Ranking every Charlie Chan film

If what I have concluded from my intrepid detective work is correct, I watched my first Charlie Chan movie, Dead Men Tell (directed by Harry Lachman, 1941), starting at 8 pm on July 25, 1976 on Philadelphia’s now-defunct Channel 48. As you can see, it was the first film in a double feature with a Sherlock Holmes film.

TV lising July 24 1976

I have written extensively (see here, here and here) about my decades-long love for the Charlie Chan film series produced by Fox Film Corporation (later 20th Century Fox Productions). In fact, I am devoting an entire chapter of my book to why I love these films so much. (Teaser: It has a lot to do with my relationship with my own father).

Overall, 47 Charlie Chan films were released between 1926 and 1949, averaging two per year for 24 years. The first two films (The House Without a Key [10-part serial directed by Spencer Gordon Bennet); The Chinese Parrot [Paul Leni, 1927]) were produced by Pathe Exchange and Universal Pictures, respectively; neither survives. In these two films, Chan was portrayed by Japanese actors George Kuwa and Sojin Kamayama, respectively.

The first Fox Film Corporation Charlie Chan film, and the oldest surviving one, is Behind That Curtain (Irving Cummings, 1929). Asian-American actor E. L. Park appears as “Police Inspector Charlie Chan” for a nearly-irrelevant five minutes at the end of this forgettable 91-minute film (it has a lousy 4.9 Internet Movie Database [IMDB] rating).

These three films, nominally derived from three of the first five Charlie Chan novels written by Earl Derr Biggers, were essentially “one-off” releases. Then, on April 12, 1931, Fox released Charlie Chan Carries On (Hamilton MacFadden), starring Swedish-born actor Warner Oland as the titular detective. This was the fourth film to be drawn directly from a Biggers novel.

Oland would portray Charlie Chan in 16 films, of which 12 survive[1]. Of the first five Fox-Oland Charlie Chan films, the only one still in existence is The Black Camel (Hamilton McFadden, 1931), the last of the five original Biggers Charlie Chan novels to be filmed.

However, it was Charlie Chan in London (Eugene Forde), released on September 12, 1934 that heralded the “modern” Fox Charlie Chan film series. Or, at least, this was the first of a cycle of 22 Charlie Chan films that Channel 48 aired every Saturday night (11:30 pm to 1 am) from September 11, 1976 (Charlie Chan at the Circus [Harry Lachman, 19360) to April 26, 1980 (Charlie Chan in Reno [Norman Foster, 1939]), setting my social life back years. Granted, I was still just a few weeks shy of my 10th birthday when Channel 48 shifted their Saturday Charlie Chan screening from 8 pm to 11:30 pm.

After Oland died in 1938, American actor Sidney Toler stepped into the role, playing a sharper, less “scientific” (but more proto-noir, I contend) Charlie Chan for 11 films until 20th Century Fox ceased producing this series following 1942’s Castle in the Desert (Harry Lachman). “Poverty Row” studio Monogram Pictures picked up the series in 1944, beginning with Charlie Chan in the Secret Service (Phil Rosen). Toler starred as Charlie Chan for Monogram in 10 more films before his death in 1947. American actor Roland Winters then assumed the role for six final films before Monogram mercifully ended the series after the April 27, 1949 release of The Sky Dragon (Lesley Selander).

I say “mercifully” because, while the 28 Fox films may not always have been high art, they were demonstrably superior (at least when viewed from a modern vantage point) to the 17 Monogram films.

Using data obtained from IMDB.com, the Monogram films were an average 5.1 minutes shorter (65.4 vs. 70.5) and had a 4.5-point lower average IMDB rating (6.9 vs. 7.4). Rotten Tomatoes audiences were even harsher: an average of just 39.1% liked the Monogram Charlie Chan films, compared to the average 51.6% who liked the Fox Charlie Chan films; these audience members gave the Monogram Charlie Chan films an average rating (1 to 5) of 3.0, lower than the 3.3 average rating they gave the Fox Charlie Chan films. Film critic Leonard Maltin, in his 2003 Film and Video Guide[2], gives the Monogram Charlie Chan films an average of 1.6 stars (1 [BOMB] to 4), much lower than the respectable 2.5 star average he assigns to the Fox Charlie Chan films.

