- Jin-Hyuk Kim, Associate Professor, University of Colorado Boulder
- Yidan Sun, Illinois Institute of Technology
- Liad Wagman, Professor of Economics, Illinois Institute of Technology
This blog article is derived from the authors’ paper titled The Value of Technology Releases in the Apple iOS App Ecosystem, a project of the Economics of Digital Services (EODS) reserch initiative led by Penn’s Center for Technology, Innovation and Competition (CTIC) and The Warren Center for Network & Data Services. CTIC and The Warren Center are grateful to the John S. and James L. Knight Foundation for its generous support of the EODS initiative.
Apple iOS, the operating system that interfaces hardware and software on iPhones, is now installed on over one billion active devices worldwide. With iOS, iPhones are able to handle files and applications, allocate memory space, control hardware features, interface with various smartphone sensors, prevent unauthorized access, and provide other smartphone services. As one of the most vital components of the iPhone, iOS serves as an interface between the users and the hardware and helps manage the smartphone’s hardware and software resources. Software mobile applications (apps), developed by firms or individuals, which can be downloaded on iPhones through the Apple App Store, leverage the iOS operating system to offer various features and functionalities to users such as a game, a map, or an online shopping platform.
One of the major ways through which apps interface with the iOS operating system and with other technology platforms is called Software Developer Kits, or SDKs, comprising ready-to-use codes, libraries, Application Programming Interfaces, and other packages or modules. SDKs, which can be developed and extended by both the operating system operators as well as by third-party technology platforms, enable app developers to interface with the operating system as well as link their apps to external services platforms such as payment, database, advertising, and inventory systems. Thus, SDKs provide critical app infrastructure without which an individual or a small startup may not have the capacity or ability to develop apps.
The smartphone digital ecosystem has been most often studied from the perspective of multi-sided markets where there are smartphone users, apps and app developers, advertisers, and other interested stakeholders on different sides, with cross externalities among these sides. This perspective focuses on the top layer of smartphone markets, often ignoring an additional layer of SDK technologies specifically created to aid app developers in creating apps. More specifically, third-party technology platforms develop and offer SDKs to leverage a smartphone’s software and hardware features; app developers, engineers, and designers come up with innovative product ideas, which can result in new apps being offered to users relatively quickly because of the availability of SDKs. Prior research has paid relatively little attention to the underlying technological drivers that have facilitated the rapid growth of mobile apps.
In this study, we empirically assessed how the releases of iOS SDKs affect the iPhone app ecosystem, including the effects on iPhone sales and new app releases, and estimated the consumer surplus that iPhone apps generate. Our findings quantify SDKs as a substantial driver of new app releases and the consumer surplus that users derive from apps. In particular, we found the effect of SDKs on new app releases to be twice as large as the effect from iPhone sales. More broadly, our findings demonstrate that hardware and software releases, and new product and app provision, positively link with each other in the iOS mobile ecosystem and that the SDK layer plays a crucial role in the iOS digital ecosystem. Hence, from a policy perspective, while much scrutiny in both the European Union and in the United States has recently focused on the rules that app store operators establish for app developers, the technology layer underlying app development—a layer over which operating system designers have considerable influence—merits attention as well to ensure that its dynamics are efficient.
Estimating the impact of the SDK layer on new app releases
In the worldwide mobile operating system market, iOS has a market share of around 28%. In our dataset, obtained from Apptopia, there are approximately 1,300 iOS SDKs that support over 4.9 million iOS apps released over the period 2009-2020. For each app, our dataset comprises estimates (extrapolated by Apptopia based on data from over 125,000 apps that directly share their metrics) of the app’s number of downloads, its daily and monthly active users (which Apptopia uses to calculate the app’s ‘Engagement’ metric as DAU/MAU), the app’s rating and number of reviews, the app’s total revenue from downloads, and its revenue from in-app purchases. We also obtain figures for quarterly iPhone sales directly from Apple’s 10-Q forms. (Apple stopped reporting product unit sales figures in FY2019.)
The vast majority of the SDKs in our data—over 90%—were released by third-party technology platforms rather than by the developers of the operating system (Apple). The functions of SDKs are grouped into 38 categories. As shown in Figure 1, the most commonly offered SDKs are associated with functionalities concerning “DEV_TOOL,” “ANALYTICS,” “GEO_LOCATION,” “DEV_PLATFORM,” and “AD_NETWORK.” By releasing SDKs, technology platforms benefit from apps connecting to and utilizing their platforms, which could in turn result in a broadening of their user bases and/or direct monetary rewards (e.g., from licensing fees—such fees are mostly privately negotiated but some SDKs are open source).
Figure 1: iOS SDK Counts by Function
We first examined the effects of SDK releases by using a platform-level time-series data from 2009:Q1 to 2020:Q1. Figure 2 shows the trends of quarterly iPhone sales, SDK releases, and new app entries over the 2009-2020 period.
