Advances and Declines

About Advances and Declines

This site gathers advance and decline data for the NYSE, AMEX, and NASDAQ stock exchanges. We gather data from various publicly-available sources and report the median values found, and make available historical data in comma-delimited text format.

Why don't the different sources agree?

There doesn't seem to be a standard way to define whether an issue is advancing, declining, or unchanged.

For example, if a stock starts out at $100 and ends the day at $100.01, does such an insignificant change warrant counting it an "advancing" issue like a stock that went from $100 to $102? Or do we consider it effectively "unchanged"?

We suspect that the different data sources may have different criteria for what defines an "unchanged issue". And of course, the number of unchanged issues will affect the number of advancing and declining issues.

For that matter, what is an "advancing issue"? There are a couple possibilities:

• A stock closed higher today (exceeding the "unchanged" threshold) than yesterday's close.
• The stock closed higher today than it opened today.

Either of those definitions will result in different numbers of advancing issues.

The universe of stocks counted by each data source may also be different, causing different results. For example, sources may or may not count low-priced stocks, or stocks that fall below some minimum capitalization threshold, or ETFs or REITs. A data source might also not monitor all stocks offered by each exchange.

So what do we do about it?

One thing we can do is choose one source and stick with it. However, doing this has a higher risk of exposure to errors than monitoring several sources. Sometimes negative volumes or other erroneous values are reported in one source but not others.

Another alternative is to average together the sources. A prior incarnation of this web site did this. Averages, however, are skewed by data errors.

For the purposes of this web site, we assume that the median value of advancers, decliners, etc. from all available sources is a fair representation of the actual market data. Unlike averages, medians are insensitive to outliers and errors. Not all sources are available every day, but enough should be available each day to get a good idea of the median values.

Potential problem

The distributions of data are often bimodal, which render the median values somewhat questionable. You can see this on the home page; view the analysis (click "Show") and scroll down to the bottom where the sorted lists of all values are displayed.

Observe, for example, the NYSE advancing and declining volumes. You'll see two groups of values: smaller and larger, with the larger values being 5 to 10 times greater than the smaller ones. If you consider that the Wall Street Journal is likely a reliable source for NYSE data, you'll agree that the smaller values more closely match what the WSJ reports.

A problem arises here because a median value of a bimodal distribution is essentially a report on the majority consensus, and we may not always get data reflecting that consensus.

Fortunately, the majority of the NYSE advancing and declining volumes we obtain are the smaller values, presumably more correct than the larger values if one accepts WSJ numbers. Unfortunately, some of the sources reporting these "more correct" values may become unavailable (which happens occasionally). If the consensus then becomes evenly split, the median value becomes the average of a smaller and larger value. If the larger values become the majority, then the median will select from those values.

The more data sources we can get, the better. If you find any other besides what we're already getting, please suggest it at the address given at the bottom of this page.

Commentary on the data sources

If we assume that the WSJ reports "correct" data, we make the following observations about the data sources this site uses.

Yahoo

Volumes, new highs, and new lows seem consistently wrong for NYSE and AMEX, OK for NASDAQ. All other data looks OK. The page always displays the current date and time, not the date associated with the data.

Data no longer available from this source.

BigCharts

All data OK except for volumes, which are occasionally reported as negative numbers, so it's hard to trust when they aren't negative.

eTrade

Pretty good data source, although new highs and new lows aren't reported. No longer used here, requires customer login.

Data no longer available from this source.

DailyStocks

The numbers reported here used to be obtained from barchart.com, with different numbers from Yahoo's but exhibiting the same problems with volume data, new highs, and new lows. They now appear to use data from NASDAQ directly, which we already get.

Data no longer available from this source.

Financial_Post

Seems to be OK for everything except NASDAQ data, for which unrealistically low values are reported.

Data no longer available from this source.

MarketVolume

Seems to report unrealistic unchanged issues for all markets, and exhibits similar problems as Yahoo with volume values for NYSE and AMEX. This source often reports values at the extreme end of the range of values reported by other sources. We have dropped this source for that reason, about the same time their data became subscription-only.

Data no longer available from this source.

NASDAQ

This is the official NASDAQ source, so it's likely reliable for NASDAQ data. It seems to report reasonable values for NYSE and AMEX also.

SmithBarney

If you round the Wall Street Journal values to the nearest 1000, you get results identical to SmithBarney's. Assuming the WSJ reports "correct" values, we count SmithBarney twice to bias the median somewhat. Unfortunately, although data for Monday through Thursday are posted in the afternoon, SmithBarney sometimes waits until Monday to post Friday's data.

Data no longer available from this source.

Wall_Street_Journal

This source is what we consider most correct, and, like SmithBarney, rarely (if ever) shows outlier values.

A possible improvement to this web site is to score the sources based on how often they report outlier values, and weight the ones that report the most outliers less than the others. This isn't planned; just an idea.