I gathered these data for the Charlie Chan chapter of my book, meaning that I created a Microsoft Excel spreadsheet consisting of the following data points for each Charlie Chan film from Behind that Curtain to The Sky Dragon, where available:

Title, Chan actor, Son/actor, Release Date/Year, Director, Running Time (minutes), IMDB score/# stars[3]/# raters, RottenTomatoes Tomatoemeter[4] freshness score/average rating/# raters, %RottenTomatoes “Like”/average audience rating/number of raters, Maltin stars, Studio, Survive/Not.

Given this range of “quality” metrics, I thought it would be instructive and illuminating to construct a single index that would rank every surviving Charlie Chan film from highest to lowest “quality.”

Just bear with me while I briefly present my methods (feel free to skip the following italicized paragaphs). All analyses were conducted using SPSS 17.0[5].

The first step was to convert “actor” and “studio” into numeric values. I coded Park as 1, Oland as 2, Toler as 3 and Winters as 4; I coded Fox as 1 and Monogram as 2. Statistical purists will note that I am assuming a monotonic linear increase/decrease in quality going from Park to Oland to Toler to Winters. While Park is not superior to Oland (who, arguably, was at least equal to Toler, with both far superior to Winters), since Park he appeared in only one film, I chose to use the single variable[6].

Next, while I had complete data (excluding Tomatoemeter ratings) for 30 Charlie Chan films, I was missing %RottenTomatoes “Liked” and Average Audience Ratings for 11 films[7]. My goal was to create a single “quality” score for all 41 films, so I decided to estimate missing values using ordinary least squares (OLS) linear regression. OLS linear regression estimates the expected value of a dependent variable (Y) given the values of a set of independent variables (Xi), using the formula Y=B0 + B1*X1 + B2*X2 + etc..

The best model (highest adjusted r2, fewest independent variables) proved to be one that only included actor, studio and number of RottenTomatoes audience raters. The estimated %RottenTomatoes “Liked” and Average Audience Ratings yielded by these models differed little from estimates using more complicated models.

I chose to combine these scores using factor analysis. This method uses correlations between variables to group them into a smaller number of dimensions, or “factors.” Factor loadings (correlations between a variable and each factor) can then be used as weights to sum variables into a single score[8].

All that remained was to decide which variables to include in the factor analysis. I began with every numeric variable (actor, studio, running time, IMDB score, IMDB raters, %RottenTomatoes Liked, RottenTomatoes audience rating, RottenTomatoes audience raters, Maltin stars).

Analysis (and reflection) suggested that running time was not truly a measure of “quality,” so I excluded it from the final factor analysis[9]. One factor accounted for 55.9% of the total variance across the remaining eight components.

That is, I used factor analysis to combine eight metrics (actor, studio, IMDB score, IMDB raters, %RottenTomatoes Liked, RottenTomatoes audience rating, RottenTomatoes audience raters, Maltin stars) into a single Charlie Chan film “quality” score. Higher scores indicate higher perceived quality.

The full results can be seen here.

Charlie Chan films, 1929-1949, Ranked

The highest-rated films are Charlie Chan at the Opera (H. Bruce Humberstone, 1936) and Charlie Chan in London, at 1.56 and 1.52, respectively. Opera is notable for two reasons. One, it costars legendary actor Boris Karloff. Two, it features an original opera (Carnival) composed by Oscar Levant especially for the film. London kick-started the Oland series of Fox Charlie Chan films; it is the film being made in 2001’s Gosford Park (Robert Altman).

Other films with score>1.00 are Charlie Chan in Egypt (1.25; Louis King, 1935), Charlie Chan at Treasure Island (1.23; Norman Foster, 1939) and Castle in the Desert. The latter two films are the highest-rated Sidney Toler films. Treasure Island, featuring an excellent performance by Cesar Romero, is my favorite Charlie Chan film, with Charlie Chan in Reno (0.89) a close second. I am not, however, particularly fond of Castle, despite its high ranking. Opera and London are quite good, but my favorite Oland film is Charlie Chan on Broadway (0.76; Eugene Forde, 1937).

