rather than assuming either that every result is invalid or that the revision is cosmetic.

Set completion criteria around explanation rather than around an arbitrary small percentage difference. For an educational close-based chart, a fully documented convention change may be acceptable after relabeling. For a reconstruction of past information, lack of the original adjustment version may remain a substantive limitation. Mark that limitation in the research record. The useful final answer states what changed, which outputs depend on it, and which comparisons remain unresolved. It does not claim that a smooth replacement series establishes the truth of the earlier study or that numerical agreement alone settles the meaning of the data.

Adjusted data are not automatically suitable for simulated execution. A retrospectively adjusted historical value is not necessarily a price at which a transaction could have occurred at that time. Nor does an adjusted series prove that the dataset includes delisted securities, uses historically available information, or correctly represents every corporate action relevant to your analysis.

Unadjusted data are not inherently bad either; they can be appropriate for questions about historical quoted prices when event effects are handled separately. The choice follows the question. A sound research note states the convention, explains the calculation, and identifies what the output leaves out. That is a more useful result than calling one column correct and another wrong without specifying the task they are meant to serve.