Figure 2: Time series data for iOS (in log scale)
To evaluate the macro effect from the release of SDKs, we used a Structural Vector Autoregression (SVAR) model—one of the most widely-used models in macroeconomics—to identity the association between market dynamics and various shocks or changes at the aggregate level. The variables we used, comprising smartphone sales, SDK releases, and new app entries, as well as their lagged counterparts, form a system of equations (e.g., with lag 1 below):
In the above system, the
are 3x3 matrices of parameters, and the
are structural shocks. Our identifying assumption is that the release of apps and SDKs can respond to a sudden change in the number of iPhone sales within a quarter but not the reverse. This is because iPhone sales are often driven by new model releases, and app developers and tech platforms can plan on releasing apps and SDKs based on the iPhone’s hardware update schedule. In addition, we assumed that the release of apps can respond to a sudden change in the release of SDKs within a quarter, not the reverse. This is because app developers can swiftly pick up newly released SDKs for their app development. We imposed no restrictions how a change in one variable can affect the other variables over time.
We estimated the above system of equations using four lags, as is common in the literature. The results can be intuitively presented using so-called “impulse response functions,” demonstrating the effects of a change in one variable on the other variables over time. Figure 3 shows how quarterly app releases are influenced by (orthogonalized) structural shocks to the three variables. That is, a positive shock to iPhone sales has a delayed effect on app releases with its effect becoming significant after four quarters, lasting several quarters. A positive shock to SDK releases has a relatively quick, positive effect on app releases, lasting for multiple quarters. And finally, a positive shock to new iOS app releases additionally increases app releases (e.g., due to scale efficiencies, due to replication in other geographic markets, due to competitors entering, etc.) in the next four quarters, with its effect diminishing over time. Hence, the impulse response graphs indicate how the shocks to each variable impact the development of iOS apps.
Figure 3: Impulse Response Function (IRF) with quarterly % changes
Applying a methodology called Forecast Error Variance Decomposition, we can estimate how much each structural shock is responsible for the variation in the number of app releases, which is depicted in Figure 4. The results indicate that after about eight quarters, the relative contribution of shocks in each of the variables on new app releases is stabilized. Specifically, app releases themselves can explain about 41% of the variation in further app releases, iPhone sales explain about 17%, and SDK releases explain about 42% of app releases. It thus follows that the traditional force from the perspective of multi-sided markets—the effect of iPhone sales on new app releases—explains less than half of the variation in app releases in comparison to the effect from SDK releases.
Figure 4: Error Variance Decomposition (EVD) with quarterly % changes
Estimating consumer surplus from apps
Next, in order to evaluate the economic value to consumers (‘consumer surplus’) of iOS apps, we used the price of apps, defined as the Average Revenue Per User, or ARPU, and worldwide download data on a given day (e.g. June 1st, 2020, as shown in Figure 5), to estimate the demand function for iOS apps. We then repeated this process to other dates. The ARPU is calculated by dividing the sum of the total revenues from app downloads and in-app purchases by the total number of downloads. It is thus important to keep in mind that transactions outside the Apple App Store (such as e-commerce billings) are not counted in this measure. In that sense, our estimate of consumer surplus aims to quantify the surplus consumers derive directly from apps rather than from external transactions.
Figure 5: Scatter plot for iOS apps (June 1, 2020)
When estimating the iOS app demand function, we used the number of app ratings to instrument for the number of downloads, given potential concerns about endogeneity. We estimate the following log-linear inverse demand function:
where Engagement is a ratio between daily active and monthly active users and serves as a proxy for app quality. We estimated the demand curve separately for game apps and non-game apps. The consumer surplus for each app can then be calculated by integrating the estimated demand curve over the range of its observed number of downloads and price, incorporating the estimated coefficients and the app’s engagement metric.
Table 1 showcases results for daily worldwide consumer surplus from iOS Apple App Store apps over the first week of June 2020. The daily worldwide consumer surplus exceeds $60 million with the numbers relatively stable across days. Given that iPhone sales account for about 17% of new app releases, our estimates indicate that about $10 million of the daily consumer surplus is the indirect result of iPhone sales, whereas SDK releases account for about $25 million.
|Date (2020)||June 1||June 2||June 3||June 4||June 5||June 6||June 7|
|No of Apps||523,803||526,692||528,230||530,072||529804||527,880||524,883|
Table 1: Worldwide daily consumer surplus for iOS App Store apps on several days in 6/2020
Our findings thus demonstrate that hardware and software releases, and new product and app provision, positively link with each other in the iOS mobile ecosystem. In particular, iOS apps, which generate substantial value for users, are most significantly associated with iOS SDK releases. While much regulatory scrutiny has recently been focused on the rules that app stores establish for app developers, much less attention has been given to the technology layer underlying app development—a layer over which operating system designers and third-party platforms have influence. Our findings suggest that this technology layer plays a crucial role in app development and thus in value generation for consumers.