And Dead Men Tell, almost certainly the first Charlie Chan film I ever saw, 41 years ago? It ranks 19th overall (with only five Fox Charlie Chan films ranked lower) at 0.41.

The Fox Charlie Chan films are far more highly rated (average score=0.68, 0.74 excluding Behind That Curtain [-0.68]) than the Monogram Charlie Chan films (-0.96)[10]. In fact, with the exception of Curtain, EVERY Fox Charlie Chan film has a higher score than every Monogram Charlie Chan film.

The highest rated Monogram Charlie Chan film, and the only one with a positive score (0.04), is Charlie Chan in the Secret Service. The lowest rated Fox Charlie Chan films, somewhat unfairly I think, are the last Oland Charlie Chan film, Charlie Chan in Monte Carlo (0.09; Eugene Forde, 1938) and the Toler film City in Darkness (0.06; Herbert I. Leeds, 1939. In the latter, veteran character actor Harold Huber replaces Jimmy Chan (Victor Sen Yung) as the Watson-like foil.

The Oland Charlie Chan films are higher-rated (0.78), on average, than the Toler Charlie Chan films (0.01), which are in turn higher-rated than the Winters Charlie Chan films on average (-1.47). However, when you split the Toler Charlie Chan films into those produced by Fox (0.69) and those produced by Monogram (-0.68), you see that there was almost no drop in overall quality when Toler replaced Oland.

Which is good, because I prefer the Fox Toler Charlie Chan films, as a rule, to the Oland Charlie Chan films (I have seen very few of the Monogram Charlie Chan films, to be honest, but the ones I have seen were not impressive). It is not at all clear to me that I would have become as enamored with these films as quickly as I did, that Saturday night in July 1976, had it been an Oland film.

And that would have been a shame.

Until next time…

[1] Not surviving are Charlie Chan Carries On, Charlie Chan’s Chance (John G. Blystone, 1932), Charlie Chan’s Greatest Case (Hamilton MacFadden, 1933) and Charlie Chan’s Courage (Eugene Forde and George Hadden, 1934). However, Carries On was essentially remade as Charlie Chan’s Murder Cruise (Eugene Forde, 1940) and Greatest Case was reworked into Murder Over New York (Harry Lachman, 1940).

[2] Maltin, Leonard ed. Leonard Maltin’s Movie & Video Guide: 2003 Edition New York, NY: New American Library.

[3] I ultimately did nothing with this variable as it was always either one star or not-yet-rated.

[4] The only Charlie Chan film to receive a Tomatoemeter score, along with rating and number of raters is Charlie Chan at the Wax Museum (Lynn Shores, 1940).

[5] SPSS Version 17.0 (SPSS Inc. Released 2008. SPSS Statistics for Windows, Version 17.0. Chicago: SPSS Inc.)

[6] It would probably have been statistically more valid to create three “dummy” variables: Oland (0=No, 1=Yes), Toler (0,1) and Winters (0,1). However, with only 41 surviving films for which I have ANY data, I chose to minimize the number of construct variables.

[7] Behind That Curtain, Charlie Chan at the Race Track (H. Bruce Humberstone, 1936), The Red Dragon (Phil Rosen, 1946), Dark Alibi (Phil Karlson, 1946), Shadows Over Chinatown (Terry O. Morse, 1946), Dangerous Money (Terry O. Morse, 1946), Docks of New Orleans (Derwin Abrahams, 1948), The Shanghai Chest (William Beaudine, 1948), The Golden Eye (William Beaudine, 1948), The Feathered Serpent (William Beaudine, 1948), The Sky Dragon.

[8] This would be the sum of each factor loading multiplied by the z-score (number of standard deviations from the mean) of each variable.

[9] Correlation between scores generated with and without running time=0.997. All factor analyses used principal axis factoring with varimax rotation, forced to one factor.

[10] The difference in means, 1.64, has a 95% confidence interval of 1.28 – 2.00.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s